Epic Systems Reaches $5.7B Revenue and $50B Value via Trust Structure

Acquired

Epic Systems dominated the healthcare software market by maintaining a unified database architecture and a strict private trust that prevents public offerings or corporate acquisitions.

AI-assisted summary of Epic Systems (MyChart) — original source. Check the original for full context.

Source published: 2025-04-21

Detailed notes

## Background Published on YouTube by the channel Acquired on April 21, 2025, this episode covers Epic Systems and its online patient portal, MyChart. The description highlights how this vital healthcare giant is run by its wealthy founder, Judy Faulkner, from a whimsical farm in rural Wisconsin under a strict trust structure ensuring it remains private.

## TL;DR The episode explores Epic Systems, a bootstrapped enterprise software titan that acts as the central nervous system for the American healthcare industry. It chronicles founder Judy Faulkner's journey from a single computer program to winning massive contracts like Kaiser Permanente through a unified database architecture called Chronicles. The hosts also examine how federal legislation like the 2009 HITECH Act accelerated industry digitization, Epic's fierce operational strategy involving zero acquisitions, and its ironclad succession plan designed to keep the company independent forever.

## Atmosphere & style The podcast features an analytically deep yet highly engaging and narrative-driven storytelling style between hosts Ben and David. It shifts from an awe-inspired fascination with Epic's eccentric, themed Verona campus to a calculated, rigorous breakdown of healthcare enterprise software economics. The discussion feels collaborative and deeply researched, carrying a tone of admiration for Faulkner's 'wacky brilliance' and uncompromising business discipline.

## Key takeaways - [00:00](https://www.youtube.com/watch?v=FNRU9ZNqcls&t=0s) Epic Systems is a highly resilient, 47-year-old private medical software company based in rural Wisconsin with near-perfect customer retention and an estimated annual revenue closing in on $6 billion. - [00:00](https://www.youtube.com/watch?v=FNRU9ZNqcls&t=0s) The company operates under a unique trust structure created by founder Judy Faulkner, ensuring it will never go public, never be acquired, and always remain an independent entity. - [33:32](https://www.youtube.com/watch?v=FNRU9ZNqcls&t=2012s) Epic relies on Chronicles, a centralized, single-core database engine built on a descendant of MUMPS, giving it a massive architectural advantage of native data integration over competitors. - [39:55](https://www.youtube.com/watch?v=FNRU9ZNqcls&t=2395s) Epic bootstrapped its entire global expansion using only a $70,000 bank loan and $70,000 in early angel investments, completely avoiding outside venture capital. - [46:05](https://www.youtube.com/watch?v=FNRU9ZNqcls&t=2765s) Unlike traditional software giants, Epic employs virtually no formal sales or marketing departments, leveraging organic upmarket demand and top-down CIO enterprise relationships instead. - [01:01:14](https://www.youtube.com/watch?v=FNRU9ZNqcls&t=3674s) The launch of MyChart in 2000 pioneered web-based, consumer-facing patient portals within highly restricted, HIPAA-regulated environments. - [01:10:24](https://www.youtube.com/watch?v=FNRU9ZNqcls&t=4224s) Securing the massive Kaiser Permanente contract in 2003 validated Epic's unified database scaling capacity and established it as the premier gold standard for enterprise health networks. - [01:33:13](https://www.youtube.com/watch?v=FNRU9ZNqcls&t=5593s) The company enforces a unique corporate culture combining an eccentric, highly whimsical physical campus with a demanding, hyper-systematized 'up-or-out' engineering execution engine. - [01:53:17](https://www.youtube.com/watch?v=FNRU9ZNqcls&t=6797s) Epic leverages a dedicated customer account methodology where every client is assigned a 'BFF' (Best Friend Forever) who issues an annual report card performance grade. - [02:12:24](https://www.youtube.com/watch?v=FNRU9ZNqcls&t=7944s) The federal 2009 HITECH Act injected over $27 billion into health IT adoption, acting as a massive regulatory tailwind that consolidated the market in favor of premium, compliant suites like Epic. - [02:40:40](https://www.youtube.com/watch?v=FNRU9ZNqcls&t=9640s) Epic deliberately avoided massive DoD and VA government contracts, leaving main rival Cerner to become bogged down in multi-billion dollar federal implementation delays. - [02:55:40](https://www.youtube.com/watch?v=FNRU9ZNqcls&t=10540s) The company's massive market footprint includes training over 90% of US medical students and housing anonymized data for 295 million patients via its free collaborative database, Cosmos. - [03:05:52](https://www.youtube.com/watch?v=FNRU9ZNqcls&t=11152s) The hosts estimate Epic's true enterprise value at upwards of $100 billion, which would make founder Judy Faulkner the wealthiest self-made woman in history. - [03:33:57](https://www.youtube.com/watch?v=FNRU9ZNqcls&t=12837s) Epic aggressively buffers its core market from micro-point solutions by deploying the Microsoft playbook of bundling newly engineered modular additions into the base license for free. - [03:38:05](https://www.youtube.com/watch?v=FNRU9ZNqcls&t=13085s) Epic acts as a critical choke point and integration partner for emerging ambient AI scribes that aim to automate data entry and resolve massive administrative burnout among clinicians. - [03:44:07](https://www.youtube.com/watch?v=FNRU9ZNqcls&t=13447s) Epic represents the ultimate success story in Vertical Market Software (VMS), succeeding because it focused deeply on a highly complex industry with zero tolerance for operational risk.

## Detailed walkthrough ### [00:00](https://www.youtube.com/watch?v=FNRU9ZNqcls&t=0s) Introduction to a Healthcare Behemoth

The spring 2025 season of *Acquired* kicks off with hosts Ben Gilbert and David Rosenthal diving into Epic Systems, a quietly dominant medical software company based in rural Wisconsin. While most people only recognize Epic through its consumer-facing patient portal, MyChart, the company plays an unseen, massive role across American healthcare. Epic is described by the hosts as an extraordinarily unusual enterprise. In its 47-year history, the company has famously never raised venture capital, never made an acquisition, and remarkably, employs virtually no sales or marketing teams. They don't negotiate prices, they don't offer discounts, and they regularly reject prospective customers who approach them. Furthermore, the company completely rejects remote work; its 14,000-plus employees work entirely in person on a sprawling, eccentric campus designed with buildings resembling the Land of Oz, a Wizards Academy, and a treehouse, alongside an 11,000-seat underground auditorium.

The most staggering statistic shared by the hosts is Epic's near-perfect customer retention rate. Serving the vast majority of major U.S. hospital systems, Epic has famously never lost a single one of its 600-plus customers—with one technical exception where a customer left for a mere six months before returning. Today, the private company is closing in on $6 billion in annual revenue. At 81 years old, founder Judith Faulkner remains at the helm. She has established a strict trust structure for her voting shares to ensure that her stated goal remains intact forever: Epic will never go public and will never be acquired, despite decades of interest from giants like GE, Microsoft, and Google. Gilbert posits that Faulkner is "the most successful female entrepreneur in history," a claim the hosts promise to unpack further by the end of the episode.

- **Revenue & Scale:** Nearing $6 billion in revenue with over 14,000 employees. - **Ownership:** Privately held; Faulkner and her family foundation own roughly half the company. - **U.S. Healthcare Cost:** Healthcare spending has ballooned to consume 18% of U.S. GDP, driven by runaway costs and administrative waste.

[03:46](https://www.youtube.com/watch?v=FNRU9ZNqcls&t=226s) Housekeeping and Upcoming Live Event

Before transitioning into the historical narrative, the hosts share several community updates and show announcements. They encourage listeners to join the *Acquired* email list at `acquired.fm/email` to receive episode hints, corrections, and behind-the-scenes trivia. They also invite the audience to join their active Slack community, noting that a significant portion of the medical ecosystem hangs out there. For listeners seeking more content between the main monthly episodes, they plug their interview spin-off show, *ACQ2*.

The hosts also reveal a highly anticipated "save the date" for a live event in New York City on July 15th, partnered with J.P. Morgan Payments. Prompted by years of intense listener demand, Gilbert promises it will be "a night of absurdity." They close the introductory segment by thanking their presenting partner, J.P. Morgan Payments, and issuing their standard disclaimer that the podcast is strictly for informational and entertainment purposes and does not constitute investment advice.

[05:25](https://www.youtube.com/watch?v=FNRU9ZNqcls&t=325s) The Genius and DNA of Judy Faulkner

The biographical narrative begins in August 1943 with the birth of Judy Greenfield (later Faulkner) in Earlton, New Jersey, a suburb of Philadelphia. The hosts jokingly note that the Delaware River area was quite a fertile ground for entrepreneurship, spawning not only corporations like DuPont but also contemporaries like Taylor Swift. The timing of her birth was auspicious; just four years later and slightly up the road in Murray Hill, Bell Labs scientists invented the transistor, laying the physical foundation for the software industry. Interestingly, the earliest medical records pioneers were actually massive hardware and defense companies like Lockheed and Siemens rather than dedicated software firms.

Judy’s early life was shaped by an eccentric and highly driven family. Her father, Lou, was a small-town entrepreneur who ran a local pharmacy and soda fountain called "Lou's Soda Fountain." Her mother, Dell Greenfield, was described as an "absolute freaking dynamo." Dell graduated high school at age 15, worked as a secretary, helped run the pharmacy, and later became a deeply passionate peace advocate during the Vietnam War, eventually directing the South Jersey Peace Center. Later in life, Dell moved to Portland, Oregon, and became the executive director of Oregon Physicians for Social Responsibility. In a spectacular piece of historical trivia uncovered by Rosenthal via old obituaries and confirmed by the company, Dell's organization partnered internationally to win a share of the Nobel Peace Prize in 1985. This lineage gifted Judy a potent mix of entrepreneurial grit and divergent, mission-driven thinking.

- **Birthdate & Location:** Born August 1943 in Earlton (Cherry Hill), New Jersey. - **Father's Business:** Lou's Soda Fountain and pharmacy. - **Mother's Achievement:** Dell Greenfield shared in the 1985 Nobel Peace Prize through her leadership in anti-nuclear physician advocacy groups.

[08:44](https://www.youtube.com/watch?v=FNRU9ZNqcls&t=524s) From Math to the Genesis of Computer Science

Growing up, Judy's primary obsession was mathematics. She frequently shares a formative story from the seventh grade when a teacher introduced a number theory puzzle, asking why any number whose digits add up to a multiple of three is itself divisible by three. The elegance of that mathematical law hooked her instantly, cementing her conviction that her future lay in mathematics. She graduated high school in 1961 and attended Dickinson College as a math major. During a pivotal summer job at the University of Rochester's particle physics lab, she was handed a manual and told to learn Fortran to assist with running experiments. Judy taught herself the language in a single week, quickly becoming the best programmer in the laboratory.

The hosts emphasize that Judy was entering computer programming at a time when software engineering did not exist as a formal discipline. Universities were only just beginning to conceptualize computer science departments, meaning Judy was explicitly intertwined with the very birth of the field. She fell completely in love with the medium, later reflecting that programming felt "like a kid playing with clay." For Judy, it represents an intoxicating intersection of cold mathematics, linguistic structure, and artistic expression. Rosenthal notes that her background, creative streak, and entrepreneurial timeline give off distinct "Bill Gates vibes," while Gilbert chimes in that her peace-oriented, slightly counter-culture upbringing carries echoes of Steve Jobs.

[11:36](https://www.youtube.com/watch?v=FNRU9ZNqcls&t=696s) The Wisconsin "Sliding Door" Moment

Upon returning to Dickinson College, Judy applied to five PhD programs in mathematics, gaining admission to all of them. However, because she had listed her Fortran experience on her CV, two of the universities—Stanford and the University of Wisconsin—had just unilaterally shifted her application over to their brand-new, experimental computer science departments. Judy accepted the offer from the University of Wisconsin and moved to Madison. The hosts flag this as a massive "sliding door moment" in tech history. Had she chosen Stanford, she likely would have been institutionalized by the distinct Silicon Valley ethos regarding venture capital, equity, and corporate scaling, resulting in a completely different trajectory.

While at Wisconsin, Judy enrolled in a pioneering course called "Computers in Medicine," taught by Dr. Warner Slack of the Wisconsin Medical School. It was arguably the first class of its kind in the world, arriving just 15 to 20 years after the invention of ENIAC, a period when mainframes were finally escaping military and government exclusivity. Judy naturally became the star student. Recognizing her talent, Dr. Slack hired her for $5 an hour to write a computer program to optimize on-call scheduling for hospital physicians. Because she was a hyper-efficient genius who wrote code incredibly fast, she joked that she actually made very little money trading time for dollars, even after they doubled her pay to $10 an hour.

As she bounced around different medical departments—from psychiatry and OB/GYN to inpatient care and the ICU—doctors began approaching her with a unified, critical complaint. They were treating patients who were simultaneously seeing other specialists within the same hospital, but there was absolutely no infrastructure to share patient data across departments. The clinicians explicitly told her they needed a single, longitudinal database that could track a patient's entire medical history across the University of Wisconsin Medical Center. This precise academic frustration in the mid-1960s serves as the exact origin point for Epic and the electronic health record (EHR), moving away from an incredibly disorganized, paper-based tracking system that hospitals had failed to successfully standardize since initial attempts at Mass General in 1912.

[16:29](https://www.youtube.com/watch?v=FNRU9ZNqcls&t=989s) The Early Roots of Medical Standardization

The historical push for standardized medical documentation in the United States traces back to Massachusetts General Hospital (MGH) in Boston, the largest teaching hospital of Harvard Medical School. In 1912, representatives from the American Medical Association and the American Hospital Association gathered at MGH to establish uniform entry practices for physicians' "patient diaries." It took seven years of collaboration before they officially introduced a standardized "treatment diary" for national distribution in 1919.

However, getting physicians to adopt these templates proved incredibly difficult. Doctors are described as a "notoriously independent-minded group" who resisted being told how to design their clinical notes. The speakers note that this resistance made sense given the prestige, intensive education, and high compensation associated with the profession. Furthermore, medicine at the time—as it remains today—was viewed as "equal parts art and science." For a highly trained physician, being told how to standardize patient interactions felt anathema to their professional autonomy.

- **1912:** First meeting of the American Medical Association and American Hospital Association at MGH to discuss standardizing physician notes. - **1919:** Introduction of the first standardized "treatment diary" for national distribution.

[18:10](https://www.youtube.com/watch?v=FNRU9ZNqcls&t=1090s) Policy, Pre-Tax Benefits, and the Rise of Employer-Sponsored Care

The standardization movement gained institutional backing in 1928 with the creation of the Association of Health Record Librarians of North America, an official body funded by the Rockefeller Foundation to oversee the collection, storage, and retrieval of patient data. The speakers connect this back to their past episodes on Standard Oil, noting that the Rockefeller Foundation was a primary early funder of American medical schools. Despite these institutional efforts, paper records remained a bottleneck until the 1960s; reams of paper still had to be manually transported between hospital departments, making true data portability impossible.

``` [1942: Stabilization Act] ──> [Wage Controls / 10% Insured] ──> [Insurance Offered as Bonus] │ [1946: 30% Insured] <── [Tax-Deductible & Pre-Tax Status Approved] <──┘ ```

The landscape shifted permanently due to federal policy during World War II. In 1942, Congress passed the Stabilization Act, which imposed strict wage and price controls. At this time, only 10% of Americans had health insurance, and the prevailing healthcare model was simple fee-for-service. To bypass wage caps and attract top talent, employers began offering health insurance as a fringe benefit. Labor unions quickly embraced this strategy to maximize total worker compensation, lobbying the National War Labor Board to explicitly legitimize the practice.

Subsequent lobbying by employers and emerging insurance providers successfully convinced the government to make health insurance premiums tax-deductible for corporations and non-taxable as income for employees. This regulatory pivot established the modern system of pre-tax payroll deductions. The impact was immediate and dramatic: the percentage of insured Americans climbed to 30% by the end of the war in 1946, and reached 80% by 1964, firmly embedding employer-sponsored health insurance as the baseline American model.

- **1928:** The Association of Health Record Librarians of North America is established with funding from the Rockefeller Foundation. - **1942:** The Stabilization Act introduces wartime wage controls, unintentionally sparking the employer-sponsored health insurance model. - **Growth of Health Insurance:** Insured Americans rose from 10% (1942) to 30% (1946), and reached 80% by 1964.

[23:40](https://www.youtube.com/watch?v=FNRU9ZNqcls&t=1420s) The Social Safety Net and the "Faustian Bargain" of Billing

By the mid-1960s, the dominant employer-sponsored framework left a glaring coverage gap for individuals outside the traditional workforce: the elderly and the low-income populations, who also happened to be the heaviest consumers of medical services. While leaders like FDR, Truman, and Kennedy had previously failed to pass a UK-style single-payer system due to a lack of domestic political will, a major legislative compromise was achieved in 1965. Under the Social Security Act, Congress created Medicare for citizens over 65 and Medicaid for low-income individuals, federating the administration of the latter out to the states.

This policy framework fundamentally altered the economics of American healthcare. Because costs became heavily laundered through third-party private insurers and government programs, the vast majority of patients stopped feeling the direct cost of their healthcare encounters. The speakers point to this abstraction as a primary reason why U.S. healthcare spending skyrocketed to a staggering 18% of GDP, compared to 11% in the United Kingdom and just 6% in Singapore.

``` [Direct Patient-Doctor Payment Model] │ (1965 Medicare/Medicaid Legislation) │ ▼ [Second & Third-Party Payer System] │ (Existential Need for Documentation) │ ▼ [Standardized Electronic Medical Billing (EMR)] ```

This structural shift forced medical records to evolve from simple clinical diaries into strict financial instruments. To receive payment from insurance companies or the government, hospitals and doctors could no longer just state that they performed a service; they needed bulletproof, highly standardized documentation to prove it. This created what the speakers call a "Faustian bargain" or a "slippery slope" for the medical profession. To secure reimbursement for increasingly complex clinical procedures, physicians had to accept tightly regulated, standardized workflows, giving birth to the modern Electronic Medical Record (EMR) and electronic billing industry.

- **1965:** Congress passes the Social Security Act, establishing the Medicare and Medicaid programs. - **Healthcare Spending as % of GDP:** United States: 18%, United Kingdom: 11%, Singapore: 6%.

[28:52](https://www.youtube.com/watch?v=FNRU9ZNqcls&t=1732s) MUMPS, CO-STAR, and the Birth of Meditech

In 1966, the year following the creation of Medicare and Medicaid, a team at Massachusetts General Hospital began developing CO-STAR (Computer Stored Ambulatory Record), the first true computerized medical record system. Funded by the National Institutes of Health (NIH), CO-STAR was designed to handle patient scheduling, registration, clinical data, and electronic billing interfaces. However, building a highly concurrent transaction system that could support numerous simultaneous hospital users on the limited processing power and memory of 1960s mainframes proved impossible with existing programming languages like Fortran.

To overcome these technical constraints, the MGH programming team wrote an entirely new language: the Massachusetts General Hospital Utility Multi-Programming System, universally known as MUMPS. The language featured two revolutionary design choices: 1. **Integrated Language and Database:** The database structure was native to the language itself, removing the need for programmers to switch contexts or write external SQL queries. This integration made data processing highly efficient. 2. **Built-in Concurrency Control:** It was engineered from the ground up to handle multiple simultaneous users across different departments, preventing data collisions and loss when updating the same patient record.

The lead programmer of this MGH team was a recent MIT graduate named Neil Papalardo. In 1968, Papalardo spun out from the hospital to launch a commercial enterprise around the software. Originally naming the venture Medical Information Technology (MIT), he quickly changed it to Meditech—hypothetically at the request of the university. Today, Meditech remains the number three player in the EMR industry behind Epic and Cerner. The speakers highlight that Epic still relies on a modern descendant of MUMPS called Caché, making the 1960s medical language the de facto standard underpinning modern healthcare infrastructure. Meanwhile, Judy Faulkner was working with Dr. Slack at the University of Wisconsin on medical applications; when Dr. Slack later relocated to Harvard and began collaborating with Papalardo at Meditech, the close-knit community of computer scientists remained dedicated to building the unified, longitudinal patient records that healthcare networks still chase today.

- **1966:** Development of CO-STAR begins at MGH, leading to the creation of the MUMPS programming language. - **1968:** Neil Papalardo founds Meditech (originally named Medical Information Technology). - **Industry Hierarchy:** Meditech stands as the number three EMR vendor globally, trailing only Epic and Cerner.

[33:32](https://www.youtube.com/watch?v=FNRU9ZNqcls&t=2012s) The Living Room Epiphany and the Genesis of Chronicles

The conceptual "holy grail" of healthcare software was a unified system that could track a patient through the entire continuum of care while seamlessly billing the payers. In the mid-1970s, Judy Faulkner learned the MUMPS (referred to here as MS) programming language and set out to solve this exact problem. Legend has it that she was sitting in her living room on a sunny day when she experienced a massive epiphany about how to construct it. Faulkner recalled the moment vividly: "The sun was shining. I was disattentive. I was just sitting there and suddenly it all came to me. Here's how you build it. The integrated system." She immediately ran to her kitchen, grabbed a pad of paper, and began writing "code, code, code, code, code."

This initial burst of programming laid the foundation for Chronicles, a database named for its ability to document a patient's entire medical care journey. While none of Faulkner's original 1970s source code remains in production today, Chronicles still serves as the absolute core database for Epic Systems. This design choice remains Epic's primary competitive edge; whether an institution is utilizing the clinical electronic medical records (EMRs), the "Resolute" billing module, the Cosmos data platform, or specialized applications like Stork or Beans (for kidneys), hundreds of distinct applications all pull data directly from this single, centralized architecture.

- **The Core Differentiator:** Epic relies on exactly one single database that every application directly communicates with, preventing data silos between clinical and billing departments. - **Chronicles Applications:** The system feeds everything from standard EMRs to specialized modules like Resolute (billing) and Cosmos.

[35:11](https://www.youtube.com/watch?v=FNRU9ZNqcls&t=2111s) The Green Screen Era: Hospital Mainframes and Viral Demand

In the 1970s, modern graphical user interfaces did not exist, meaning Faulkner’s breakthrough was purely a backend database rather than a standalone app. The early product operated within a mainframe environment. Different departments at the University of Wisconsin (UW) Medical Center wrote their own basic query screens to overlay on top of Chronicles. Healthcare workers accessed this centralized data using primitive, 80-character-wide, text-only Unix terminals—the classic "green screens"—stationed directly in patient rooms and administrative offices.

Faulkner originally built Chronicles purely as an internal tool for her employer, the Wisconsin medical system. However, the product began to spread virally through word-of-mouth. Academic physicians from UW traveled to medical conferences across the country, boasting to peers about the incredible database their staff programmer had built. Soon, health systems nationwide began calling Wisconsin and Faulkner directly, asking if she could build the system for them. Faulkner resisted initially, waving off suggestions to launch a business by insisting she had only made a single instance and that "we don't need to build a whole company around this thing." Eventually, the overwhelming demand forced her to capitulate. In 1979, she and a few colleagues transitioned to part-time work on the project and founded a company under the remarkably generic name Human Services Computing, which would later become Epic Systems.

- **Hardware Infrastructure:** Early operations required running Chronicles on a large hospital mainframe, connected to 80-column text-only terminals. - **Viral Expansion:** Academic doctors acting as brand advocates at national conferences created organic, nationwide demand for Faulkner's software.

[37:30](https://www.youtube.com/watch?v=FNRU9ZNqcls&t=2250s) Modern Healthcare Infrastructure and the Complexities of Payments

The story of Epic's origin highlights how critical backend systems are to industry innovation, a theme that mirrors modern business dynamics. In the current market, whether a company is a massive multinational or an early-stage startup, having an advanced payments infrastructure is foundational. Entire modern sectors—such as ride-sharing, the creator economy, software-as-a-service (SaaS) marketplaces, and complex B2B supplier networks—simply could not function without the sophisticated transactional engines built over the last decade. Infrastructure partners handle these back-end logistics so founders can focus on their primary product differentiators.

When looking at the contemporary healthcare landscape, massive waves of innovation are arriving via telehealth, artificial intelligence, preventative treatments, and clinical trial management. However, unlocking the full potential of these medical advancements depends entirely on seamless, secure payment systems to improve the overall patient experience. Zooming out, healthcare represents an exceptionally convoluted ecosystem of providers, insurance networks, specialists, and monitoring services. Determining who is paying whom, when they are paying, and under what contractual terms—all while strictly adhering to rigorous medical data privacy regulations—creates an incredibly high-friction financial environment.

- **Macro Payment Scale:** Industry pioneers like JP Morgan Payments move roughly $10 trillion daily across global business ecosystems. - **Healthcare Transaction Management:** Advanced cloud-based systems rely on patented technologies to securely combine electronic transactions with medical data movement. - **Industry Gathering:** The premier healthcare industry event takes place annually every January, serving as a focal point for frontier innovations.

[39:55](https://www.youtube.com/watch?v=FNRU9ZNqcls&t=2395s) Bootstrapping Epic: The $70,000 Capital-Light Phenomenon

When Faulkner left her job to commit to Human Services Computing, she was a pure software developer with zero experience running a business. Her immediate hurdle was securing a computer to write software on, an expensive endeavor in an era sitting awkwardly between giant mainframes and microcomputers. The microprocessor had not yet shifted the landscape, meaning businesses relied on "mini-computers." These machines, like the 16-bit Data General Eclipse that Faulkner targeted, were roughly the size of a standard washer-dryer set and typically lived in basements. While a consumer would never buy one, they were perfect "half-steps" for small businesses or isolated corporate departments that couldn't afford a multi-million dollar IBM mainframe.

To fund this purchase and launch the company, Faulkner secured a $70,000 bank loan. She then rounded up a group of friends, family members, physicians, and fellow programmers from the University of Wisconsin to invest an additional $70,000 in equity. This group valued the company at a $70,000 pre-money valuation, creating a $140,000 post-money structure where Faulkner accepted a 50% dilution alongside her $70,000 in debt. Amazingly, those entities were the *only* primary investors in Epic Systems history. Today, the company generates $5.7 billion in annual revenue and completely dominates American healthcare IT, having never raised another dime of outside equity—a capital-efficient feat reminiscent of IKEA or Microsoft. Faulkner's generational talent as a programmer served as a cornered resource; because she could write world-class software herself, the fledgling company never needed to spend capital hiring an army of developers.

- **The Mini-Computer Standard:** A Data General Eclipse or DEC mini-computer cost approximately $70,000 in the late 1970s. - **Epic Financial Foundations:** Epic launched on exactly $70,000 in equity and $70,000 in bank debt, scaling to billions in revenue from that sole injection. - **The Silicon Valley Intersection:** While Epic historically avoided venture capital, a notable chunk of those original 1970s angel shares eventually found their way into the hands of Sequoia Capital by the 2000s.

[44:26](https://www.youtube.com/watch?v=FNRU9ZNqcls&t=2666s) The Basement Office and the Shared History with American Girl

Epic's first official headquarters was located in the basement of an apartment building at 2020 University Avenue in Madison, Wisconsin. In its earliest years, the company remained incredibly small, consisting of just "three halftime people in the basement" who managed to scale the product remarkably far on their own. Funnily enough, this exact same basement office space was shared concurrently with another iconic Madison startup: the American Girl Doll Company. Founded by Pleasant Rowland, the doll company operated alongside Epic in the shared workspace, and when American Girl eventually moved out to larger facilities, Epic actually purchased some of their leftover office furniture.

- **American Girl Destiny:** The Madison-born doll company was eventually acquired by toy giant Mattel in the late 1990s. - **The $250 Catalog:** The speakers joke about the premium nature of the brand, noting how quickly young children gravitate toward the high-priced items in the American Girl catalog.

[46:05](https://www.youtube.com/watch?v=FNRU9ZNqcls&t=2765s) The MIT Playbook: How Meditech’s DNA Shaped Epic

Recognizing that Faulkner understood code but lacked corporate training, mentor Warner Slack intervened. He invited her out to Boston for a three-day crash course with Neil Pappalardo, the founder of Meditech and a member of the original MUMPS development team at MIT. This introduction was a profound "history turns on a knife-point" moment for Epic. Rather than pairing Faulkner with a traditional business school graduate who might have installed corporate "suits" or replaced the founder, Slack introduced her to a fellow programmer who ran a business *like* a software developer. Pappalardo handed Faulkner his literal corporate manuals, providing standardized, "hardened APIs" for human resources, payroll, and benefits.

Crucially, Meditech’s playbook relied on recruiting fresh, high-agency college graduates from local Boston universities rather than hiring expensive, experienced programmers. Faulkner adopted this exact framework for Epic, pulling talent first from the University of Wisconsin and later expanding to other Midwestern schools. Epic operates under the explicit assumption that smart kids with technical, non-medical majors can easily be taught the nuances of healthcare. This practice of empowering raw college graduates to run free was highly unusual in 1979; traditional career fairs didn't exist, and even Microsoft—which would later become famous for campus recruiting—was still a tiny outfit that had just left Albuquerque for Washington and had not yet begun its massive hiring scaling.

- **Anti-Venture Bias:** Faulkner's lifelong aversion to venture capital stemmed from watching other 1970s founders take funding, only to be promptly fired and replaced by traditional business executives. - **The Corporate DNA:** To this day, Epic remains a massive organization composed almost entirely of programmers, logicians, and implementation teams, with virtually no traditional "business people" in its ranks. - **The 1979 Timeline:** At this point in history, Microsoft had no DOS, had yet to experience its major hiring boom, and was just establishing its footing in the Pacific Northwest.

[50:04](https://www.youtube.com/watch?v=FNRU9ZNqcls&t=3004s) Upmarket Expansion and the Birth of Epic Systems

In the late 1970s, the personal computer industry as we know it did not exist, and Microsoft's primary product was still a basic interpreter. Epic's early Unix-based system could only find traction at large academic hospitals and university medical systems that possessed the necessary mainframe or minicomputer infrastructure. Outpatient facilities and average medical clinics simply lacked the hardware to adopt such a system. Out of deference to competitors like Metatech, Judy Faulkner intentionally steered her company away from the small hospital market, choosing instead to target the most complex, upmarket enterprise institutions—academic training hospitals, children's hospitals, and eventually integrated delivery networks (IDNs).

By 1979, the firm had only four initial customers. Four years later, in 1983, they officially changed their name to Epic Systems. The transition highlighted Faulkner’s creative, somewhat quirky sensibilities. Rather than a declaration of raw corporate ambition, the name was a nod to the concept of a Greek epic—a massive, longitudinal historic tale. Faulkner viewed a patient's life and medical history as an epic narrative. The business itself grew at a remarkably slow, deliberate pace for a software company, functioning essentially as a small, local enterprise in Madison, Wisconsin. By the end of its first decade in 1988, Epic had only 24 customers, a handful of employees, and a modest \$1.5 million in revenue.

- **1979:** Epic starts out with four initial customers. - **1983:** The company renames itself to Epic Systems; the customer base stands at nine. - **1988:** Epic marks its first decade with 24 customers and \$1.5 million in revenue.

---

[53:06](https://www.youtube.com/watch?v=FNRU9ZNqcls&t=3186s) The Resolute Billing Module and Integrated Architecture

A significant reason for Epic's slow initial customer adoption was that it only handled clinical medical records during its first eight years. It lacked a billing system, which is what hospitals desperately needed to generate revenue and justify expensive software investments. This changed dramatically in 1987 with the launch of "Resolute," a dedicated revenue cycle and billing application. Crucially, Resolute was built directly on top of Epic’s existing Chronicles single-core database rather than as a standalone product.

This architectural decision remains Epic’s single greatest competitive advantage over its rivals. While nearly every competitor in the industry expanded by acquiring separate billing systems and gluing disparate products together through erratic mergers and acquisitions, Epic remained a single unified company. This native integration ensures extreme reliability and predictable implementations. When Epic quotes a budget and timeline, they actually deliver "on time and on budget." Because the clinical side and the billing side are seamlessly stitched into one codebase, hospitals don't suffer from data drops or mismatched records when transitioning from patient care to financial reimbursement. As a hospital administrator, if your goal is to log medical practices and seamlessly monetize those events, this architecture is the ideal solution.

> "Epic becomes the natural choice for enterprise decision makers precisely because of its integrated system architecture. Rather than managing multiple vendors and systems, buyers get a comprehensive platform with a single database..."

---

[57:26](https://www.youtube.com/watch?v=FNRU9ZNqcls&t=3446s) Entering the PC Era with Epicare

By the late 1980s and early 1990s, the computing landscape shifted fundamentally as the PC era arrived. Hospitals could suddenly buy computers for a few thousand dollars instead of tens of thousands, blowing the market wide open. Riding this wave, Epic launched "Epicare" in 1992, which stands as the industry's very first Windows-based, graphical user interface (GUI) electronic medical record (EMR) application.

Previously, utilizing Chronicles meant interacting via terminal access directly into a Unix mainframe—a user experience entirely unsuited for the average doctor, nurse, or medical assistant. Epicare completely transformed this paradigm by providing a clean, accessible Windows interface for ambulatory and outpatient settings. When combined with their billing system and their scheduling application, called "Cadence," Epic achieved a highly compelling "holy grail" product suite for health practice administrators. By 1995, Epic's revenue surged to \$18 million, representing a ten-fold increase over seven years. Despite this phenomenal growth, its trajectory contrasted sharply with Microsoft, which went public, released Windows 95, and reached \$6 billion in revenue that same year. Epic was fundamentally a vertically focused healthcare player, though it was tethered to an industry representing 18% of American GDP.

- **1992:** Launch of Epicare, the first GUI-based EMR application, initially restricted to ambulatory (non-overnight) settings. - **1995:** Epic achieves \$18 million in revenue, while horizontal giant Microsoft hits \$6 billion. - **Cadence:** Epic's proprietary scheduling application that completed the outpatient trifecta alongside billing and clinical records.

---

[01:01:14](https://www.youtube.com/watch?v=FNRU9ZNqcls&t=3674s) Achieving the Holy Grail and Pioneering MyChart

Epic fully realized its enterprise vision in 2001 by launching the inpatient version of Epicare, expanding its coverage to complex, high-revenue overnight hospital stays. With Chronicles acting as the backbone database, Epicare handling both ambulatory and inpatient care, and Resolute driving the billing, the company locked in its philosophy of comprehensive breadth. They realized that hospital systems did not want to buy piecemeal software; they wanted an all-in-one vendor ecosystem.

Remarkably, right around the dot-com bubble in the year 2000, Epic also launched "MyChart"—an internet-based, consumer-facing medical records portal. Launching a consumer web platform in a highly restricted, HIPAA-regulated environment at that time was incredibly innovative and pioneering. The project originally branched off from "Epic Web" in 1997, which was designed to let doctors remotely review and update clinical files from home. A young programmer fresh out of college named Sumit Rana was tasked with the project. Feeling unchallenged, Rana directly approached Epic's then-president, Carl D'Vorah, and asked, "Hey, you know, this is good that I'm working on this, but I'm kind of bored. Can you give me something hard to do?"

This interaction highlighted Epic’s unique internal culture. Out of that bold request, MyChart was born. It became the industry's first integrated patient portal, fundamentally changing the consumer healthcare experience overnight. Much like Zillow altered real estate by opening up home pricing data, MyChart made it impossible for patients to return to a world of faxed records, physical medical requests, and tedious phone tag. It became an indispensable lifesaver for everyday tasks, especially for adults remotely managing the complex healthcare needs of elderly parents.

- **2000:** MyChart launches, establishing the first integrated web-based patient portal. - **2001:** Epic releases the inpatient version of Epicare to manage overnight hospitalizations. - **Sumit Rana:** The early-career programmer who requested a harder challenge and went on to design MyChart; he currently serves as the president of Epic.

[01:06:42](https://www.youtube.com/watch?v=FNRU9ZNqcls&t=4002s) The Power of Patient Self-Scheduling

The introduction of patient-side virality transformed operations for both Epic and its healthcare customers. While hospital administration initially found the rollout of self-scheduling within MyChart quite intimidating, it quickly became an extraordinary asset. The workflow savings were massive, eliminating a staggering volume of phone calls that staff previously had to manage manually to schedule simple appointments.

Furthermore, MyChart completely revolutionized how hospitals handle "no-shows" and waitlists. Before these automated systems, a patient who failed to appear for an appointment represented completely lost revenue—creating a substantial, unrecoverable hole in a hospital's daily financial operations. With MyChart dynamically managing patient waitlists, the system can now automatically identify an opening and slot in another patient to fill the gap. Today, this highly efficient system has achieved massive scale, boasting 191 million active, de-duplicated, unique users. Following this momentum, Epic launched MyChart in 2000, followed by EpicCare Inpatient in 2001, pushing the company's annual revenue past the $50 million mark.

- **Active MyChart Users:** 191 million unique, de-duplicated individuals. - **Financial Milestone:** Epic surpassed $50 million in revenue by expanding to 88 health systems, methodically penetrating the market hospital by hospital. - **Product Launch Timeline:** MyChart was released in 2000, and EpicCare Inpatient debuted in 2001.

---

[01:08:32](https://www.youtube.com/watch?v=FNRU9ZNqcls&t=4112s) Acquired Show Sponsor: Fundrise

The hosts take a brief interlude to thank their long-term partner and sponsor, Fundrise, noting how the investment platform has evolved since they first began working together three years prior. Originally renowned as the largest real estate investment platform in the United States for retail investors, Fundrise expanded its horizons after observing how technology markets were scaling exponentially due to Moore's Law. Recognizing that premier tech companies are staying private much longer, Fundrise saw a massive opportunity to democratize access to venture investing for everyday people—a highly contrarian move in 2022 that many historical predecessors had failed to execute successfully.

Today, Fundrise provides ordinary American investors with pre-IPO access to elite private global companies, allowing tens of thousands of regular individuals to participate in massive value creation. When Service Titan went public last December, regular investors got to celebrate alongside traditional venture capitalists, institutional LPs, and company employees. Given the explosive growth of AI, the timing has been ideal for expanding value within the technology sector.

- **Venture Portfolio Highlights:** Fundrise has backed major names including DataBricks, Canva, Anduril, Ramp, Vanta, Anthropic, and Service Titan. - **Contact Information:** Founders or retail investors can view the complete portfolio at `[fundrise.com/venture](https://fundrise.com/venture)`.

---

[01:10:24](https://www.youtube.com/watch?v=FNRU9ZNqcls&t=4224s) Kaiser Permanente and the EMR "Central Nervous System"

In 2003, Epic received a monumental call from Kaiser Permanente, the largest single hospital system in the United States at the time. Headquartered in California, Kaiser operates as a fully integrated, managed-care consortium, meaning they uniquely act as both the insurance health plan provider and the actual hospital system. Ben describes it as "the closest thing that you can sort of have to a single payer here in the country," though they also interface with Medicare through their Medicare Advantage plan for older patients. In 2003, Kaiser's footprint was staggering: 30 hospitals, over 400 clinics, 11,000 physicians, and 8.5 million patients.

When Kaiser put out a Request for Proposal (RFP) for a new, system-wide Electronic Medical Record (EMR) system, it became clear that an EMR is far more than just digital charts or billing software. It is the core operating system—or more accurately, the complex central nervous system—of a healthcare organization. It acts as an incredibly intricate web tying together thousands of screens, authorization levels, authentication protocols, and user roles. Calling it just an operating system takes it too lightly; it is an operating system, an ERP system, and an application layer all rolled into one.

- **Epic's Revenue Jump:** Winning the Kaiser contract catapulted Epic from a $50 million company to $162 million in annual revenue, effectively doubling its business overnight. - **The True Financial Scope:** While headlines widely reported it as a $4 billion, three-year project, Epic’s actual software and implementation portion was around $400 million. The $4 billion headline figure factored in multi-year implementations, internal hospital headcount, and the economic impact of temporary dips in doctor productivity during training.

---

[01:15:14](https://www.youtube.com/watch?v=FNRU9ZNqcls&t=4514s) Internal Politics and the Battle of the Bids

The internal story of how Kaiser chose its vendor is highly chaotic. At the time, Kaiser's Northern and Southern California divisions functioned almost like entirely separate entities under the corporate umbrella, utilizing completely different management teams and disparate EMR systems. Meanwhile, a smaller regional "stepchild" operation in the Pacific Northwest, based in Portland, Oregon, had quietly started using Epic for its ambulatory, outpatient clinics. While the massive California factions were aggressively burning cash with consultants like Accenture to build their own proprietary, homegrown EMR systems—harboring pipe dreams of commercializing and selling their own IP—the physicians migrating from Portland down to California began speaking up. They told their peers, "Like we've got this thing called Epic up in Portland... and it's way better than the stuff that you're trying to build."

Eventually, a truce was called. Northern and Southern California agreed to abandon their competing internal projects and issue an open contract bid, initially selecting IBM for a massive $4 billion project. However, the IBM implementation failed completely and collapsed, which is a frequent and devastating occurrence in the ERP and hospital world. Because Epic staked its entire company reputation on ensuring they "don't have failed implementations," they earned a unique layer of trust among hospital executives (CEOs, CIOs, and CFOs). While the medical staff and clinicians act as the critical "on-air talent" of a hospital system, it is the administration that calls the business shots, and they desperately needed stability.

- **The Epic vs. Cerner Rivalry:** After IBM's failure, Kaiser reopened the RFP, narrowing the selection down to Epic and its chief rival, Cerner. - **Cerner's Scale in 2003:** Founded the exact same year as Epic (1979) by Neil Patterson in Kansas City, Cerner took a radically different corporate path, utilizing venture capital, executing 24 corporate acquisitions, and going public to become a near-billion-dollar international giant. - **The Split-Baby Proposal:** Kaiser initially tried to compromise by proposing a dual system: using Cerner for their foundational inpatient hospitals and Epic for their outpatient ambulatory clinics.

[01:22:40](https://www.youtube.com/watch?v=FNRU9ZNqcls&t=4960s) The Kaiser Architecture Bake-Off and the Excel All-Nighter

When the massive healthcare provider Kaiser Permanente was deciding on an electronic medical records (EMR) provider, Epic founder Judy Faulkner took a "calculated high-risk decision" by insisting they pick a single system rather than splitting the contract. At the time, Epic's inpatient product was brand new and lacked market trust, whereas Cerner boasted a billion dollars in revenue to Epic’s roughly \$50 million. Despite being a "pipsqueak" in comparison, Judy was "super ballsy" and stuck to her guns, knowing Epic excelled in both ambulatory and inpatient systems.

This culminated in an "11th hour" technical due diligence meeting in California to determine if the systems could handle Kaiser's massive concurrent transaction volume of 8.5 million patients and 11,000 physicians. Kaiser had the Epic and Cerner teams operating out of separate conference rooms in the same building, bouncing back and forth every hour to grill each team using questions formulated from the competitor's pitch. While Epic's team initially prepared a theoretical architectural presentation, Epic President Carl Dvorak flew in the night before and completely upended the plan. Recognizing their performance advantage, Dvorak told the team, "Guys, no, we need to model out in Excel exactly what Kaiser's transaction flow is going to be," leading to a grueling all-nighter. The next day, Epic presented a concrete spreadsheet model proving their excess bandwidth, while Cerner failed to provide a similar level of proof, turning the tide of the deal.

- **Kaiser Volume Data:** 8.5 million patients and 11,000 physicians. - **Financial Disparity:** Cerner had \$1 billion in revenue; Epic had slightly over \$50 million. - **The Strategy Change:** Carl Dvorak forced a pivot from a theoretical formula presentation to an exact, data-driven Excel transaction model.

[01:26:12](https://www.youtube.com/watch?v=FNRU9ZNqcls&t=5172s) Equity Ultimatums and the Flooding Gates of Revenue

Even after Epic's stellar performance in the architecture bake-off, Kaiser attempted to leverage its position as the biggest health system in America to demand equity warrants. Kaiser approached both competitors, asking for stock in exchange for the historic contract. Cerner desperately wanted the deal and agreed, offering Kaiser a 10% equity stake in their company. When Kaiser presented this to Epic, asking what they had to say about it, Judy flatly refused. She declared, "No, we're not going to do it. We're not going to do it for you. We're not going to do it for anybody." She argued that giving in would mean offering equity to all major customers, which would ultimately harm the system as well. Epic stood their ground, and Kaiser chose them anyway.

Winning the Kaiser contract completely transformed Epic, causing its revenue to double or triple almost overnight. *The Los Angeles Times* famously wrote that due to its massive scale, the Kaiser-Epic system "could become the Model T of its industry." Once the multi-year implementation went live without tipping over, the floodgates opened around 2006–2008, making Epic the new gold standard for hospital CEOs and CIOs everywhere. By 2007, Epic's revenue rocketed to \$500 million—nearly a 10x increase from its pre-Kaiser days. To cope with this hyper-scale, Epic went from hiring about 10 college graduates a month to hundreds of new hires flowing through the doors.

- **The Equity Offer:** Cerner offered Kaiser 10% of their company; Epic offered 0%. - **Revenue Growth:** Epic grew from north of \$50 million to \$500 million by 2007. - **Historical Mergers:** Cerner's architecture was a patchwork of 24 merged companies, which later grew by another 12 when they acquired Siemens in 2014.

[01:29:59](https://www.youtube.com/watch?v=FNRU9ZNqcls&t=5399s) The Philips Scar Tissue and Corporate Autonomy

The narrative features a critical "alternate history" detour from the early 2000s involving a failed partnership with the Dutch company Philips. Looking for international distribution and a foothold in the radiology segment, Epic agreed to a co-development deal where Philips would market a licensed, branded version of Epic's IT system called "Philips Externity Enterprise." Because Philips sold medical machinery worldwide, it seemed like an appealing way for Epic to go international, prompting them to hire a team in the Netherlands and spend years building it out. However, the entire venture folded within a year or two of launch.

This expensive detour left Epic with intense corporate "scar tissue" regarding external partnerships. The leadership team internalized the lesson that partnerships equal an "uncontrollable dependency" full of risk outside their organization. This failed experiment shaped Epic’s deeply ingrained DNA to ignore fancy partner integrations, stay entirely self-reliant, and go directly to the customer—a closed approach for which they would take "tons of arrows" over the next 15 to 20 years.

- **The Product Name:** Philips Externity Enterprise. - **Analogy Used:** The hosts compared the deal to the "HP iPod" from Apple's weaker era, where HP branded the back of Apple's technology for distribution.

[01:33:13](https://www.youtube.com/watch?v=FNRU9ZNqcls&t=5593s) The Whimsical, Hard-Driving DNA of the Verona Campus

With their eyes on the prize and soaring revenues, Epic decided to invest heavily in its future by building a massive corporate campus in Verona, Wisconsin. The inspiration came when Judy visited her son, a developer at Microsoft’s Redmond campus. She and Carl Dvorak were highly impressed by Microsoft’s "fairy tale tech campus" layout—which was "Google before Google"—featuring low-rise buildings, sports fields, outdoor walking spaces, and private offices for everyone. Judy bought 1,000 acres of farmland in the bucolic Madison suburb of Verona to replicate this strategy, ensuring that spiritually, "Epic owns Verona" just as Microsoft owns Redmond.

However, Epic rejected Microsoft's strict utilitarian aesthetic in favor of something highly whimsical. The inspiration began at their old renovated schoolhouse headquarters in Madison, where designers had added a cozy, Wisconsin-lodge-style fireplace to a main conference room. Adorned with snowshoes, furs, and an axe on the wall, the "lodge room" became so immensely popular with visiting clients that leadership decided to maximize the theme on the new campus. To execute this vision, Epic hired two architecture firms: one that built parts of Microsoft’s campus, and the other responsible for the 2008 Disneyland California Adventure renovation. The result is a "bananas" campus filled with Alice in Wonderland, Harry Potter, and Wizard of Oz themes designed to attract highly competitive, fresh college graduates. This environment reflects Judy Faulkner's dual personality: a goofy, whimsical fairy tale aesthetic juxtaposed with a hard-driving, "win at all costs," fierce competitive execution.

- **Campus Scale:** 1,000 acres of former farmland purchased in Verona, Wisconsin. - **Design Inspiration:** A blend of Microsoft's Redmond layout, a cozy Wisconsin lakeside lodge, and Disney-tier amusement park whimsy.

[01:38:32](https://www.youtube.com/watch?v=FNRU9ZNqcls&t=5912s) The Subsurface Mega-Auditorium and the Software Factory

To attract talent to Verona, Wisconsin, Epic set out to create a literal paradise. The corporate campus spans 1,700 acres, with 410 acres dedicated to the campus itself and the rest left as farmland. The footprint includes 89 buildings and four indoor auditoriums that boast a combined seating capacity of 18,000. The crown jewel of this infrastructure is Deep Space, the world's largest subsurface auditorium. Going down 74 feet beneath the earth, it features 11,400 seats—a capacity that is described as "two radio city music halls full of people smashed together" or closer to a professional basketball arena like the Chase Center than a traditional corporate hall. The original logic was that they would never outgrow it, providing a definitive venue for company all-hands and the entire client ecosystem, though Epic has since grown so massive that even their own employee base no longer fits inside. The inspiration stemmed from their early days when they held monthly all-hands meetings in a local movie theater and wanted to carry that communal tradition onto the new campus at an unbelievable scale.

This grand, unusual campus reflects Judy Faulkner's core philosophy: she explicitly views Epic as a "software factory." Because Epic famously never acquires other companies and maintains no corporate competencies outside of building code, the organization functions as a pure manufacturing plant. It takes in raw developer talent and turns out specialized software for the medical industry. The hosts joke that Judy was decades ahead of her time, pioneering the factory concept long before Nvidia and Dell popularized the phrase "AI factories."

- Total campus size: 1,700 acres (410-acre campus, remaining acreage is farm). - Total facilities: 89 buildings and 4 indoor auditoriums with 18,000 total seats. - Deep Space auditorium: 11,400 seats, subterranean (74 feet below the surface).

---

[01:40:54](https://www.youtube.com/watch?v=FNRU9ZNqcls&t=6054s) The Ten Commandments of Epic

Epic's corporate culture is heavily anchored by its "ten commandments," which are famously posted in every single bathroom and break room across the campus. Because the campus is entirely open to the general public, any visitor or client can walk in and read them. The first two commandments are explicitly structural: number one is "do not go public," and number two is "do not acquire or be acquired." The hosts find humor in the fact that these are framed as commandments for general staff, considering only the CEO has the power to execute an IPO or an acquisition. However, broadcasting these rules serves a vital dual purpose. For employees, it embeds Judy’s long-term philosophy deeply into their daily mindset. For visiting customers, it acts as a constant promise that Epic will remain independent and stable, assuring them that "this is one system forever."

The remaining eight commandments dictate operational excellence and corporate behavior, completely bypassing any standard healthcare platitudes. There are no mentions of putting patients first or saving lives; instead, they serve as a strict guide on how to run a highly disciplined business. The list includes mandates like "software must work," "reality equals expectations," and keeping even unspoken commitments. It demands a high bar by telling staff to "focus on competency" and to "not tolerate mediocrity." The eighth commandment warns, "what you put up with is what you stand for," while the final rules dictate teaching culture and remaining strictly frugal by never taking on operational debt.

- **The Ten Commandments of Epic:** 1. Do not go public. 2. Do not acquire or be acquired. 3. Software must work. 4. Reality equals expectations. 5. Keep commitments, even the unspoken ones. 6. Focus on competency. Do not tolerate mediocrity. 7. Have standards. Be fair to all. 8. Have courage. What you put up with is what you stand for. 9. Teach philosophy and culture. 10. Be frugal. Do not take on debt for operations.

---

[01:42:51](https://www.youtube.com/watch?v=FNRU9ZNqcls&t=6171s) Squeezing Through the Epic Machine

When a new client signs a contract with Epic, wedding bells play campuswide via the loudspeaker system. This "wedding march" is Epic's version of ringing a sales gong, symbolizing a lifelong marriage commitment between the company and the health system. For young, ambitious college graduates, Epic serves as an incredibly intense and rewarding training ground. While critics often accuse the culture of being "cultlike," the hosts argue that this intense immersion is highly effective. Epic takes young talent fresh out of school and systematically teaches them the exact, hardened practices iterative testing has proven to be the best. There is a precise, mandated "Epic way" to write emails and even a specific method for taking notes on a yellow legal pad.

This hyper-systematized approach builds an environment of extreme internal trust, allowing Epic to operate with very few middle managers and no traditional corporate budgets. Judy notoriously hated standard corporate budgeting; she recalled stories of clients pushing off crucial decisions because they lacked the budget or rushing wasteful spending just to avoid losing their annual allowance. Calling those practices "stupid," she eliminated traditional budgets at Epic. The company is also incredibly light on corporate titles; whether someone has been at Epic for twenty years or six months, their business card might simply read "implementation." To maintain close ties to the real world, every single employee must go on mandatory "immersion trips," spending time in fast-paced clinical environments like operating rooms to directly observe workflows. New hires complete five immersion trips out of the gate, with ongoing annual requirements to keep them grounded in customer reality.

- New hires go through a rigid 6-month intensive training program. - Implementation staff routinely work 10 to 12 hours a day in high-adrenaline roles. - All employees are required to do 5 clinical immersion trips at the start of their career, plus annual follow-ups.

---

[01:45:43](https://www.youtube.com/watch?v=FNRU9ZNqcls&t=6343s) The Zero-Bug Methodology and Up-or-Out Retention

On the engineering side, Epic employs a highly prescriptive software methodology explicitly designed to minimize bugs. Because a software glitch in a prescription dose could quite literally kill patients, or cause a hospital to get sued for federal medical fraud due to revenue cycle errors, Epic values a "zero bug environment" over sheer speed or cutting-edge innovation. The engineering pipeline calculates and minimizes the exact number of hours between when a line of code is written and when it gets tested. If a bug is caught, the system dictates an unyielding rule: the original developer must drop everything else to fix it immediately while the context is completely fresh in their mind. This prevents bugs from compounding into massive structural defects over time.

The environment operates on a strict "up or out" philosophy. Epic aggressively trims the bottom tier of performers and works its staff incredibly hard. Because of the sheer volume of new college graduates coming in, the company uses the first few years to filter out those who cannot cut it, finding it more cost-effective to replace underperformers than to carry dead weight. However, this creates a high-trust workforce where anyone who survives the initial years is known to be exceptionally reliable. The organizational hierarchy remains flat enough that junior programmers regularly find themselves coding side-by-side with senior executives, such as company president Carl Dvorak, creating a thrilling environment for high performers. Isolated out in the countryside near Madison, Wisconsin, employees find it easy to let Epic completely consume their lives, drawing tight structural parallels to the early days of Microsoft in the 1980s and 1990s or the intense deployment cultures found at Palantir.

- High-stakes logistics managing what Peter Drucker called "the most complex form of human organization"—the hospital. - Code issues must be resolved immediately by the original author to maintain absolute mental context.

---

[01:50:09](https://www.youtube.com/watch?v=FNRU9ZNqcls&t=6609s) The Teenage Interview Test and Customer Obsession

Epic's historic methodology for hiring is legendary and revolves heavily around rigorous pre-employment testing rather than standard conversation. The origin of this practice dates back to the late 1980s when Judy Faulkner realized that traditional interviews were terrible predictors of engineering talent. At the time, her son was only around the seventh grade but was frequently winning computer science and programming competitions across the state. Recognizing that these timed tests were highly predictive of actual capability, Judy asked her teenage son to write a customized programming test for Epic to screen applicants. For 18 years, that teenager's test was the primary gatekeeper for an Epic job offer, often bypassing the interview process entirely. Though that specific test was eventually retired after its answers leaked onto the internet, it set a precedent. Today, every applicant takes a battery of systemized assessments, including logic exams for the culinary staff and the Rembrandt profile for corporate roles, a progressive hiring strategy Epic pioneered long before Google popularized analytical screening.

Externally, the resulting culture manifests as total customer obsession. The hosts quote Jeff Gotney, a CIO within the Rush University System for Health, who remarked that despite Epic being "not cheap," a buyer "gets what you pay for 100% of the time." Epic has achieved the enviable institutional status once held by IBM: "no one gets fired for buying Epic." They never overpromise, and their integration is completely reliable. To maintain this hyper-focus, Epic structures its entire product roadmap around its users; during the massive annual customer conference held on campus, clients pitch features, vote on priorities, and directly dictate what Epic’s software factory will build next.

[01:53:17](https://www.youtube.com/watch?v=FNRU9ZNqcls&t=6797s) Lean Operations and the BFF System

Epic operates with an incredibly lean corporate structure that features basically only three primary roles: software developers, project managers (who handle active implementation for new customers), and technical specialists. Strikingly, the company lacks a traditional sales or marketing department. Instead, a tiny group of about eight "salespeople" merely react to inbound requests, and every single one of them was internal, having transitioned from project management or technical specialist roles.

The technical specialists make up the single largest group at Epic. The company ensures that every single one of its 607 or so hospital system customers has a dedicated technical specialist team assigned to every single individual product they utilize, whether it is Epicare EMR, MyChart, Resolute, or Cosmos. On top of these product-specific teams, each customer is assigned a single dedicated point of contact known as a "BFF" (Best Friend Forever). The sole mission of the BFF is to guarantee the customer's success. As part of this relationship, the BFF issues an annual report card directly to the client's CEO, CIO, and CFO. This document grades the hospital system on a scale of 1 to 5 across multiple operational dimensions and explicitly benchmarks their performance data against a peer set of similarly sized hospital systems.

[01:55:09](https://www.youtube.com/watch?v=FNRU9ZNqcls&t=6909s) High Leverage and "Altruistic Capitalism"

Epic maintains immense leverage in its customer interactions. While the company is ultimately customer-focused and will eventually defer to a client's final demands, they approach disagreements by laying out incredibly compelling arguments for why the Epic way is correct. This highly opinionated stance manifests in their strict push for a "standard package." Unlike competitors like Cerner, where every implementation looks entirely different, Epic vigorously pushes standard setups to ensure seamless update rollouts, module additions, and reliable interoperability. They back this up with a powerful negotiating position; Epic only takes on 10, 20, or maybe 30 new customers a year and is perfectly content to tell a hesitant hospital, "I don't think you're ready to be a customer yet," and walk away until they are compliant.

This rigid standard-package philosophy is tied directly to Epic's pricing structure, as adhering to standard or mostly standard implementations is practically the only way a hospital can secure a discount. Additional financial incentives are offered if clients stay completely up-to-date on database maintenance and software versioning. Despite this aggressive accumulation of operational power, the hosts note that there is an almost altruistic dimension to the company's culture, driven by the personality of its founder, Judy Faulkner. The dynamic is colorfully summarized as "ultra competitive, ultra value maximizing, altruistic capitalism."

- Epic historically limits annual price increases to an average of around 2%, meaning their price adjustments trend below inflation. - The company claims to have never changed the core pricing of MyChart throughout its entire history, though the hosts note this specific detail might be a bit of a "red herring" since outsiders lack access to complete enterprise contracts to see the total bundled cost.

[01:58:00](https://www.youtube.com/watch?v=FNRU9ZNqcls&t=7080s) Bundling Strategies and Customer Obsession

The fundamental reason a health system chooses Epic is simple and pragmatic: to turn a patient interaction into as many dollars as possible with the lowest possible risk. Epic aligns its goals with this reality to win deals, maintain lifetime retention, and ultimately achieve what the hosts half-jokingly call "world domination." To achieve this, Epic deploys incredibly clever bundling tactics under the guise of customer service. For instance, when telehealth suddenly became critical in 2020, Epic explicitly told its clients to use a HIPAA-compliant version of Zoom in the interim while openly disclosing that Epic was developing its own telehealth module. They promised the upcoming module would be completely free to existing customers, effectively killing any incentive for hospitals to shop around for third-party alternatives. Over time, as the native tool matured, hospitals naturally adopted the Epic version as part of their enterprisewide agreements—a strategy that closely mirrors Microsoft's historic bundling plays.

This customer obsession is framed not as pure altruism, but as an acute understanding of long-term self-interest. If you deliver absolute value for the hospital executives, you deliver for Epic over the long haul. This corporate philosophy echoes a classic Jeff Bezos quote from around 1999, which noted that in the long run, there is no divergence between what is good for the customer and what is good for the platform.

[01:00:21](https://www.youtube.com/watch?v=FNRU9ZNqcls&t=3621s) Intermission: Service Now and the Human Middleware

While the history of Epic transitions into the late 2000s, the hosts take a brief detour to thank their sponsor, Service Now, drawing parallels between the operational philosophies of both companies. Service Now has quietly grown a massive, billion-dollar-plus CRM business by acting as a single platform for an enterprise’s data and operations. Traditional CRM platforms built decades ago were designed merely to record data and report on activities rather than to execute actions or seamlessly tie distinct business divisions together. By contrast, Service Now focuses on delivering comprehensive organization-wide service, eliminating the need for disjointed applications.

When a company relies on separate siloed applications for sales, support, and logistics, employees are forced to act as "human middleware"—manually typing and updating the exact same information across multiple disconnected systems. Utilizing a unified platform provides a single source of truth where a sales interaction instantly updates across the entire enterprise ecosystem. This approach is particularly powerful given the modern rise of AI agents; an AI agent can instantly pull cross-departmental data from inventory, logistics, and customer history to actually resolve a customer issue, such as a delayed shipment, rather than just logging a static complaint.

[02:03:21](https://www.youtube.com/watch?v=FNRU9ZNqcls&t=7401s) Bipartisan Winds of Digital Healthcare Reform

Moving deeper into the 2000s, the narrative shifts to 2006 during the George W. Bush administration. In his State of the Union address, President Bush explicitly called for the wider adoption of electronic records and health information technology to control soaring medical costs and curb dangerous clinical errors. The hosts find it incredibly ironic that popular public perception usually attributes the massive push for electronic medical records (EMRs) and "meaningful use" guidelines entirely to the Obama administration and Obamacare, when in reality, the momentum was completely bipartisan and heavily initiated under Bush.

The shifting political landscape of the era highlights how dramatically partisan lines have moved. In that same speech, Bush advocated for health insurance coverage to be entirely portable so American workers could easily switch jobs without fear of losing their healthcare. Just four years later, during the fierce political battles over the Affordable Care Act, a Republican making that exact same argument would have faced intense backlash from within their own party. This political push was born out of a universal realization that the American healthcare system was suffering from rampant "cost disease"—consuming a massive portion of GDP while remaining deeply inefficient. Digitization was championed as the ultimate promise to deliver the country from these systemic failures, at a time when the industry was profoundly analogue.

- In the mid-2000s, standard data indicates that only 13% of healthcare facilities in America utilized an EHR system of any kind. - While some industry insiders push back on that 13% figure by arguing that the technical definition of an EHR shifted during this era, it remains the baseline metric for how un-digitized American medicine was at the time.

[02:06:11](https://www.youtube.com/watch?v=FNRU9ZNqcls&t=7571s) The Three Buckets of Interoperability

To understand the controversies surrounding Epic and electronic health records, one must understand the concept of "interoperability," which breaks down into three distinct operational buckets. The first and simplest bucket is Epic-to-Epic transfers, where a patient moves between two entirely separate hospital systems that both happen to use Epic software. Because they share the underlying architecture, data transfers smoothly. The second bucket is Epic-to-competitor transfers, such as moving data from Epic to Cerner, Meditech, or Allscripts. This is vastly more difficult technically because the architectures are entirely different, and unlike modern B2B SaaS verticals, the healthcare industry has historically struggled to maintain uniform data standards.

Beyond the technical hurdles of the second bucket lie powerful misaligned business incentives. Hospital systems are fierce local competitors and have little natural business incentive to make it easy for a patient to take their records and walk to a rival facility. Similarly, Epic has zero market incentive to make data sharing with rival software systems seamless. While all parties publicly insist they want to prioritize patient care, the raw capitalistic incentives run completely counter to external data integration.

The third bucket involves interoperability between Epic and third-party applications attempting to sit on top of or extract data from Epic's ecosystem. Epic has historically been extraordinarily defensive and restrictive in this category, fiercely guarding its perimeter. While application developers find Epic notoriously difficult to integrate with compared to other software markets, Epic defends its closed stance by pointing to an unblemished data privacy record, proudly noting they have completely avoided a "Cambridge Analytica situation" by preventing patient data leaks to external developers.

[02:08:36](https://www.youtube.com/watch?v=FNRU9ZNqcls&t=7716s) The "Care Everywhere" Strategy and Its Accidental Success

The speakers contrast the strict regulatory restrictions of HIPAA data with Epic’s commercial incentives, noting that keeping data walled off aligns perfectly with Epic's business strategy. However, the first scenario—Epic-to-Epic data sharing—is surprisingly seamless. Epic developed a tool called "Care Everywhere," which facilitates the exchange of 20 million patient records daily. One host shares a personal anecdote about how he easily linked all of his MyChart accounts across three competing health systems in Seattle while preparing for the episode.

The origin of Care Everywhere involves a dramatic corporate gamble by Epic's founder, Judy Faulkner. She personally mandated that the interoperability feature could not be toggled on or off by individual customers; if a hospital enabled it, they had to share data with all other Epic customers, including local rivals. The software launched successfully due to "pure luck" when one of Epic's early clients unknowingly agreed to the interoperability clause in their contract. The health system's CEO later admitted he would have flatly rejected the feature had he noticed it. Capitalizing on this accidental win, Epic made Care Everywhere mandatory for all future contracts and even retrofitted older software versions. Today, this nationwide network serves as Epic's primary talking point to defend its record on interoperability.

[02:10:20](https://www.youtube.com/watch?v=FNRU9ZNqcls&t=7820s) The Friction of Capitalist Incentives in Healthcare

Despite Epic's internal ecosystem, broader industry interoperability entering 2008 remained remarkably poor. Hospital administrators and CIOs frequently described healthcare data sharing as "laughable," noting that commercial entities have zero incentive to share valuable assets with competitors unless legally compelled to do so. One CIO framed the issue by asking if United Airlines would ever willingly hand over its proprietary customer data to Delta just because it was "customer-friendly."

This reality creates a sharp emotional tension for listeners. The hosts point out that anyone tracking this story will feel a conflict between their "inner capitalist"—who recognizes that hospitals operate as businesses making rational commercial decisions—and their "inner human being," who feels that public health infrastructure shouldn't be governed by the cold mechanics of profit maximization. They conclude that this friction is a defining feature of the highly regulated, for-profit American healthcare landscape.

[02:12:24](https://www.youtube.com/watch?v=FNRU9ZNqcls&t=7944s) The 2009 HITECH Act: Government as a Software Startup's Best Friend

The narrative shifts to the 2008 Great Financial Crisis, prompting a humorous aside where the hosts express a desire to dedicate an entire future episode to the global financial collapse and subsequent bailouts. To stimulate the flatlining economy, the federal government looked for "shovel-ready projects"—initiatives that were widely agreed to be beneficial and could absorb massive fiscal stimulus. Digitizing the nation's archaic medical records became the ultimate shovel-ready target, giving birth to the Health Information Technology for Economic and Clinical Health (HITECH) Act of 2009, which was folded into the nearly $1 trillion American Recovery and Reinvestment Act.

The mechanics of the HITECH Act created a dream scenario for Electronic Health Record (EHR) vendors. The bill injected $27 billion in direct incentive payments to hospitals adopting EHR systems, a figure that climbed to $36 billion when accounting for broader health IT training and data infrastructure. The hosts compare this to the government handing out dedicated-use credit cards to a startup’s customers, paying them first to buy the software and then paying them more to actually use it. Conversely, a failure to adopt the software within five years triggered severe financial penalties.

- **Direct Incentives:** Provided $44,000 to $64,000 per physician over several years, paid out to the hospital systems. - **Scale Example:** For a massive provider like Kaiser Permanente with 11,000 physicians, the available government funding represented an enormous windfall.

[02:16:33](https://www.youtube.com/watch?v=FNRU9ZNqcls&t=8193s) The "Meaningful Use" Carrot, Stick, and Market Consolidation

The HITECH Act introduced the industry-triggering phrase "Meaningful Use" (or "MU"), which functioned first as a carrot and later as a financial stick. While the scrappy, $2 billion EHR vendor community initially griped about the government's strict list of boxes to check, the massive influx of cash forced them in line. Rusty France, CEO of Epic competitor NextGen Healthcare, noted that the entire industry simply jumped at the check dangling in front of them.

While the legislation forced Epic to shift its development priorities and hop through regulatory hoops, it ultimately tilted the competitive landscape heavily in their favor. When faced with choosing an EHR system funded by free government money, risk-averse hospital executives bypassed cheaper or more innovative startups. Instead, they bought the most expensive, reliable, and established option on the market because they knew it was guaranteed to work and qualify them for the stimulus. The hosts liken it to receiving a government voucher for a free handbag and choosing a Birkin bag over a target brand because expense is no longer a constraint.

- **Market Penetration:** EHR adoption skyrocketed from a mere 9% of US hospitals in 2009 to a staggering 95% by 2014. - **Historical Comparison:** The hosts note that this explosive five-year adoption curve is virtually unseen in software history, with the only modern parallel being Zoom's rapid growth during the COVID-19 pandemic.

[02:20:18](https://www.youtube.com/watch?v=FNRU9ZNqcls&t=8418s) The Mixed Legacy of HITECH: Costs, Billing, and Political Influence

While digitization was a massive victory for patient convenience—allowing people to message doctors, schedule appointments, and manage family care via portals like MyChart—its secondary goals saw mixed results. Though one study claimed a 10% reduction in hospital costs, strong counterarguments suggest EHRs actually drove healthcare costs higher. Critics point out that electronic systems streamlined the ordering of unnecessary care and automated the maximization of billing codes.

The hosts cite an *American Prospect* article titled "An Epic Dystopia," which highlights how hospital management teams leveraged EHRs to audit clinical behavior for revenue maximization. In one instance, a doctor recounted being reprimanded by a supervisor asking why an appointment was billed as a level "two" instead of a higher-paying level "three." Because hospitals are commercial entities staffed by professionals who want to make a good living, the computerized systems naturally became optimization tools for higher-dollar insurance billing.

Furthermore, the HITECH Act failed to mandate strict data standards, allowing vendors to ignore cross-platform interoperability. This omission directly benefited Epic, whose closed, self-contained suite didn't require external integration to be valuable. Addressing a common industry criticism, the hosts acknowledge that Judy Faulkner sat on President Obama's health IT advisory council during this period. However, they clarify a frequently omitted nuance: major rivals like Cerner and Allscripts also held seats on those same councils, meaning Epic was not uniquely rewriting federal policy in a vacuum.

The Cost of Compliance and Market Distortion

The conversation shifts to the long-term consequences of the HITECH Act and its "meaningful use" guidelines on the Electronic Health Record (EHR) market. The speakers reflect on how government intervention inevitably creates market distortions and complications, particularly when regulating how products are developed and used. By explicitly defining what constituted meaningful use, Congress inadvertently designed software features and dictated how doctors should perform their day-to-day jobs. Healthcare providers now find themselves constantly clicking through menus just to satisfy legislative checklists and protect their hospital systems from steep financial penalties, rather than focusing on patient care.

This compliance-driven environment has dramatically increased the overall regulatory and operational overhead for the entire healthcare industry. Even after the initial federal stimulus funds dried up, the mandatory fields and rigid workflows remained permanently embedded within hospital software. The speakers note that the system became optimized for legal adherence rather than fulfilling the original spirit of the law, which was to help doctors and patients. This increased cost of doing business acted as an accelerant for hospital consolidation, forcing health systems to merge to absorb the rising overhead costs.

- The operational overhead of the industry was permanently increased due to lingering mandatory workflows after stimulus funds expired. - Hospital consolidation accelerated as independent healthcare systems merged to handle the rising costs of regulatory compliance.

The Digitization Silver Lining and a "Jobs Program" Perspective

Despite the glaring flaws, the speakers ponder the alternative, admitting that remaining on a paper-based medical system was unsustainable. While the specific execution of the digitization was heavily flawed, the broad achievement of shifting the industry to digital records is still viewed as a positive baseline. Zooming out further, they discuss whether the HITECH Act could be viewed through a different lens: as a highly successful economic stimulus and jobs program. Given that healthcare and related fields account for 18% of US GDP, the legislation successfully birthed an entire domestic industry of tech administrators, IT professionals, and specialized Epic systems coordinators.

David pushes back on praising government-sponsored jobs programs, noting that "mailing checks would have stimulated the economy" just as well. He views the resulting electronic medical records as a "free stapled coupon" that came along with the broader financial stimulus package of that era—hardly a perfect system, but ultimately better than what would have existed without intervention.

- Healthcare and its supporting industries make up roughly 18% of the United States GDP. - The legislation successfully generated thousands of domestic, import-independent IT and administrative careers across American health systems.

Pulling the Future Forward and Regulatory Tailwinds

Returning to the Epic corporate narrative, the hosts explore how the $36 billion federal push altered the competitive landscape. Epic was already on track to become the dominant industry player following its major contract win with Kaiser, but the massive influx of government cash effectively closed the door on future innovation. By forcing hospitals to undergo Request for Proposal (RFP) processes all at once to capture federal funding, the government locked in the existing competitive set for decades due to the immense switching costs of EHR infrastructure. This effectively locked out theoretical, more innovative startups that might have emerged down the road.

Rather than labeling this phenomenon as pure "regulatory capture," the hosts describe it as a "regulatory tailwind." Epic did not necessarily lobby for or love the legislation, as it forced them to build their software in a fragmented, "Frankenstein style way" to check bureaucratic boxes. While it solidified their market victory by accelerating industry demands by five years, it simultaneously degraded the core product and burdened doctors with checking "57 boxes on every patient visit."

- The federal government injected $36 billion into the healthcare IT sector to incentivize rapid EHR adoption. - A 2016 study highlighted the software's administrative burden, revealing that doctors spend about 2 hours entering data into EHRs for every 1 hour of direct, hands-on patient care. - Digital compliance has eliminated the practical "slack" in hospital workflows; previously, medical assistants could queue up pharmacy or scan orders for doctors to sign off on later, but strict digital tracking now forces doctors to input these orders personally, accelerating burnout. - A historical study from 1970 revealed that communication and physical record management once accounted for 35% to 39% of total hospital operating costs, illustrating that administrative burdens have long plagued the sector.

Reflecting on Digital Transformation vs. Digitization

Even the architects of the bill recognized its shortcomings. In 2017, former President Barack Obama stated to Vox that the HITECH legislation had not fully lived up to his expectations, lamenting that "nurses are spending all their time on this administrative work" and that lowering overall healthcare costs to match the rest of the world remained elusive. Summarizing the era, a prominent hospital CIO quoted by the hosts remarked that the HITECH Act "wildly succeeded at digitization of the industry. It did absolutely nothing on digital transformation."

However, the hosts express optimism for the future. Now that the grueling foundational work of converting records into data is complete, the stage is set for true digital transformation. They anticipate that emerging technologies, such as ambient AI tools, will eventually step in to automate these onerous data-entry processes and liberate humans from administrative monotony.

Epic's Ascendancy and Competitor Consolidation

Amidst the industry-wide compliance struggles, Epic continued to experience explosive financial growth. By 2011, the company surpassed $1 billion in annual revenue, routinely securing enterprise contracts for prestigious institutions including Johns Hopkins, Cedars-Sinai, and UCSF. Conversely, the strict requirements of meaningful use triggered a frantic wave of consolidation among Epic's fragmented competitors, which ultimately played further into Epic's hands as combined legacy suites grew increasingly disjointed and difficult to navigate.

- Epic reached $1 billion in revenue in 2011 while sweeping major tier-one academic medical center RFPs. - Allscripts merged with Misys in 2008, and the newly formed entity merged again with Eclipsys in 2010; the consolidated company operates today under the name Veradigm. - Meditech acquired LSS Data Systems in 2011 to shore up its platform. - Cerner acquired Siemens Health Services for $1.3 billion in August 2014, a massive integration effort that many insiders believe triggered Cerner's eventual downward spiral. Even by 2025, legacy customers still bifurcate themselves as either Siemens or Cerner users.

Sponsor Message: Crusoe AI Infrastructure

The hosts briefly pause the healthcare chronology to introduce a segment on their long-time sponsor, Crusoe. Described as an exciting glimpse into the future of computing, Crusoe is highlighted as a vertically integrated AI infrastructure company that builds its stack from the ground up—sourcing stranded energy, constructing data centers, and managing the cloud software layer.

The hosts share details regarding Crusoe's massive 1.2-gigawatt AI data center project in Abilene, Texas. After breaking ground on an empty field in June 2024, the company managed to begin spinning up its first operational buildings within a single year. To contextualize the scale of the Abilene site, the hosts point out that the entire existing data center hub of Northern Virginia—which handles roughly 70% of global internet traffic—amounts to approximately 4.5 gigawatts. The hosts also announce their personal participation as investors in Crusoe's latest funding round alongside prominent venture firms.

- Crusoe is constructing a 1.2-gigawatt AI data center facility in Abilene, Texas, which broke ground in June 2024. - The total data center capacity of Northern Virginia sits at roughly 4.5 gigawatts. - Crusoe recently launched two new operational features: a managed inference service and an "autocluster" service designed to simplify large-scale model training. - Ben and David invested in Crusoe's recent financing round alongside Nvidia and Founders Fund.

The 2015 Department of Defense Megacontract

The narrative shifts back to the EHR industry, arriving at the massive Department of Defense (DoD) contract of 2015. The hosts frame this bidding war as the ultimate high-stakes prize for any electronic medical records company, while simultaneously calling it a prime historical example of American governmental waste.

[02:39:06](https://www.youtube.com/watch?v=FNRU9ZNqcls&t=9546s) The Battle for the Billion-Dollar Military Contracts

The US Department of Defense (DoD) decided to bid out a massive global electronic health record (EHR) contract to cover all of its hospital and healthcare services across every branch of the military. At its inception, this deal was strictly for active-duty US military personnel, keeping it entirely separate from the Veterans Administration (VA) system that would come later. Signed in 2015, this historic $4.3 billion deal surpassed even the landmark Kaiser contract to become the largest healthcare IT contract of all time. It held that record until just two years later in 2017, when the VA contract was put up for bid at a staggering $10 billion.

Epic naturally bid on both of these high-stakes contracts. Just as they had seen with the Kaiser deal, the final competition narrowed down to Epic and Cerner. However, they couldn't compete directly due to the "insanity" of the government procurement system. Unless a company makes bidding on federal contracts its full-time focus, navigating the bureaucracy directly is nearly impossible. To participate, tech firms must partner as subcontractors with established "big government primes"—defense contractors that make their bread and butter working with the US government. The bidding process quickly turned into a high-stakes drama where "the knives come out on all sides" because winning either contract promised to sustain a company's revenue for decades.

* **DoD Contract Value:** A $4.3 billion deal awarded in 2015 for active-duty military personnel. * **VA Contract Value:** A $10 billion contract put out for bid in 2017 for the Veterans Administration. * **The Procurement System:** Strict government regulations required commercial EHR vendors like Epic and Cerner to bid as subcontractors under primary government contractors rather than bidding directly.

[02:40:40](https://www.youtube.com/watch?v=FNRU9ZNqcls&t=9640s) The Brutal Reality of Government Implementation

Cerner ultimately triumphed, winning both the massive DoD and VA contracts as a subcontractor. However, in a twist that the speakers note will "perhaps not shock you at all," both projects instantly fell into a quagmire of massive time and budget overruns. The active-duty DoD system, initially bid out in 2015, only achieved full deployment late in 2024—representing a grueling nine-year implementation timeline.

The VA system has fared significantly worse and remains far from fully live. According to the VA's latest official announcement, the optimistic target to have the system live across all VA sites is "as early as 2031." The speakers laugh in disbelief at this timeline, pointing out that even this best-case scenario represents a 14-year project timeline from its 2017 origin. While the system is operational at a few select test sites, the glacial pace serves as a point of frustration for American taxpayers. The host highlights a cynical reality of the military-industrial complex: due to perverse incentives, "you make more money on failed government contracts than successful ones" because ongoing implementation allows contractors to continually accumulate and bill for additional costs.

* **DoD Timeline:** Took 9 years to fully implement, starting in 2015 and going fully live in late 2024. * **VA Timeline:** Initiated in 2017, the system is projected to be fully live across all sites "as early as 2031" under the most optimistic scenario. * **Systemic Failure:** Multi-decade projects consuming billions in taxpayer dollars due to compounding layers of bureaucracy across the military, government, and electronic medical record (EMR) systems.

[02:43:26](https://www.youtube.com/watch?v=FNRU9ZNqcls&t=9806s) Epic’s Secret Blessing and Cerner’s Corporate Chaos

While Epic desperately wanted to win these contracts—noting that the $10 billion VA price tag alone was double Epic's total annual revenue at the time—losing turned out to be a massive blessing in disguise. Today, Epic's customers and Chief Information Officers (CIOs) are "down on their hands and knees thankful" that the company lost. While Cerner got dragged deep into a bureaucratic "shitshow" and trapped in the mud of federal implementation, Epic remained completely unburdened.

Compounding Cerner's misery, the company was already struggling to integrate a patchwork of 36 different acquired companies, including its massive acquisition of Siemens Healthineers' EHR division. To make matters worse, Cerner's visionary founder and long-time leader, Neal Patterson, was diagnosed with cancer and passed away in 2017 right in the middle of this high-stakes transition. Deprived of its founder, Cerner cycled through a rapid succession of different leaders over the subsequent years. Epic took advantage of Cerner's internal chaos by maintaining absolute focus on Judy Faulkner's unconventional playbook: staying private, avoiding acquisitions, listening intently to users, and keeping a rigid, uniform product architecture.

* **Founder Loss:** Cerner's long-time leader and founder Neal Patterson passed away from cancer in 2017. * **Corporate Disruption:** Cerner faced leadership instability and technical friction from trying to stitch together 36 legacy acquisitions. * **Epic's Growth Strategy:** Avoided artificial growth or short-term step-functions, prioritizing a slow, deliberate march to build deep structural momentum over 20 to 30 years.

[02:45:19](https://www.youtube.com/watch?v=FNRU9ZNqcls&t=9919s) Epic’s Total Domination of the Healthcare Market

Unshackled from government contract overruns, Epic went on an unprecedented winning streak across the private sector. They secured massive enterprise deals with Partners Healthcare in Boston (the mega-entity formed by Harvard and Massachusetts General Hospital), the entire Mayo Clinic network, Intermountain Health, and Common Spirit Health. They even successfully expanded overseas, winning Cambridge in the UK. By 2018, Epic's revenue climbed to $2.7 billion. More importantly, they announced they had won contracts with all 20 of the top academic medical hospitals ranked by *US News & World Report*.

Securing top-tier academic institutions gave Epic an ironclad competitive advantage. By controlling the university hospital systems, Epic effectively captures the training ground for the next generation of medical professionals. Today, a staggering 90% of all medical school students in the United States are formally trained on Epic software. Even though Epic sells the classroom educational licenses separately from the clinical hospital deployment, the systemic familiarity ensures that newly minted doctors and nurses enter the workforce demanding Epic at their future employers. By 2019, Epic's revenue surged to $3.2 billion, while Cerner’s momentum completely stalled out.

* **2018 Revenue:** Epic reached $2.7 billion in annual revenue. * **Academic Sweep:** Epic captured 100% of the top 20 academic medical hospitals in the United States. * **Training Monopoly:** Approximately 90% of current US medical students complete their clinical training using Epic's interface. * **2019 Revenue:** Epic's annual revenue grew to $3.2 billion.

[02:46:37](https://www.youtube.com/watch?v=FNRU9ZNqcls&t=9997s) The Fall of Cerner and the Oracle Acquisition

The final chapter of the Epic-Cerner rivalry culminated in December 2021 when Oracle purchased Cerner for a massive $28 billion. Although Cerner remained a large international corporation pulling in roughly $5.5 billion in annual revenue—bolstered significantly by the steady influx of federal cash from the ongoing DoD and VA contracts—its core business growth had been completely flat to declining since 2018.

The acquisition has not aged well for Oracle, which has since stopped reporting Cerner's financials as a separate segment. Instead, corporate leadership frequently laments Cerner and Oracle Health during earnings calls, characterizing them as a "headwind" to Oracle's overall growth and profitability. The division has been plagued by waves of corporate layoffs, leading Wall Street analysts to widely view it as one of Oracle's worst acquisitions of all time. This downfall perfectly validated Judy Faulkner's long-term thesis to her clients: that a single, integrated platform built entirely through organic software development—backed by a company that promises it will never go public, never be acquired, and never buy another vendor—is the only stable choice for healthcare IT.

* **The Acquisition:** Oracle bought Cerner in December 2021 for $28 billion. * **Stagnant Financials:** Cerner's revenue had flattened around $5.5 billion and began declining prior to the acquisition. * **The Aftermath:** Oracle Health is now classified internally as a financial drag, resulting in widespread layoffs and negative analyst sentiment.

[02:49:56](https://www.youtube.com/watch?v=FNRU9ZNqcls&t=10196s) Hospital Consolidation and the "Grid of Care"

Looking toward the future, Epic's market outlook has never been brighter. International business is growing rapidly, now accounting for 10% to 15% of their total revenue, highlighted by their largest-ever single implementation in 2023 at London’s Guys and St. Thomas' NHS Foundation Trust. Furthermore, Epic is riding a massive structural tailwind in the US: the rapid consolidation of the healthcare industry. Spurred by the immense regulatory compliance burdens of the Affordable Care Act and heightened HIPAA penalties introduced under the HITECH Act, independent community clinics and local practices are rapidly merging into mega-giant regional hospital systems just to survive. Because these pseudo-monopoly hospital networks are the exact complex entities Epic optimized its software for decades ago, Epic is left as the only vendor capable of serving them.

With an iron grip on the provider network, Epic is now deploying its ultimate playbook: expanding into adjacent verticals like health insurance payers, home health, post-acute rehab, and pharma/biotech companies—a massive web they call the "grid of care" or "system of connectedness." A prime example is the automation of prior authorizations. Because both doctors and insurance companies despise the tedious, manual paperwork required to approve expensive drugs or procedures, Epic can step in and say, "We have all the data, we can just automate this for you." Epic's ultimate strategy remains unchanged: win a baseline contract with a hospital, identify every other piece of third-party IT software that the hospital uses, and systematically replace those vendors until Epic becomes the all-in-one enterprise software solution for the entire healthcare ecosystem.

* **International Footprint:** Global sales make up 10% to 15% of Epic's business, led by the massive 2023 rollout at London's Guys and St. Thomas' NHS Trust. * **Industry Tailwinds:** Increased compliance costs from the Affordable Care Act and strict HIPAA penalties have forced local practices to consolidate into massive hospital systems. * **The "Grid of Care":** Epic's forward-looking growth engine focused on cross-selling automated tools like prior authorizations directly to insurance payers and pharmaceutical companies.

[02:55:22](https://www.youtube.com/watch?v=FNRU9ZNqcls&t=10522s) The Launch of Cosmos and the Utopian Dream of Data

After saturating their primary hospital market, Epic began exploring new avenues to leverage their assets, looking into payers and pharmaceutical companies to see who else they could sell to. This brings the discussion to the modern ecosystem. While there are easily 50 things Epic does that won't fit into the episode, the hosts highlight one exceptionally cool project launched a few years ago: Cosmos. Epic realized that a massive amount of structured patient data was sitting across a scattered web of servers—some on-premise at hospitals, others on Epic's cloud, AWS, or Azure. They decided to anonymize this data and bring it into a single, highly queryable, Epic-hosted instance.

The scale of Cosmos is staggering, containing the anonymized data of 295 million patients and spanning 15 billion individual patient encounters. This completely changes the answer to whether Electronic Medical Records (EMRs) truly benefit patients. While Epic boasts that in 2023 its system prevented 66 million potential adverse drug interactions and 250,000 surgical errors, Cosmos represents the "utopian dream" long held by the Bush and Obama administrations: having all medical data in one centralized, actionable place.

The hosts illustrate this power by contrasting a paper-based world with a digital database. If something is wrong with a local water supply but all records are on paper, an outbreak goes unnoticed unless a massive research grant allows someone to manually pull individual files. By looking horizontally across a digital database, patterns emerge instantly—which is exactly how the water crisis in Flint, Michigan was uncovered. Cosmos takes that concept "on steroids," making the database completely free and accessible to any institution that contributes its own data.

- **Cosmos Database Size:** 295 million anonymized patients and 15 billion individual doctor-patient encounters. - **EHR Impact Study:** 45% of patients reported improved care quality with EHRs, while only 6% noticed a decline. - **Epic 2023 Safety Stats:** Automatically prevented 66 million potential adverse drug interactions and 250,000 potential surgical errors. - **Access Model:** Completely free to any Epic customer institution that chooses to opt in and contribute its own data.

[02:59:25](https://www.youtube.com/watch?v=FNRU9ZNqcls&t=10765s) Productizing "Lookalikes" and Prefiguring AI

Epic has productized Cosmos in a way that allows doctors to run real-time queries during care. If a physician encounters a patient with an incredibly rare or baffling condition, they can turn to Cosmos to see if it has ever happened to anyone else in the world, what the outcome was, and who treated them. The system surfaces these historical matches as anonymized "lookalike patients."

The hosts note that this is already happening frequently for rare diseases and complex cases, and it will become massively useful for clinical trials and medical research. While it is entirely free right now, they wonder how Epic might price and package it in the coming years. More than anything, this massive pool of structured data "precages" a major topic for later analysis: the unbelievable potential of training AI on this data.

[03:00:32](https://www.youtube.com/watch?v=FNRU9ZNqcls&t=10832s) The Business Today: Massive Leverage, Surprising Revenues

Taking stock of Epic's business scale, the company currently has 607 customers managing 3,200 hospitals between them. Insanely, they have never lost a single customer. They steadily add 10 to 25 new health systems every year. Their network encompasses 590,000 physicians, 495,000 staffed beds, and 325 million patients worldwide—280 million of whom are in the US. This means almost every single American has their healthcare touched by an Epic system at some point.

In 2024, Epic generated \$5.7 billion in revenue, representing a 13% year-over-year growth rate over the last five years, and a 16% jump from the previous year's \$4.9 billion. One host admits that Epic actually does "less revenue than I would have figured" given its reputation as the "central nervous system" of healthcare. Dividing revenue by customers equates to roughly \$10 million a year per customer. For a massive enterprise software system that an entire hospital relies on, Epic captures a surprisingly small minority of the total value created. For instance, a mid-sized health system requires 100 full-time internal employees just to administer Epic, resulting in massive internal and consulting spend that doesn't go to Epic itself.

The hosts debate whether this pricing is intentionally altruistic or just highly strategic. While Epic used to saturate their customers' willingness to pay, they now possess dominant pricing power but choose not to fully exploit it. The reason is a long-term orientation: they never want to give a customer an incentive to rebid or consider a competitor. It mirrors the classic Jeff Bezos quote about AWS in its infancy, where he wanted choosing a competitor to feel like an "irrational and irresponsible decision." To highlight the financial contrast, United Healthcare pulls in \$400 billion a year in revenue and generates \$35 billion in EBITDA. United Healthcare’s pure profit is six times Epic's entire top-line revenue, illustrating how Epic acts as an incredibly powerful industry lynchpin without commanding the largest profit pool.

- **Customer Count:** 607 customers across 3,200 hospitals, with zero historical customer churn. - **Annual Growth:** Adding 10 to 25 new health systems annually; 2024 revenue hit \$5.7 billion (up 16% from \$4.9 billion). - **Network Reach:** 590,000 physicians; 495,000 staffed beds; 280 million US patients (roughly 79% of the US population). - **Workforce & Margins:** 14,000 total employees; estimated EBITDA margins sit between 30% and 35%, yielding roughly \$1.7 billion in conservative EBITDA. - **Market Share:** 42% of all hospitals (skewed toward the largest, highest-volume centers) and 58% of ambulatory physicians.

[03:05:52](https://www.youtube.com/watch?v=FNRU9ZNqcls&t=11152s) Valuing the Most Durable Software Company in History

Valuing Epic is a bit of a "silly question" because it is impossible to sell shares to anyone other than the company itself, meaning there will never be a traditional liquidity event. However, the hosts point out that Forbes' 2021 estimate of founder Judy Faulkner’s net worth at \$7.6 billion—which implies a \$15 billion company valuation—is "wildly wrong" and "ludicrous."

If you apply a conservative 30x EBITDA multiple (the rate Oracle paid to acquire a struggling Cerner) to Epic’s estimated \$1.7 billion EBITDA, the company is worth at least \$51 billion. A standard public market software multiple of 9x revenue yields a similar \$50-ish billion figure. Yet, Epic's revenue streams are far more durable and growing much faster than Cerner's ever were. The hosts argue that Epic is likely "the most durable software company in history." If it were public, investors would easily value it in the \$100 billion neighborhood. This would peg Judy’s half-share of the company at \$50 billion, making her the wealthiest self-made woman in the world and placing Epic alongside giants like Shopify, ARM, Lockheed Martin, and Starbucks.

Furthermore, if a tech titan like Microsoft, Google, Apple, Amazon, or even Berkshire Hathaway had the opportunity to acquire Epic, a \$100 billion price tag would be a no-brainer. That amount represents a mere 4% of Apple or Microsoft's total value. Beyond the financial returns, the strategic value of owning Cosmos's 15 billion patient interactions for the modern AI landscape is immeasurable.

[03:09:09](https://www.youtube.com/watch?v=FNRU9ZNqcls&t=11349s) Judy's Trust, Succession Playbook, and the Cost of Integration

With Judy Faulkner now 81 years old, a leadership transition will inevitably occur in the coming decades. Judy owns roughly half of the company economically and holds 100% of the voting shares. She has signed the Giving Pledge and is progressively transferring her non-voting shares to her foundation, "Roots and Wings," which funds itself by selling those shares back to the company at an annually adjusted price.

Upon her death, Judy has announced that 100% of her voting control will transfer into an ironclad "purpose trust." This trust will be uniquely managed by three separate constituencies: her family (her husband Gordon and their three children), a group of five longtime senior Epic managers, and three CEOs from major Epic customer health systems. The trust's bylaws dictate two fascinating mandates. First, Epic can never be sold or taken public. Second, the next CEO must meet two strict criteria: they must be a longtime Epic employee, and they must be a software developer. The hosts express deep admiration for this "very Costco" or "Nike in its heyday" approach, praising the beauty of a company being run by someone who came from its core operational and design skill set.

Finally, looking to the future, the conversation turns to Epic's history of interoperability. While notoriously protective and resistant to outside partnerships or open integration in the past, a lot has changed. However, diving deeply into their modern integration methods would require "a whole another 4-hour podcast." Today, Epic utilizes many different integration programs, each carrying different standards, names, and revenue-sharing models that kick back cash to Epic. In rare cases of true partnership, Epic's integration even requires warrants to own a chunk of the partnering company.

- ** Succession Trust Structure:** Governed by 3 family members, 5 senior Epic managers, and 3 customer CEOs. - **CEO Prerequisites:** Must be both a veteran Epic employee and a software developer. - **Partnership Terms:** Modern external integrations often require revenue shares paid back to Epic, or financial warrants granting Epic equity ownership in the partner company.

[03:11:51](https://www.youtube.com/watch?v=FNRU9ZNqcls&t=11511s) The Business "Power" of Epic Systems

The conversation shifts smoothly into a formal business analysis, utilizing the framework from Hamilton Helmer's book *Seven Powers*. The speakers explain that "power" in this context refers to the specific attributes that allow a company to achieve persistent differential returns and maintain sustainable profitability over its closest competitors. They map out how Epic Systems transitioned from having little to no structural power during its extended takeoff phase to wielding a massive amount of power as a scale incumbent today.

During the early takeoff phase, Epic's eventual primary differentiator—having a single, unified platform—did not actually matter much. Because the market was so young, computers were not widely adopted by hospitals, and everyone else was just starting out. Instead of traditional business power, Epic effectively used time as a resource. While most entrepreneurs feel pressured by their capital structures to launch quickly and rely on third-party applications, Epic was content to remain a small, self-sustaining business for a long period, slowly building out a massive software suite until the market was mature enough to demand it.

Today, however, Epic possesses an extraordinary amount of power, driven primarily by high switching costs. The hosts joke that Epic is essentially "an ERP on steroids" and arguably represents the single biggest switching cost piece of software in existence. It functions as the core nervous system of a hospital network, with implementations routinely costing hundreds of millions or even billions of dollars. Crucially, the stakes are incomparably high; if an enterprise resource planning (ERP) or customer relationship management (CRM) system goes down, it is an inconvenience, but if an electronic medical record (EMR) system goes down, "people die."

- **The Seven Powers Framework:** Counterpositioning, scale economies, switching costs, network economies, process power, branding, and cornered resource. - **Scale Economies:** Epic can amortize its massive, ongoing investments in product breadth across a giant pool of existing customers, allowing them to bundle in free software updates that a new market entrant could never afford to develop from scratch. - **Network Economies (The "Care Everywhere" Effect):** Epic achieved a tipping point where hospitals now adopt it because other neighboring hospitals use it, allowing seamless data interoperability for patients and shared medical records. - **The Talent Network Economy:** Hospitals preferentially choose Epic because it is the industry standard that incoming doctors and medical staff are already trained on and know how to use. - **Branding as Risk Mitigation:** Epic has captured the most prestigious medical institutions, establishing a "don't get fired for buying IBM" tier of trust where clients know the system will definitively work. - **Defensive Process Power:** Epic trains its young developers fresh out of college in specialized, non-transferable internal systems and languages licensed from InterSystems, such as Caché and MUMPS, meaning their skills do not easily port over to competitors like Cerner.

---

[03:19:28](https://www.youtube.com/watch?v=FNRU9ZNqcls&t=11968s) Research, Mountain Climbing, and the Law

Analyzing Epic’s operational playbook reveals a staggering commitment to internal research and development. Drawing on data from Arvin Navaratnam at Worldly Partners, the hosts highlight that Epic allocates 35% of its operating expenses to R&D. Because Epic famously employs zero traditional sales and marketing teams—aside from a tiny handful of about eight headcount classified as salespeople—and maintains low general and administrative (G&A) overhead, they can aggressively compound their software advantage. This internal R&D strategy allows them to build everything strictly from scratch rather than buying outside companies.

The conversation takes a philosophical turn regarding Epic's growth mindset. Founder Judy Faulkner once described her strategy as climbing a mountain by focusing entirely on the next hill directly in front of her rather than trying to map out the entire peak at once. This closely mirrors Silicon Valley icon Paul Graham's famous 2012 essay *Startups Echo Growth*. Graham argued that weekly hill-climbing forces essential action over procrastination, and that good startup ideas are almost always adjacent to even better ones. The hosts marvel at how Faulkner independently extracted these foundational tech principles in Wisconsin without ever reading Graham's work or following traditional MBA rulebooks.

The co-hosts also address Epic’s fierce, protective litigation strategy, which stands in stark contrast to their customer relations. Epic is highly aggressive when it comes to defending its intellectual property against patent trolls and enforcing strict employment non-compete clauses. They even won a landmark U.S. Supreme Court case regarding forced employee arbitration. Currently, they are locked in a massive antitrust lawsuit with a startup called Particle Health over alleged violations of the Sherman Antitrust Act. Yet, remarkably, Epic has a sacred boundary: they have never engaged in a lawsuit or mediation with an actual customer, keeping their corporate culture laser-focused on client satisfaction.

- **R&D Expense Comparisons:** Epic's 35% R&D spend rivals tech giants like Apple (36%) and Amazon (28%), easily outstripping healthcare tech competitors like Athenahealth (10%) and Oracle overall (23%). Google is one of the few to beat them at 45%. - **Capital Structure Resonance:** The hosts muse that if Y Combinator instructed founders to build independently without venture capital instead of raising massive funds, the tech landscape would have far more self-sustaining "Epic Systems" running around. - **Litigation Moat:** Epic aggressively uses the legal system against external tech competitors and former employees to deepen its competitive moat and keep its proprietary data secrets locked down.

---

[03:25:26](https://www.youtube.com/watch?v=FNRU9ZNqcls&t=12326s) Enterprise Sales on Healthcare Steroids

The segment concludes with a deep dive into why consumer-style software mechanics fail to disrupt the healthcare landscape, a phenomenon the hosts call "enterprise sales on steroids." Standard enterprise software is already notorious for grueling sales cycles, massive customization needs, long implementation windows, and uninspired user interfaces filled with endless dropdowns and buttons. These visual design limitations happen because the functional breadth required by enterprise software is too vast to constantly redesign alongside modern consumer UI trends.

When enterprise software meets the healthcare sector, the complexity intensifies exponentially due to rigid regulatory compliance, extreme data sensitivity, and the literal life-or-death stakes of medical operations. This environment completely suffocates popular modern tech strategies like "product-led growth" (PLG) or bottoms-up user adoption, which allowed tools like Slack and Figma to spread organically within corporations.

The host references insights from a prominent health API writer on Substack to explain the legal wall preventing grassroots software adoption in medicine. Under Health Insurance Portability and Accountability Act (HIPAA) regulations, any software handling patient data requires a Business Associate Agreement (BAA). A BAA cannot simply be clicked through or signed by an individual doctor or a single clinic practice manager; it must be executed by an executive with the specific, legal authority to bind the entire multi-billion-dollar healthcare organization. This legal barrier completely blocks individual providers from testing out or adopting nimble, bottoms-up software alternatives.

[03:27:04](https://www.youtube.com/watch?v=FNRU9ZNqcls&t=12424s) The Wacky Brilliance of Judy Faulkner

The enterprise healthcare market operates on a completely different playbook than standard B2B SaaS. In healthcare, product-led growth (PLG) is a massive uphill battle; you cannot expect a small team within a giant corporation to simply buy software on a credit card. Instead, these are strictly top-down CIO sales. At the helm of Epic is Judy Faulkner, whom the hosts affectionately describe as a "wacky, wacky founder," placing her in the same tier of singular, obsessed entrepreneurs they frequently study. Judy famously mirrors Warren Buffett's joke that she is more likely to die than retire. Her absolute focus is on making Epic great, exhibiting the same customer-feedback obsession as Jeff Bezos and a deep conviction that "the nerds will prevail in the long run."

Judy is almost certainly the most successful female founder in history by any metric. While companies like Estée Lauder hit a $100 billion market cap during the pandemic, their value has since cratered to around $20 billion. Other cultural icons like Oprah Winfrey and Taylor Swift—whose enterprise value the hosts tracks with a casual, ongoing "Taylor Inc. calculator"—don't quite reach the top spot, with Taylor Inc. estimated at around $11 billion. Currently, Forbes considers Diane Hendricks, the founder of the massive roofing supplier ABC Supply, to be the wealthiest female founder in America. Remarkably, this means the two most successful female founders in the country are both based in Wisconsin. However, because ABC Supply operates at much lower margins on its $20 billion in annual revenue, Epic's valuation comfortably clears it at a minimum of $100 billion based on what an acquirer would realistically pay. When challenged to put their money where their mouth is, the hosts admit they would eagerly buy and hold Epic shares for life at a $100 billion valuation if they could receive profit distributions, though they wouldn't touch it if they were banking on a traditional liquidity event.

* **Judy Faulkner's Tech Philosophy:** Strongly believes that "the nerds will prevail in the long run." * **Wisconsin Powerhouses:** Both Judy Faulkner (Epic) and Diane Hendricks (ABC Supply) are based in Wisconsin. * **Taylor Inc. Calculator:** An ongoing banter metric between the hosts tracking Taylor Swift's enterprise value at approximately $11 billion. * **Epic Valuation:** Pegged at a minimum of $100 billion based on hypothetical acquisition willingness.

---

[03:30:26](https://www.youtube.com/watch?v=FNRU9ZNqcls&t=12626s) The Power of Founder Continuity and the Bear Case for Epic

The sheer impact of founder continuity is staggering when looking at Epic's 47-year history under a single leader. For context, Jensen Huang has been running Nvidia for 31 years, and Mark Zuckerberg has led Facebook for 21 years. Aside from Warren Buffett’s 50-plus year tenure at Berkshire Hathaway, a company scaling to Epic's magnitude while preserving the exact personality of its founder for nearly five decades is incredibly rare. Judy remains highly engaged and actively runs the daily operations, drawing a direct parallel to Jensen's hands-on style at Nvidia.

Shifting the conversation to the bear and bull cases, the hosts dissect the structural vulnerabilities Epic might face. The company has historically feasted on massive US healthcare budgets fueled by a large population and a bloated healthcare system. If future growth depends heavily on international expansion, Epic will face a much lower willingness to pay for healthcare administration software abroad. Furthermore, antitrust threats loom large. A current "particle suit" alleging violations of the Sherman Act could be a company-changing event if it gains material traction. Finally, recent legislation via the Cures Act explicitly targets "information blocking." This makes technologies like screen scraping, Chrome extensions, and robotic process automation (RPA) entirely legal mechanisms for extracting data from Epic, bypassing Epic's traditional data-privacy lockouts. While a nightmare scenario involves a third-party UI extension capturing 70% of Epic's user base and eventually swapping out the backend entirely, the hosts agree the actual probability of a competitor displacing Epic this way is "near zero."

* **Founder Tenures for Comparison:** Judy Faulkner (47 years), Warren Buffett (50+ years), Jensen Huang (31 years), Mark Zuckerberg (21 years). * **International Headwinds:** Lower global willingness to pay for complex healthcare administration. * **The Cures Act:** Landmark legislation making data extraction techniques like screen scraping and RPA legal, loosening Epic's grip on data isolation.

---

[03:33:57](https://www.youtube.com/watch?v=FNRU9ZNqcls&t=12837s) Best of Breed vs. The Microsoft Playbook

Even if a shift toward healthcare interoperability allows "best of breed" point applications to challenge Epic's all-in-one integrated paradigm, Epic has a highly effective defense strategy: the Microsoft playbook. If a hospital CIO wants a specialized application, Epic can simply offer a comparable module bundled directly into their existing enterprise license for a single price, neutralizing the need for outside vendors. Looking at more existential "handwavy" threats, paradigm shifts like AI or a massive transition to value-based care could theoretically alter what a health system looks like. Yet, even if new care delivery models emerge where patients rarely visit traditional doctors, it remains incredibly difficult to imagine a world where a massive foundational system like Epic isn't required.

* **The Microsoft Playbook:** Bundling auxiliary modules into the core enterprise license to crowd out point solutions. * **Paradigm Shift Threats:** AI integration, value-based care, and evolving clinical delivery models.

---

[03:35:16](https://www.youtube.com/watch?v=FNRU9ZNqcls&t=12916s) The Bull Case and the $800 Billion Healthcare Waste Pool

The true bull case for Epic rests on its flawless track record of cross-selling. They successfully expanded from core EMR and billing software into endless clinical specialties and administrative modules. Now, Epic is leveraging its dominant position to stretch its customer base entirely outside of traditional health systems, targeting pharma companies, insurance payers, and researchers. Prior authorization handling represents an obvious, lucrative frontier. While health systems are notoriously low-margin customers, insurance companies have deep pockets—even if it makes the hosts "philosophically" uncomfortable to watch profit pools grow in a deeply flawed system.

The baseline reality of American healthcare is defined by staggering inefficiency. Citing research, the hosts point out that roughly 30% of US healthcare spending—amounting to a jaw-dropping $800 billion—is pure waste. To put it in perspective, "the entire GDP of Switzerland's worth of waste exists in our health care system." Consumers acutely feel this extraction. One host shares a striking anecdote about paying $25,000 to $30,000 annually in family insurance premiums, noting that over any 10-year period, the sheer out-of-pocket reality never justifies the entirely "made up" top-line and negotiated rates on medical bills. Consumers simply are not getting a good deal, which paves the way for software that can cut through the administrative bloat.

* **Waste in the System:** Approximately $800 billion—equivalent to the entire GDP of Switzerland—is lost to waste annually in US healthcare. * **New Customer Vectors:** Expansion into pharmaceutical companies, insurance payers, and research institutions.

---

[03:38:05](https://www.youtube.com/watch?v=FNRU9ZNqcls&t=13085s) Ambient AI and the Future of the EHR Operating System

The most tangible technological catalyst for Epic is ambient listening. The single biggest complaint from physicians is doctor burnout caused by endless "clicky-clicky" data entry, dropdown alerts, and screen time. While some hospitals hire human scribes, doctors still find themselves trapped behind monitors. Ambient AI scribes promise to completely revolutionize this workflow by listening to patient encounters, categorizing data, and structuring notes automatically. Epic has positioned itself as the industry's ultimate gatekeeper for this technology. They currently partner with major players like Microsoft (which acquired voice-recognition pioneer Nuance and its Dragon dictation tech) alongside high-flying startups like Abridge and Suki. Because Epic acts as the central choke point determining which breakthrough innovations actually reach hospital workflows, they stand to benefit enormously via partnerships, revenue splits, warrants, or by eventually building a proprietary version.

The ultimate "mega bull case" portrays a utopian future where the user-facing EHR completely fades into the background. A real CIO from a major hospital system pitched this exact vision to the hosts: as ambient AI perfects data capture, it could directly ship clinical and billing data straight to insurance payers or the government for instant adjudication. In this world, the massive administrative bloat of healthcare is quietly absorbed by an AI operating system. While the hosts express healthy skepticism about administrative jobs and costs ever truly evaporating—noting "I'll believe it when I see it"—Epic is flawlessly positioned to become that very operating system. Combined with a zero-churn install base where customers simply never leave, Epic has reached an incumbent status where even heavy government regulation acts as a defensive moat.

* **Ambient AI Partners:** Microsoft/Nuance, Abridge (based in Pittsburgh), and Suki. * **The AI Operating System Vision:** A theoretical future where user-facing EHR software vanishes, and ambient AI handles everything from clinical documentation to automated insurance adjudication.

[03:42:45](https://www.youtube.com/watch?v=FNRU9ZNqcls&t=13365s) Regulatory Moats and Platform Futures

The conversation opens with a discussion on how regulation fundamentally impacts market dynamics, noting that compliance frameworks naturally "shut out new entrance and entrench the older companies". Because established corporations possess the sheer scale and resources to comply quickly, they hold an overwhelming advantage. A prime example in the healthcare sector is TEFA, an upcoming data interoperability framework. The hosts point out that Epic is perfectly positioned to be the first to implement it smoothly, further solidifying its dominant market presence.

Turning to Epic's future strategy, the hosts touch on the company's significant pricing headroom. While squeezing this could extract massive value and essentially "pull forward some of the future into today," they both express heavy skepticism that Epic would ever exploit its customers this way. Instead, the real curiosity lies in whether Epic will transition into a true platform layer. Despite minor past efforts like App Orchard, Epic historically built everything itself. The hosts wonder if they will ever open up for other applications to build on top of them, though they acknowledge that in vertical software, this standard tech playbook might simply not apply.

---

[03:44:07](https://www.youtube.com/watch?v=FNRU9ZNqcls&t=13447s) Quintessence: Why Vertical Software Won in Healthcare

Transitioning to their "quintessence" section, the hosts tackle the ultimate question: why did building an entire platform and every single application completely in-house uniquely work for Epic? In almost any other industry, an investor pitched with this kind of closed strategy would immediately pass and call it a stupid idea. The answer, they realize, lies in the distinct playbook of Vertical Market Software (VMS). Unlike horizontal giants like Slack, Salesforce, or Microsoft Office—which must avoid catering too deeply to any single customer segment to prevent bad design—VMS thrives on going exceptionally deep into one industry's operational problems. "When you're building vertical market software... you absolutely want to listen to your customers," which is Epic’s greatest superpower.

Adding another layer, the hosts emphasize that healthcare introduces severe "sticks" rather than carrots. If horizontal software fails, an email goes undelivered; if healthcare software fails, "they die." This absolute intolerance for operational risk makes a single-vendor, fully integrated approach the only logical choice to prevent data leaks or the catastrophic loss of Medicare and Medicaid subsidies. This high-stakes environment explains why a single-vendor playbook was correct, even if it took 47 years to build the dominant empire.

Furthermore, the hosts marvel that Epic is likely the most valuable vertical market software company in the world. Usually, VMS companies are strictly limited by their niche, but the American healthcare market is uniquely massive. The only other peers that operate in this tier are Bloomberg and the public giant Viva Systems—though Viva is valued at a comparatively modest $20 billion market cap. Constellation Software is also mentioned, though it functions as a roll-up conglomerate of various smaller verticals rather than a single organic monster.

- **TEFA:** A new data interoperability framework Epic is poised to implement first. - **Viva Systems:** A healthcare VMS peer with roughly a $20 billion market cap, which the hosts note is valued far below Epic. - **The Core Difference:** Horizontal software requires synthesizing varied feedback to avoid niche traps; vertical software requires building exactly what the core customer (the hospital CIO and CFO) demands.

---

[03:49:25](https://www.youtube.com/watch?v=FNRU9ZNqcls&t=13765s) Carveouts: Rally Cars, the Switch 2, and Brat Summer

Wrapping up the main discussion, the hosts pivot to their traditional "carveouts" segment to share personal recommendations. The first host offers a "today" carveout and a "preemptive" one. His first recommendation is a classic 2012 YouTube video featuring the late rally car driver Ken Block tearing through the closed-off streets of San Francisco. Comparing it to the iconic movie *Bullitt* with Steve McQueen, he highlights the awe of watching a 600-horsepower rally car drift past familiar city blocks. Digging into Block's history, he shares the surprising fact that Block was actually a co-founder of DC Shoes before pivoting to professional racing. His preemptive carveout is the highly anticipated Nintendo Switch 2, praising Nintendo's decision to preserve full backwards compatibility and online accounts, and expressing pure excitement to finally play the upcoming Mario Kart with his older daughter.

The second host counters with three recommendations of his own. First is a rewatch of the movie *Knives Out* starring Daniel Craig, describing it as "pure popcorn" and visually stunning. Next, he jokingly admits he might be the last person on earth to discover it, but confesses his intense obsession with Charli XCX's hit 2024 album *Brat* after watching her perform at the Grammys. The first host laughs, joking that his own music taste is "frozen in amber at like 2006," while the second host guarantees *Brat* will make listeners feel like they are at "a much cooler party than you've been invited to in 20 years." Finally, he plugs ODESZA's new remix album, *Music to Refine To*, which features tracks from the Apple TV+ show *Severance*.

---

[03:53:27](https://www.youtube.com/watch?v=FNRU9ZNqcls&t=14007s) Closing Credits, Show Notes, and Special Thanks

The episode draws to a close with a reminder for listeners about the upcoming Acquired live event in New York City on July 15th, directing fans to their event link to get the latest updates. The hosts extend their gratitude to their seasonal sponsors: JP Morgan Payments, ServiceNow, Crusoe, and Fundrise. They emphasize that this deep dive required interviewing dozens of experts across the medical and tech industries to piece together Epic's notoriously private history.

In an extraordinary turn of events for the show, the hosts reveal that because there is no existing long-form canonical piece written about Epic, they cold-emailed the legendary founder Judy Faulkner directly. To their delight, she responded and granted them an interview. They express immense gratitude to Judy, along with Carl D'vorak, Sumit Rana, and Lee Leavon from Epic, for walking them through the company's history. They conclude by directing listeners to check out their Novo Nordisk episode, their ACQ2 interview with ServiceNow CEO Bill McDermott, and inviting everyone to join the conversation in the Acquired Slack channel.

- **NYC Live Event Date:** July 15th at acquired.fm/nyc. - **Key Industry Thought Partners Thanked:** Arvin Navaratnam (Worldly Partners), Brendan Keeler (Health API Guy Substack), Brian Lawrence (Oakcliffe Capital), Lan Wang (JSL Health Capital), Scott Gaines, Patrick Wingo (Eleon), John Bertrren (Digital Diagnostics), and Dr. Abed Mir. - **Medical Insiders Highlighted:** Mike Pfeffer (CIO of Stanford Medicine), Dr. Molly Cantor (UCSF), and Dr. Karine Davala (Harvard MGH).

## Notable quotes > [33:32](https://www.youtube.com/watch?v=FNRU9ZNqcls&t=2012s) The sun was shining. I was disattentive. I was just sitting there and suddenly it all came to me. Here's how you build it. The integrated system. — Judy Faulkner

> [35:11](https://www.youtube.com/watch?v=FNRU9ZNqcls&t=2111s) we don't need to build a whole company around this thing. — Judy Faulkner

> [01:01:14](https://www.youtube.com/watch?v=FNRU9ZNqcls&t=3674s) Hey, you know, this is good that I'm working on this, but I'm kind of bored. Can you give me something hard to do? — Sumit Rana

> [01:15:14](https://www.youtube.com/watch?v=FNRU9ZNqcls&t=4514s) Like we've got this thing called Epic up in Portland... and it's way better than the stuff that you're trying to build. — Physicians

> [01:22:40](https://www.youtube.com/watch?v=FNRU9ZNqcls&t=4960s) Guys, no, we need to model out in Excel exactly what Kaiser's transaction flow is going to be — Carl Dvorak

> [01:26:12](https://www.youtube.com/watch?v=FNRU9ZNqcls&t=5172s) No, we're not going to do it. We're not going to do it for you. We're not going to do it for anybody. — Judy Faulkner

> [01:55:09](https://www.youtube.com/watch?v=FNRU9ZNqcls&t=6909s) I don't think you're ready to be a customer yet — Epic Systems representative

> [02:00:18](https://www.youtube.com/watch?v=FNRU9ZNqcls&t=7218s) wildly succeeded at digitization of the industry. It did absolutely nothing on digital transformation — Hospital CIO

> [03:30:26](https://www.youtube.com/watch?v=FNRU9ZNqcls&t=12626s) the nerds will prevail in the long run. — Judy Faulkner

## Mentioned Epic Systems, MyChart, Acquired, Ben Gilbert, David Rosenthal, Judith Faulkner, GE, Microsoft, Google, J.P. Morgan Payments, ACQ2, Judy Greenfield, Lou Greenfield, Dell Greenfield, DuPont, Taylor Swift, Bell Labs, Lockheed, Siemens, South Jersey Peace Center, Oregon Physicians for Social Responsibility, Dickinson College, University of Rochester, Bill Gates, Steve Jobs, Stanford University, University of Wisconsin, Dr. Warner Slack, ENIAC, IBM, Massachusetts General Hospital, American Medical Association, American Hospital Association, Association of Health Record Librarians of North America, Rockefeller Foundation, Standard Oil, National War Labor Board, FDR, Truman, Kennedy, CO-STAR, National Institutes of Health, MUMPS, Neil Papalardo, Meditech, Cerner, Caché, Chronicles, Resolute, Cosmos, Stork, Beans, Human Services Computing, Data General Eclipse, DEC, IKEA, Sequoia Capital, American Girl Doll Company, Pleasant Rowland, Mattel, Cadence, Sumit Rana, Carl D'Vorah, Carl Dvorak, Fundrise, Kaiser Permanente, Neil Patterson, Philips, Philips Externity Enterprise, Deep Space, Jeff Gotney, Rush University System for Health, Epicare EMR, Jeff Bezos, Service Now, George W. Bush, Barack Obama, Allscripts, NextGen Healthcare, Rusty France, American Prospect, Johns Hopkins, Cedars-Sinai, UCSF, Misys, Eclipsys, Veradigm, LSS Data Systems, Siemens Health Services, Crusoe, Nvidia, Founders Fund, Department of Defense, Veterans Administration, Siemens Healthineers, Partners Healthcare, Harvard, Mayo Clinic, Intermountain Health, Common Spirit Health, Cambridge, US News & World Report, Oracle, Oracle Health, Affordable Care Act, HIPAA, HITECH Act, Guys and St. Thomas' NHS Foundation Trust, AWS, Azure, Bush administration, Obama administration, United Healthcare, Forbes, Apple, Amazon, Berkshire Hathaway, Gordon Faulkner, Roots and Wings, Costco, Nike, Hamilton Helmer, Seven Powers, InterSystems, Arvin Navaratnam, Worldly Partners, Athenahealth, Paul Graham, Startups Echo Growth, U.S. Supreme Court, Particle Health, Sherman Antitrust Act, Y Combinator, Slack, Figma, Substack, Warren Buffett, Estée Lauder, Oprah Winfrey, Taylor Inc., Diane Hendricks, ABC Supply, Jensen Huang, Mark Zuckerberg, Facebook, Sherman Act, Cures Act, Chrome, Nuance, Dragon, Abridge, Suki, TEFA, App Orchard, Salesforce, Microsoft Office, Medicare, Medicaid, Bloomberg, Viva Systems, Constellation Software, Ken Block, DC Shoes, Steve McQueen, Bullitt, Nintendo Switch 2, Mario Kart, Knives Out, Daniel Craig, Charli XCX, Brat, ODESZA, Music to Refine To, Severance, Apple TV+, Lee Leavon, Brendan Keeler, Health API Guy Substack, Brian Lawrence, Oakcliffe Capital, Lan Wang, JSL Health Capital, Scott Gaines, Patrick Wingo, Eleon, John Bertrren, Digital Diagnostics, Dr. Abed Mir, Mike Pfeffer, Stanford Medicine, Dr. Molly Cantor, Dr. Karine Davala, Novo Nordisk, Bill McDermott

[Source](https://www.youtube.com/watch?v=FNRU9ZNqcls)

中文简报

Epic Systems通过信托架构实现57亿美元营收与500亿美元估值 Epic Systems通过维持统一的数据库架构以及防止公开上市或被企业收购的严格私人信托，主导了医疗健康软件市场。本期节目探讨了Epic Systems，这家自筹资金起家的企业级软件巨头扮演着美国医疗健康行业中枢神经系统的角色。节目记录了创始人Judy Faulkner的历程，讲述她如何从编写单个电脑程序起步，并借助名为Chronicles的统一数据库架构赢得如Kaiser Permanente等大型合同。主持人们还探讨了2009年《高科技法案》（HITECH Act）等联邦立法如何加速行业数字化、Epic坚决不进行收购的运营策略，以及旨在确保公司永久保持独立性的严密接班计划。 Epic Systems是一家拥有47年历史、总部位于威斯康星州乡村地区的私营医疗软件公司，具有极高的韧性，客户留存率接近完美，估计年营收已接近60亿美元。 (0:00) 该公司在创始人Judy Faulkner设立的独特信托架构下运营，确保其永远不会公开上市、永远不会被收购，并始终保持为一个独立实体。 (0:00) Epic依赖基于MUMPS衍生语言构建的中心化单核心数据库引擎Chronicles，这使其在原生数据集成方面比竞争对手具备巨大的架构优势。 (33:32) Epic仅凭7万美元银行贷款和7万美元早期天使投资自筹资金完成了整个全球扩张，完全没有引入外部风险投资。 (39:55) 与传统软件巨头不同，Epic几乎没有正式的销售或市场部门，而是依靠高端市场的自然需求以及自上而下的CIO企业关系开展业务。 (46:05)

Reading summary

Hosts Ben Gilbert and David Rosenthal argue that Epic Systems achieved near-permanent dominance in American healthcare IT by pursuing an insular, organically developed single-database architecture while strictly rejecting venture capital, acquisitions, and standard enterprise software conventions. Architectural Unity and In-House Product Development (33:32) Epic's primary competitive advantage stems from Chronicles, an integrated database conceptualized by founder Judy Faulkner in the mid-1970s. Rather than expanding through acquisitions, Epic built its entire product ecosystem—including the Epicare electronic medical record system, the Resolute billing module, and the MyChart patient portal—around this single backend. According to the hosts, this unified architecture eliminated the brittle data integrations that crippled rivals. By contrast, Epic's chief historical competitor, Cerner, accumulated a fragmented portfolio of 36 acquired companies, including Siemens Healthineers' electronic health records unit. Following the death of Cerner founder Neal Patterson in 2017 and severe implementation struggles with multi-billion-dollar military and Department of Veterans Affairs contracts, Cerner's growth stalled. Oracle acquired Cerner in December 2021 for $28 billion, but corporate leadership subsequently characterized the asset as an operational headwind plagued by layoffs and flat revenues. Governance, Non-Negotiable Standards, and Operational Discipline (100:54) Epic's operational playbook rejects prevailing corporate orthodoxy. Led continuously by Faulkner for 47 years, the privately held firm employs over 14,000 workers entirely in person on a 1,000-acre campus in Verona, Wisconsin. Its core corporate commandments—explicitly displayed across the campus—include rules against pursuing an initial public offering or engaging in mergers and acquisitions. Epic employs virtually no dedicated sales or marketing staff, enforces a zero-bug engineering workflow where original authors must immediately resolve defects, and filters applicants using rigorous, systemized aptitude assessments. This discipline translates into extraordinary commercial leverage. Epic accepts only 10 to 30 new enterprise clients annually, routinely rejects prospective hospitals that refuse to adopt standardized configurations, and refuses to negotiate prices or offer stock equity. Despite charging premium prices, Epic maintains a near-flawless retention record, having never permanently lost any of its 600-plus hospital systems. Regulatory Accelerators and Healthcare-Specific Market Realities (136:33) Epic's ascent was accelerated by federal policy shifts. The 2009 HITECH Act mandated "Meaningful Use" requirements, driving U.S. hospital electronic health record adoption from 9% in 2009 to 95% by 2014. Because the legislation provided public stimulus funds without enforcing cross-vendor data interoperability standards, risk-averse hospital executives overwhelmingly selected Epic's expensive but dependable enterprise suite to ensure compliance. The unique legal dynamics of U.S. healthcare further insulated Epic from conventional software disruption. Bottom-up "product-led growth" models commonly seen in enterprise software cannot take root because Health Insurance Portability and Accountability Act (HIPAA) rules mandate institutional Business Associate Agreements before applications can touch clinical records. Furthermore, ongoing regional healthcare consolidation pushes acquired community clinics onto the acquiring hospital network's centralized Epic system, consolidating Epic's dominance. Emerging Frontiers, Commercial Leverage, and Structural Risks (210:26) Epic is expanding from provider software into adjacent domains, including health insurance workflows, post-acute care, and clinical research through its Cosmos platform, which indexes anonymized "lookalike" patient histories. In artificial intelligence, Epic is positioning itself as an essential distribution gatekeeper for ambient voice scribing, partnering with Microsoft and specialized startups such as Abridge and Suki to automate clinical documentation. The hosts argue that conventional market valuations severely underestimate the firm; while Forbes pegged Faulkner's net worth at $7.6 billion in 2021 (implying a $15 billion valuation), the hosts estimate Epic's enterprise value between $50 billion and $100 billion based on an estimated $1.7 billion in EBITDA. However, material structural risks remain, including ongoing federal antitrust litigation under the Sherman Act, statutory crackdowns on data "information blocking" under the Cures Act, and diminished willingness to pay for complex administrative software across international healthcare markets. The hosts conclude that while Epic solved the monumental administrative problem of healthcare digitization, it produced an expensive, closed system optimized for insurance billing rather than holistic digital transformation. Epic's continued endurance depends on whether ambient AI can relieve clinician administrative burdens before regulatory antitrust pressures or enforced interoperability weaken its centralized platform power.

精炼阅读摘要

播客节目《Acquired》指出，Epic Systems 凭借完全自主研发的单一底层数据库、坚决不上市且拒绝外部并购的极致垂直经营模式，以及由高转换成本锁定的客户网络，成为了年营收接近60亿美元、实际主导美国医疗信息基础设施的非上市科技巨头。主持人认为，在医疗这种对软件故障零容忍的严苛行业中，专注于深度客户需求与极高系统稳定性的垂直软件战略，构建了现代商业中最牢固的护城河之一。单一架构与自研路线构筑底层壁垒 (33:32) 1970年代中期，创始人 Judy Faulkner 设计开发了名为 Chronicles 的核心数据库。主持团队指出，尽管最初编写的底层代码如今已全部更新，但 Chronicles 至今仍是 Epic 旗下数百款临床与计费应用（包括门诊系统 Epicare、计费系统 Resolute 和海量数据平台 Cosmos 等）唯一的共享底层架构。与依赖外部并购拼凑系统的大型软件供应商截然不同，Epic 在其47年历史中坚持不进行任何收购，所有功能模块均采取100%自主研发的方式推进。这种高度一致的架构赋予了 Epic 关键的产品一体化优势。医疗机构内部的各个科室与管理部门无需在割裂的多套系统间充当“人工中间件”去手动核对信息，所有数据都能在该单一数据库中无缝流转。节目分析强调，这种在初期看似缓慢且重资产的自研模式，伴随着个人电脑的普及与现代医院规模的剧烈膨胀，最终反超了依赖资本拼装的传统软件企业，形成了业内罕见的高客户留存率与数据一致性。反传统的企业文化与高议价运营策略 (100:54) Epic 建立了一套极度非传统的管理体系。其园区公共洗手间内公开张贴的“十大戒律”明确将“不得上市”与“不得并购或被并购”列为前两条，以此向员工和来访客户确立系统将永久由内部自研、长期保持架构一致的预期。公司完全拒绝远程办公，在威斯康星州 Verona 建造了充满奇幻主题的万人园区，并实行严苛的“零缺陷”（Zero-Bug）研发机制与人才筛选机制，确保代码即使在极高的医嘱合规要求下也不产生重大事故。在商业合作中，Epic 展现出罕见的强势与议价主导权。2000年代初竞标大型医疗组织 Kaiser Permanente 的历史性合同时，竞争对手 Cerner 主动提供10%的股权以争取订单，而 Faulkner 坚决拒绝让渡任何股权，但最终仍凭借系统稳定性中标。此战后 Epic 营收迅速攀升，并在2007年突破5亿美元。在后续运营中，Epic 始终强制推行“标准安装包”以避免系统过度定制导致的升级困难，且每年仅接纳极少数新客户，甚至主动拒绝不愿遵守其部署标准的医院。政策红利催化与竞争格局的彻底分化 (136:33) 2009年美国通过《经济与临床健康信息技术法案》（HITECH Act），以“有效使用”（Meaningful Use）奖惩机制为杠杆，推动全美医院电子病历普及率在2009至2014年间由9%飙升至95%。节目主持人指出，由于医院大多使用政府补贴资金采购软件，风险规避型的决策者普遍倾向于选择市场上最昂贵、最具声誉且绝不失败的成熟方案，Epic 借此大举整合了全美大型医疗机构市场，即便该法案在降低医疗总开支和推动跨系统数据互通方面引发了诸多批评与争议。与此同时，Epic 与主要竞争对手 Cerner 的竞争格局走向彻底分化。在2015与2017年美国国防部及退伍军人事务部（VA）总额超百亿美元的军方医疗 IT 竞标中，Epic 竞标失利，但节目指出这反而让其避开了极度繁琐的联邦官僚泥潭。反观 Cerner，在赢得大单后陷入多起并购整合消化不良、功勋创始人离世与管理层频繁更迭的危机。最终 Oracle 于2021年以280亿美元收购 Cerner，并在后续财报中将其视为拖累整体增长的业务板块。垂直市场壁垒、AI前景与反垄断风险 (191:51) 从商业模式角度分析，Epic 体现了极端垂直市场软件（VMS）的规律。在医疗这一涉及生命安危与严格 HIPAA 隐私合规的领域，消费级软件常见的“产品驱动增长”（PLG）机制完全失效，极高的转换成本使得签约医院几乎无法更换供应商。主持团队估算，若参照 Oracle 收购 Cerner 约30倍 EBITDA 的交易倍数或成熟软件公司的估值逻辑，Epic 的潜在估值可达500亿至1000亿美元区间。面向未来，Epic 正在利用沉淀海量病患记录的 Cosmos 数据库推动临床匹配，并构建连接保险机构的“照护网络”（Grid of Care）以自动化完成事先授权流程。在 AI 层面，Epic 联合微软/Nuance、Abridge 等环境智能语音（Ambient AI）公司，尝试直接从医患对话中生成结构化病历以减轻医生负担。然而，节目也指出了潜在的下行风险：包括国际市场对医疗软件付费意愿显著低于美国、美国《21世纪治愈法案》对信息封锁的监管打击，以及面临 Sherman 反垄断法诉讼的不确定性。《Acquired》总结认为，Epic 证明了在容错率极低的复杂产业中，完全依靠自研、拒绝外部资本干扰的垂直深度战略能够建立起极具统治力的商业壁垒。但其相对封闭的生态系统所招致的数据壁垒指责、愈加严格的反垄断审查，以及在创始人交棒后如何维持独特的组织文化，仍是这家掌控美国医疗数据底座的超级企业面临的长期考验。

Detailed notes — zh

## Background 该内容由知名商业播客频道 Acquired 于 2025 年 4 月 21 日发布在 YouTube 平台上。页面简介指出，本期节目深入探讨了支撑美国绝大多数医疗体系运行的软件巨头 Epic Systems，揭示其创始人 Judy Faulkner 如何在威斯康星州乡村农场中打造出一座拥有奇幻园区、法律上永不上市也永不被出售的医疗中枢帝国。 ## TL;DR 本期节目详细复盘了 Epic Systems 47 年的商业奇迹与美国医疗信息化的演进史。创始人 Judy Faulkner 凭借数学与编程天赋，早年顿悟出 Chronicles 单一数据库架构，坚持零外部风投、零收购、零传统销售团队与 100% 原生自研的极简商业法则。在拿下 Kaiser Permanente 世纪大单后，Epic 凭借无懈可击的技术交付与一体化架构脱颖而出；随后在 2008 年金融危机后的《高科技法案》（HITECH Act）数百亿美元补贴浪潮中，凭借极高的可靠性横扫全美顶级医院，彻底奠定了其在全美医疗健康系统与患者门户 MyChart 上的绝对垄断地位。 ## Atmosphere & style 整期节目散发着深度商业史诗与硬核技术考古交织的迷人魅力，两位主播 Ben 与 David 延续了标志性的抽丝剥茧式叙事，对话间充满对商业逻辑顿悟的赞叹与推演快感。叙事节奏在 Judy Faulkner 带有嬉皮反叛色彩的硬核工程师治国理念、奇幻迪士尼风格的企业园区，以及极其沉重严酷的美国医保体制与繁复监管之间自如切换。听众能直观感受到一家反直觉的软件工厂如何对抗华尔街并购风潮，并在吞吐全美 18% GDP 的复杂医疗泥潭中构筑起坚不可摧的技术堡垒，整体氛围既严谨宏大又极具戏剧张力。 ## Bulletin - [02:05](https://www.youtube.com/watch?v=FNRU9ZNqcls&t=125s) Epic Systems 坚守反常规商业模式：47 年历程中零外部风险投资、零对外并购、基本无专职销售团队且拒绝打折，年营收逼近 60 亿美元，创始人通过严密信托法律架构锁死未来上市或出售的可能。 - [01:09:50](https://www.youtube.com/watch?v=FNRU9ZNqcls&t=4190s) 2003 年 Epic 凭借统一自研架构拿下凯撒医疗（Kaiser Permanente）世纪大单，推动年营收自 5,000 万美元跨越至 1.62 亿美元，被誉为奠定医疗软件工业标杆的“福特 T 型车”。 - [02:14:28](https://www.youtube.com/watch?v=FNRU9ZNqcls&t=8068s) 2009 年美国《高科技法案》（HITECH Act）注入 360 亿美元补贴推动电子病历普及，催生医疗机构争相采购顶级系统的“爱马仕铂金包效应”，在将普及率推升至 95% 的同时客观上锁死了赛道竞争壁垒。 - [02:42:32](https://www.youtube.com/watch?v=FNRU9ZNqcls&t=9752s) 相较于 Cerner、Allscripts 等主要竞品因频繁资本并购导致多系统代码撕裂、深陷百亿美元级军政合同交付泥潭，Epic 坚持单一代码库与内生现金流驱动，实现了绝对的行业统治。 ## Key takeaways - [00:37](https://www.youtube.com/watch?v=FNRU9ZNqcls&t=37s) Epic Systems 掌控着全美医疗的核心神经中枢，在不设传统销售团队、不做任何广告且坚决不议价的前提下，服务了全美绝大多数主流医院，客户流失率近乎为零。 - [02:05](https://www.youtube.com/watch?v=FNRU9ZNqcls&t=125s) 创始人 Judy Faulkner 设立了不可撤销的信托与投票权结构，从法律层面确保公司永久私有化，恪守“不上市、不并购、不被收购”的核心戒律。 - [26:06](https://www.youtube.com/watch?v=FNRU9ZNqcls&t=1566s) 美国电子病历（EMR/EHR）的核心底层驱动力本质上是应对第三方付费体系的“电子医疗计费系统”，诊疗行为必须全面法典化、标准化与可审计，方可获得保险与政府报销。 - [32:58](https://www.youtube.com/watch?v=FNRU9ZNqcls&t=1978s) Judy Faulkner 独立自研的单一集成数据库 Chronicles，打破了传统医院各科室间的数据孤岛，成为后续临床 EMR、Resolute 财务计费与数据分析平台的一体化底层基石。 - [39:58](https://www.youtube.com/watch?v=FNRU9ZNqcls&t=2398s) Epic 仅以 14 万美元初始资本起步且此后再无外部融资，坚持以工程师为主导、以应届生理工科年轻人才为核心梯队的“自研自驱”组织文化。 - [50:04](https://www.youtube.com/watch?v=FNRU9ZNqcls&t=3004s) 1987 年推出的 Resolute 计费模块打通了临床诊疗与财务结算闭环，帮助医疗机构告别拼装怪兽般的杂乱系统，确立了极高的系统可靠性壁垒。 - [01:01:14](https://www.youtube.com/watch?v=FNRU9ZNqcls&t=3674s) 2000 年推出的集成式患者门户 MyChart 彻底颠覆了患者病历流转方式，现已沉淀超 1.91 亿去重高活跃用户，构筑了患者端强大的网络效应与品牌认知。 - [01:19:37](https://www.youtube.com/watch?v=FNRU9ZNqcls&t=4777s) 在 2003 年凯撒医疗竞标中，Judy Faulkner 严拒拆分住院和门诊系统，也不妥协出让股权，以全套系统的高并发算力模型赢下这一决定企业命运的标杆大单。 - [01:29:59](https://www.youtube.com/watch?v=FNRU9ZNqcls&t=5399s) 早年与飞利浦贴牌合作破裂带来的“伤疤组织”，促使 Epic 坚定执行 100% 自研、直销且拒绝外部贴牌及多供应商集成的独立路线。 - [01:39:58](https://www.youtube.com/watch?v=FNRU9ZNqcls&t=5998s) 位于威斯康星维罗纳的千亩奇幻园区与 11,400 座的 Deep Space 地下大礼堂，承载着极致规范的“软件工厂”哲学与全员临床轮岗、零 Bug 容忍的严苛工程师纪律。 - [01:52:17](https://www.youtube.com/watch?v=FNRU9ZNqcls&t=6737s) Epic 极度强势地推行标准化交付方案（Standard Package），反对客户魔改系统，以“定制越多费用呈指数级上升”的定价杠杆倒逼全行业向其统一规范看齐。 - [02:08:49](https://www.youtube.com/watch?v=FNRU9ZNqcls&t=7729s) 通过强制标配 Care Everywhere 模块，Epic 促成每日超 2,000 万份病历在全美跨院流转，凭借同构互联优势在事实上解决了医疗数据孤岛难题。 - [02:14:28](https://www.youtube.com/watch?v=FNRU9ZNqcls&t=8068s) 2009 年奥巴马《高科技法案》（HITECH Act）设立 360 亿美元资金池加速电子病历落地，在政府兜底资金下，医院纷纷弃用低端竞品采购最昂贵但最稳妥的 Epic，形成“爱马仕铂金包效应”。 - [02:30:49](https://www.youtube.com/watch?v=FNRU9ZNqcls&t=9049s) 法案与系统的刚性合规要求反向绑架了临床工作流，导致过度检查、向上编码以及医生“面诊 1 小时录入 2 小时”的系统性职业倦怠。 - [02:42:32](https://www.youtube.com/watch?v=FNRU9ZNqcls&t=9752s) 当对手 Cerner 因收购整合西门子医疗 IT 导致代码割裂并在军政百亿美元天价采购中深陷超支延期泥潭时，错失该类订单的 Epic 因祸得福，专注于民用顶级学术医院，保持了单一代码库的高效演进。 ## Detailed walkthrough ### [00:00](https://www.youtube.com/watch?v=FNRU9ZNqcls&t=0s) 医疗调研的终点与节目的序幕节目在一阵轻松的打趣中开场。Ben Gilbert 提到，前一天晚上 David Rosenthal 给他发短信，说自己调研美国医疗体系的资料已经一路查到了“新加坡”。在商业播客界，研究美国医疗系统有一个心照不宣的行业黑话：一旦你的研究触角不得不延伸到对比新加坡的医疗体系时，就意味着你该彻底打住调研、赶紧开录这期节目了，因为美国医疗系统的错综复杂足以把任何人吸入无底深渊。随着标志性的片头音乐与“Who got the truth?”旋律响起，两位主持人正式开启了 2025 年春季季度的全新篇章。他们将目光投向了一家坐落在威斯康星州乡村、看似低调神秘，却在每一个美国人的生命中占据举足轻重地位的庞然大物——Epic Systems。 ### [00:37](https://www.youtube.com/watch?v=FNRU9ZNqcls&t=37s) 隐秘而庞大的医疗软件帝国：Epic Systems 的反常识底色对于绝大多数美国普通民众而言，Epic 这个名字最直接的触点莫过于其患者门户应用“MyChart”。只要你在美国看过病，你的手机里大概率就装有这个软件，用它来查看化验单、预约医生或向诊所发送留言。然而，这只是冰山一角。在商业世界中，Epic 是一家在几乎所有维度上都极度反传统、甚至反常识的公司： - **零传统营销与销售**：公司不仅不做任何广告营销，甚至基本没有专职的销售团队；面对主动找上门希望购买软件的潜在客户，他们经常直接回绝。 - **绝不讲价与打折**：商业谈判中绝不接受讨价还价，无论客户规模多大，价格一律不予打折。 - **零风险投资与零并购**：在长达 47 年的创业历程中，Epic 从未拿过一分钱的风险投资（VC），也从未收购过任何一家外部公司，所有系统均由内部团队原生自研。 - **严禁远程办公**：公司要求全体员工必须全天候线下到岗。其位于威斯康星州维罗纳（Verona）农田间的巨型园区堪比奇幻乐园，里面错落着奥兹王国（Land of Oz）、哈利波特风格的魔法学院、树屋、复刻版纽约中央火车站，甚至还建有一座深藏于地下的万人级（11,000 座）超大型礼堂。 - **近乎 100% 的不可思议留存率**：全美绝大多数大型主流医院系统都在运行 Epic 的软件。在其 47 年的历史中，服务过的 600 多家顶级医疗机构客户几乎从未流失——严谨地说，历史上仅有唯一一家客户曾离开过 6 个月，随后便乖乖重回 Epic 的怀抱。 ```mermaid flowchart TD accTitle: Epic Systems 核心反常规运营法则 accDescr: 展示 Epic Systems 在商业、研发、文化及客户留存层面的反传统特征 subgraph Business["反常规商业法则"] B1["零外部风险投资 (VC)"] B2["无营销与专职销售团队"] B3["拒绝价格谈判与折扣"] B4["47年零外部并购，全部自研"] end subgraph Operations["极致的线下与园区文化"] O1["严禁远程，全员全天候在岗"] O2["维罗纳农场梦幻园区"] O3["地下万人级超大型礼堂"] end subgraph Market["统治级市场护城河"] M1["覆盖全美绝大多数主流医院"] M2[">600家大型客户，47年仅1家离开6个月后回归"] end Business --> Market Operations --> Market ``` ### [02:05](https://www.youtube.com/watch?v=FNRU9ZNqcls&t=125s) 传奇缔造者 Judy Faulkner：拒绝上市、被收购与永不妥协的治理结构这家非凡企业的掌舵人正是 Judith Faulkner（业内人称 Judy）。在两位主持人看来，她毫无疑问是当代最杰出的商业领袖之一。外界之所以对她知之甚少，是因为 Epic 至今仍是一处极为封闭的私人王国：Judy 与其家族基金会牢牢掌控着公司大约一半的股权。如今的 Epic 年营收已逼近 60 亿美元，拥有超过 14,000 名员工，但它依然公开宣布：**绝不上市，也绝不寻求被收购**。为了将这一理念化为永恒，现年 81 岁的 Judy 设立了极其周密的继任计划与信托投票权结构，从法律层面上锁死了 Epic 未来被出售或上市的一切可能。多年来，微软（Microsoft）、通用电气（GE）、谷歌（Google）等科技与工业巨头无时无刻不在觊觎 Epic，四处打探收购的可能性，但所有企图最终都撞上了不可动摇的铁壁。Ben 甚至直接给出了前置结论：创办了医疗健康领域最具价值公司之一的 Judy Faulkner，很可能是整个人类商业史上白手起家最成功的女性企业家。两位主持人也点明了理解 Epic 的必要性：今天的美国医疗系统深陷泥潭，失控的医疗费用与庞大冗余的行政管理支出已使其成本攀升至美国 GDP 的 18% 左右，体制内的繁文缛节与浪费令人窒息。要解构整个美国医疗体系的运行逻辑与权力格局，Epic 是一个绝对绕不开的神经中枢与核心切口。 ### [03:46](https://www.youtube.com/watch?v=FNRU9ZNqcls&t=226s) 播客公告与赞助商致谢在正式进入历史正文前，主持人介绍了节目的常规互动与近期动态： - 听众可订阅官方邮件列表（acquired.fm/email），获取下期节目的独家线索与过往勘误。 - 欢迎医疗与科技领域的从业者加入官方 Slack 社群（acquired.fm/slack）参与深度讨论。 - 主持人宣布了一项重磅活动预告：在听众的长期呼吁下，Acquired 将于 7 月 15 日联合 JP Morgan Payments 在纽约市举办线下专属活动（acquired.fm/nyc）。同时，节目特别致谢了长期合作伙伴 JP Morgan Payments，强调每家伟大企业的背后都离不开底层支付体系的支持；并附带了常规免责声明：本播客不构成任何投资建议，主持人可能持有讨论公司的相关资产，但均未持有非上市企业 Epic 的任何股份。 ### [05:25](https://www.youtube.com/watch?v=FNRU9ZNqcls&t=325s) 诞生于新泽西的数学女孩：跨越时代的创业与科技基因时光倒流至 1943 年 8 月，Judy Faulkner（原名 Judy Greenfield）出生于美国新泽西州的厄尔顿（Earlton，现属于彻里希尔 Cherry Hill）。这里是费城隔着特拉华河对望的典型郊区，不仅与两位主持人的故乡近在咫尺，也是后来流行巨星 Taylor Swift 的成长之地，更因靠近杜邦（DuPont）等工业传奇而孕育着浓厚的美国东岸创新气息。更为巧合的是时代的技术浪潮：就在 Judy 出生大约 4 年后，距离她家不远的新泽西默里山（Murray Hill）贝尔实验室（Bell Labs）里，威廉·肖克利（William Shockley）及其同事发明了晶体管，从而彻底拉开了现代半导体、计算机、微软乃至后来医疗数字化的序幕。在早期，涉足电子病历（EMR）的先驱往往都是洛克希德（Lockheed）、通用电气（GE）、西门子（Siemens）这类老牌硬件工业巨头。然而，童年时期的 Judy 与前沿硬核科技并无直接交集。她的商业基因最初来自于父亲 Lou Greenfield。Lou 是一位典型的小镇实体创业者，在当地经营着一家集社区药房与苏打水冷饮柜台于一体的街坊小店“Lou's Soda Fountain”。父亲经营药房与小商铺的日常，潜移默化地在幼年 Judy 心中埋下了商业与医药零售的种子。 ### [07:00](https://www.youtube.com/watch?v=FNRU9ZNqcls&t=420s) 母亲的和平奖传奇与自学 Fortran 的数学天才如果说父亲教会了她商业常识，那么 Judy 的母亲 Dell Greenfield 则赋予了她非凡的魄力与超前的思辨精神。Dell 是一位十足的行动派先锋女性，15 岁便高中毕业，早年做过秘书，后来协助丈夫打理药房和冷饮店。在经历二战洗礼后，面对越战的爆发，她极具反思精神地投身和平反战运动，担任了南泽西和平中心（South Jersey Peace Center）的负责人。等孩子们长大离家后，父母搬迁至俄勒冈州波特兰市，Dell 出任了“俄勒冈社会责任医师协会”（Oregon Physicians for Social Responsibility）的执行主任。令人惊叹的是，该组织与国际组织“国际防止核战争医生组织”（IPPNW）紧密合作，因在冷战时期倡导消除核战争威胁的杰出贡献，在 **1985 年荣获了诺贝尔和平奖**。换言之，Judy Faulkner 的母亲正是当年分享诺贝尔和平奖荣誉的团队核心成员之一。这是 David 翻阅大量早期讣告才意外挖掘出、甚至连绝大部分科技与商业媒体都未曾报道过的非凡家庭秘辛，并在随后得到了 Epic 官方的正式确认。拥有商人父亲的务实、独立母亲的抗争风骨与反叛思维，年轻时的 Judy 展现出的最大热情却纯粹归于理性的数学： - **初中顿悟数论之美**：念七年级时，数学老师在课堂上抛出了一个数论问题：“为什么任意一个能被 3 整除的数字，把它的各个数位上的数字相加之后，所得的和依然能被 3 整除？”这个问题瞬间击中了 Judy 的心智，让她立志将数学作为终生追求。 - **自学 Fortran 的天才程序员**：1961 年高中毕业后，Judy 进入迪金森学院（Dickinson College）主修数学。在大学期间的一个暑假，她前往罗切斯特大学（University of Rochester）的粒子物理实验室寻找暑期兼职。为了协助物理学家运行实验数据，实验室要求她学习计算机编程。面对这门全新的语言，实验室随手递给她一本 Fortran 编程手册，Judy 凭借过人的数学天赋，在一周之内完全通过自学啃下了 Fortran，迅速成为了团队中最高效顶尖的写码能手。一代医疗软件巨匠的计算机传奇，就此拉开序幕。 ### [09:56](https://www.youtube.com/watch?v=FNRU9ZNqcls&t=596s) 编程先驱与艺术天赋：朱迪·福克纳的早期计算机启蒙朱迪·福克纳（Judy Faulkner）在计算机实验室里的表现证明了她是个不折不扣的天才。在那个软件工程甚至还称不上一个学科的年代，编程尚未成为主流行业，大学里普遍还没设立计算机科学系，世上极少有人能够驾驭程序语言，仅有少数具有深厚数学背景的人在尝试用编程语言进行探索。朱迪恰恰深度参与并见证了计算机科学这门学科诞生的全过程。在罗切斯特的那个夏天，朱迪彻底爱上了编程。她后来回忆称，写代码的感觉就像“一个小孩子在玩黏土”（a kid playing with clay）。对她而言，计算机编程是一种不可思议的融合——它既有她热爱的严谨数学，同时又兼具语言与艺术的特质。尽管她在数学上极具天赋，但骨子里也拥有一种极富创造力的艺术特质。节目主播指出，虽然朱迪走过的道路和个性截然不同，但回顾她的成长环境、思维方式以及作为企业家的行事风格，人们会清晰地感受到一种“比尔·盖茨（Bill Gates）的气质”；而另一位主播补充道，朱迪在和平主义运动、甚至略带嬉皮色彩的环境中长大，某种程度上也带有类似史蒂夫·乔布斯（Steve Jobs）的影子。更重要的是，朱迪与盖茨是同代人，她创办 Epic 的时代背景与微软和苹果的创立几乎处于完全相同的历史时间节点，并且在后来的商业运作中展现出了深刻的“微软式 DNA”。 ### [11:37](https://www.youtube.com/watch?v=FNRU9ZNqcls&t=697s) 命运的转折点：斯坦福还是威斯康星？朱迪从迪金森学院（Dickinson College）毕业前夕，决定申请数学专业的研究生。她一共申请了五所名校的博士项目，凭借卓越的成绩全部被录取。在简历中，她特别附上了在罗切斯特的研究经历以及编写 Fortran 语言的编程经验。恰在此时，她申请的其中两所大学——斯坦福大学（Stanford University）与威斯康星大学麦迪逊分校（University of Wisconsin–Madison）刚刚着手筹备各自的计算机科学系。两所学校在审阅材料时，敏锐地意识到朱迪不可多得的计算机与数学跨界才华，正是新成立的计算机系梦寐以求的人才。于是，院系单方面做出了一个重大决定：直接将她的申请从数学系转到了新建的计算机科学系。朱迪得知后惊喜地发现自己居然可以攻读计算机科学的博士学位，并欣然接受了这个全新方向。最终，她选择了威斯康星大学麦迪逊分校开启博士生涯。这成了她人生中极具戏剧性的“滑动门时刻”（sliding door moment）： - **如果选择斯坦福大学**：她置身于硅谷初期的风险投资生态与科技公司氛围中，极可能会被截然不同的创业文化与技术理念所熏陶。 - **选择威斯康星大学**：她扎根于麦迪逊，此后虽未最终完成博士学位，却彻底在此生根发芽，并在这里创立了日后影响整个美国医疗体系的巨头。 ```mermaid flowchart TD accTitle: Judy Faulkner's University Decision accDescr: Diagram illustrating Judy Faulkner's grad school application pivot and regional divergence. A["本科就读于迪金森学院 （精通数学与 Fortran 编程）"] --> B["申请 5 所数学博士项目全部录取"] B --> C{"斯坦福与威斯康星大学 刚设立计算机科学系"} C -->|两校单方面转入 CS 录取| D["获得计算机科学博士资格"] D --> E["选择威斯康星大学麦迪逊分校"] D -.->|未选择的平行世界| F["斯坦福大学 / 硅谷创投模式"] E --> G["深耕麦迪逊：开启医疗软件革命（Epic）"] ``` ### [12:56](https://www.youtube.com/watch?v=FNRU9ZNqcls&t=776s) 沃纳·斯拉克博士的划时代课程与“排班系统” 在威斯康星大学期间，朱迪选修了一门名为“医学中的计算机”（Computers in Medicine）的先锋课程。授课教师是威斯康星大学医学院的沃纳·斯拉克博士（Dr. Warner Slack）。这门课几乎是全世界范围内最早探索计算机科学与医学实践相结合的先驱课程之一。当时大型机（mainframe）在 1960 年代才刚刚崭露头角，计算机此前几乎只用于政府和国防领域，这门课开创了医疗信息化的先河。毫不意外，朱迪成了这门课上最耀眼的学生。课程结束后，斯拉克博士主动邀请她为威斯康星大学医学院及附属医院开发一套实用软件：优化医生的值班与随叫随到排班表（on-call schedules）。这项兼职工作按小时计费，医院为她开出每小时 5 美元的时薪。有趣的是，朱迪的代码写得太好、开发效率过高，以至于她根本没能赚到多少钱。随后，全院各个科室纷纷找上门来请她开发各类软件。即便院方后来将她的薪资翻倍至每小时 10 美元，由于她总能以惊人的速度搞定需求，她依然“没拿到多少兼职收入”。 ### [14:42](https://www.youtube.com/watch?v=FNRU9ZNqcls&t=882s) 医院部门的痛点：Epic 构想的萌生随着朱迪频繁穿梭于威斯康星大学医学中心的不同部门——从精神科、妇产科、康复科，再到住院部和重症监护室（ICU）——医生和医疗从业者们向她倾诉了一个日益尖锐的共同痛点： > “我们正在接诊的这批患者，医院里的其他科室也在接诊。但是当患者穿梭于不同科室之间时，各科室之间根本无法知道在其他科室都发生了什么。我们极度渴望能有一个单一的数据库，能够在患者于威斯康星大学医学中心接受诊疗的整个纵向病程（longitudinal course of care）中，完整保存每个人的医疗记录。” 这一来自一线临床科室的迫切呼声，不仅是现代电子健康档案（Electronic Health Record, EHR）的逻辑雏形，也是 Epic 系统的真正原点。 ### [15:38](https://www.youtube.com/watch?v=FNRU9ZNqcls&t=938s) 纸质病历与百年标准化困局：从麻省总医院到洛克菲勒基金会在 1960 年代计算机介入之前，全美医疗系统的患者健康记录完全处于一种拼凑、零碎且非正规的纯纸质状态。关于医疗记录标准化的努力，最早可追溯至 1912 年： - **1912 年在波士顿发起**：美国医学会（AMA）与美国医院协会（AHA）的数名成员齐聚哈佛医学院最大的教学医院——麻省总医院（Massachusetts General Hospital, MGH），尝试为全国医生制定一套通用的病历记录规范，当时这类记录被称为“患者日记”（patient diaries）。 - **1919 年推出标准日记**：耗时整整七年，他们于 1919 年推出了全美首个标准化的“治疗日记”（treatment diary）。然而，在现实中根本没有任何机制促使医生使用它。医生群体向来独立心极强、极具自尊，且受过长期高难度精英教育，极度反感外人教他们如何写病历；在当时的医生看来，医学是一半科学、一半艺术，“我是医生我最懂，由我来决定如何提供最佳治疗”。 - **1928 年档案员协会成立与洛克菲勒资助**：1928 年，北美健康记录档案员协会（Association of Health Record Librarians of North America）正式成立，作为官方机构来规范患者数据的采集、存储与检索。这一机构的资金背后正是洛克菲勒基金会（Rockefeller Foundation）——正如在标准石油历史中所呈现的那样，洛克菲勒基金会是全美现代正规医学院建立的最重要资助者。然而，在 1960 年代计算机普及之前，所有的标准化愿景都受限于“纸质载体”的物理天花板： - **物理转移极其低效**：无论格式设计得多么规范，要把成捆成箱的厚重纸质病历从医院的一个病区送到另一个病区都极费周章，若要在不同医院、不同卫生系统甚至不同州之间调阅，无异于一场灾难。 - **数据缺乏结构化**：纸质媒介天然无法进行严格的数据结构化，医生可以在上面任意书写，字迹与表达随心所欲，无法形成可供统计、检索与横向对比的有效数据网络。这一切最终为朱迪·福克纳与计算机信息化革命的历史登场埋下了铺垫。 ### [19:41](https://www.youtube.com/watch?v=FNRU9ZNqcls&t=1181s) 计算机时代的来临与医疗政策的转折点 20 世纪 60 年代中期，医疗记录领域迎来了两个同等重要的历史性转折。首先是计算机时代的降临：数据终于可以实现数字化，医疗信息的便携性、标准化以及跨机构共享的终极梦想初露曙光。然而，另一个或许更具决定性意义的事件发生在 1965 年——美国国会正式通过法案，设立了联邦医疗保险（Medicare）与由联邦和各州合办的医疗补助（Medicaid）。在理解现代医疗软件商业模式之前，必须先理解美国的政策架构，因为该领域所有的商业结构与技术演进，归根结底都由国家政策所塑造。为了理清 1965 年这项政策从何而来，需要将时钟拨回到 1942 年的第二次世界大战时期。 ### [20:45](https://www.youtube.com/watch?v=FNRU9ZNqcls&t=1245s) 1942年《稳定法案》：雇主赞助医保体系的意外诞生在 1942 年之前，美国仅有约 10% 的人口拥有健康保险。当时的医疗模式极度简单直接——即按次付费（Fee-for-service）。病人看医生，直接从口袋掏钱付诊金；当时既不存在动辄数百万美元的失控开销，也缺乏昂贵复杂的高精尖突破性疗法。然而，二战期间美国政府颁布了《1942年稳定法案》（Stabilization Act of 1942），对国内工资与物价施加了严格统制。面对工资封顶管制，急于在劳动力短缺环境下争夺顶尖人才的企业找到了合法的“制度漏洞”：既然不能直接加薪，就为员工提供健康保险作为非工资福利。工会迅速嗅到契机并强烈支持这一做法，呼吁“让资方为工人的健康计划买单”。随后，国家战时劳工委员会（National War Labor Board）裁定雇主提供医保完全合法。紧接着，雇主群体与刚刚兴起的商业医保公司联手向国会游说，促成了极其关键的税收优惠政策： 1. 雇主为员工缴纳的医保保费可作为企业经营成本进行税前全额扣除； 2. 员工获得的医保福利属于税前福利（Pre-tax benefit），免征个人所得税。 ```mermaid flowchart TD accTitle: 雇主赞助医保机制形成流程 accDescr: 展示战时工资管制如何催生出美国雇主赞助的免税商业医疗保险体系 subgraph Policy ["政策规制"] A["1942年《稳定法案》 工资与物价管制"] --> B["企业无法涨薪招人 转而提供健康保险福利"] B --> C["战时劳工委员会裁定合法 工会全力推动支持"] end subgraph TaxRules ["税收特权确立"] C --> D["雇主端保费支出 企业税前扣除"] C --> E["员工端医保福利 免征个人所得税"] end subgraph MarketImpact ["市场结构固化"] D --> F["雇主医保覆盖率飙升 1942年: 10% → 1964年: 80%"] E --> F F --> G["雇主与大型商保公司 成为医疗服务默认支付方"] end ``` 这种制度激励彻底改写了美国医疗体系的底色。保费变成了“税前列支”，不仅降低了企业负担，也让员工乐享其成。到 1946 年战后，美国医保覆盖率从六年前的 10% 迅速蹿升至 30%；到了 1964 年，这一数字已高达 80%。由雇主主导、通过大型私营保险公司提供医保的框架，自此成为美国无法动摇的默认路径。 ### [23:40](https://www.youtube.com/watch?v=FNRU9ZNqcls&t=1420s) 1965年 Medicare/Medicaid 与成本感知的彻底脱钩雇主医保模式席卷全国后，抛下了一个严重的社会空缺：那些不在劳动力市场中的人群——老人与穷人该怎么办？老龄人口与贫困群体恰恰是整个社会中医疗需求最大、消耗医疗资源最多的群体。战后英国在强烈的全民团结凝聚力下，迅速建立了国家医疗服务体系（NHS）实行单一支付方制度（Single-payer system）；虽然罗斯福、杜鲁门乃至肯尼迪总统都曾试图推行全民公费医疗，但美国始终缺乏足够的政治共识。最终的妥协产物就是 1965 年约翰逊政府通过的《社会保障法修正案》（Social Security Amendments of 1965），正式诞生了两大支柱： - **Medicare（联邦医疗保险）**：全额由联邦政府出资支持，专门覆盖 65 岁及以上的老年群体； - **Medicaid（医疗补助计划）**：针对低收入等特殊困难人群，由联邦政府提供补助、各州政府具体分工负责与运营。这一系列政策组合，构成了当今美国医疗的核心骨架：有工作的人依赖雇主商业医保，失业个体借助《平价医疗法案》（ACA）在保险交易所自购商保，老人归 Medicare，低收入人群归 Medicaid。这种体系带来了一个致命的心理认知扭曲——**绝大多数美国患者几乎完全感知不到医疗服务的真实成本**。由于账单被层层打包转嫁给第三方支付方（私营保司、Medicare 或 Medicaid），患者往往认为看病花的是“系统”的钱，而非自己口袋里的钱。这种深度抽象化直接导致了医疗费用的全面失控：美国的医疗卫生支出一路膨胀至 GDP 的 18%，远超英国的 11%，更是新加坡（约 6%）的三倍之多。 ### [26:06](https://www.youtube.com/watch?v=FNRU9ZNqcls&t=1566s) 第三方支付与浮士德交易：电子病历与计费系统的刚需支付链条的制度化重构，瞬间将初生的电子病历推上了整个产业的历史舞台中心。在此之前，医生只需看诊并向病患当面收费；而当第三方支付方主导一切后，医院和医生必须在看诊后向保险公司或政府申请报销。无论是官僚化的政府机构还是追求利润的商业保险公司，绝不可能凭空付款，它们要求铁证如山的医疗记录，而且必须是严格结构化、标准化的凭证。 ```mermaid flowchart LR accTitle: 医疗服务向第三方报销流程 accDescr: 展现医生诊疗如何经过标准化记录转化为账单并向支付巨头报销 subgraph Clinic ["临床现场"] Doc["医生/医疗团队"] -->|"开展复杂诊疗"| Pt["患者 (接受护理)"] end subgraph Records ["数据与工作流系统"] Doc -->|"详细记录规范操作"| EMR["标准化电子病历系统 (EMR)"] EMR -->|"编码与审核"| Billing["电子医疗计费系统"] end subgraph Payers ["第三方支付巨头 (高议价权)"] Billing -->|"发送标准审计账单"| Gov["Medicare / Medicaid (老年/低收入)"] Billing -->|"发送标准审计账单"| Ins["商业保险公司 (商业医保)"] end ``` 医学界由此签下了一场“浮士德式的契约”： - 医生和医院如果想要拿到报销，并且随着手术复杂度的提升赚取更高的收益，就必须对每一项医疗行为进行高度法典化、受监管且标准化的记录。 - 诊室里发生的一切与发给支付方的账单之间，必须建立滴水不漏的数据流与工作流。 - 因此，行业对**电子病历系统（EMR）**产生了生死攸关的刚需；而更底层、更核心的驱动力，其实是**电子医疗计费系统（Electronic Medical Billing System）**。由于保险的本质是资金池聚合，Medicare、Medicaid 以及大型商业保司掌握着对独立诊所与医院系统的绝对议价权。面对掌控资金生杀大权的强势对手，医疗机构必须将自己的病历系统做得极度严谨、极其可审计、极其标准化，才能在审查中顺利获得回款。 ### [28:52](https://www.youtube.com/watch?v=FNRU9ZNqcls&t=1732s) 早期探索：麻省总医院与 COSTAR 系统的技术挑战 1966 年，也就是 Medicare 和 Medicaid 设立后的第二年，麻省总医院（Massachusetts General Hospital）的研究团队在美国国立卫生研究院（NIH）的资助下，启动了人类历史上首个真正意义上的计算机化医疗记录系统——**COSTAR**（Computer-Stored Ambulatory Record）。 COSTAR 不仅涵盖了患者预约挂号、就诊登记和临床数据记录等全套现代电子健康档案功能，而且最为关键的是直接承担了与第三方支付方对接的计费模块。然而，在 20 世纪 60 年代的技术条件下，开发这种系统的技术难度极其严苛： - **高并发事务挑战**：整个医院网络内的众多医护人员与行政管理人员需要同时在线读写，系统对高并发事务处理（High-concurrency transaction processing）有着极高的要求。 - **硬件资源极度受限**：当时的计算环境受制于昂贵的主机（Mainframe），存储空间与算力极其有限，而系统还需要连接并协调分布在各个终端的大量设备。 - **编程工具匮乏**：当时像 FORTRAN 这类主流科学计算语言，完全无法胜任此类复杂的非结构化临床数据、关系网络以及并发状态的管理，促使早期的开发者必须从底层重构编程语言与数据库架构。 ### [30:09](https://www.youtube.com/watch?v=FNRU9ZNqcls&t=1809s) 医疗专属编程语言 MUMPS 的诞生与 MEDITECH 的崛起在马萨诸塞州总医院（MGH）推进早期医疗信息系统项目的工程师们很快发现，现有的编程语言根本无法满足构建这类软件的需求。于是，他们着手开发了一门专门针对医疗场景的新编程语言——MUMPS（全称为 Massachusetts General Hospital Utility Multi-Programming System）。这门名字恰好与“腮腺炎”同名的语言，如今依然深刻影响着整个医疗软件行业：Epic 至今依然在使用 MUMPS 及其现代演进版本 Caché。虽然许多竞争对手早已不再使用 MUMPS，但凭借 Epic 如今不可撼动的市场主导地位，它几乎成为了事实上的行业标准。 ```mermaid flowchart TD accTitle: MUMPS 核心架构与 MEDITECH 起源 accDescr: 展示 MUMPS 语言的两大核心架构特性，以及 Neil Pappalardo 创立 MEDITECH 的过程。 subgraph MUMPS_Architecture["MUMPS 语言核心特性"] A["语言层与数据库深度整合"] --> C["计算高效 / 无需中途编写 SQL"] B["天生支持多用户并发访问"] --> D["杜绝数据写入冲突与数据丢失"] end subgraph History["历史演进与商业化"] E["MGH 核心程序员 Neil Pappalardo"] -->|1968 年创立公司| F["Medical Information Technology (MIT)"] F -->|更名避开 MIT 校名冲突| G["MEDITECH"] G --> H["当前美国 EMR 市场第 3 大巨头"] end ``` MUMPS 之所以需要被重新发明，核心在于两大关键架构设计： - **编程语言与数据库深度一体化**：数据结构直接内置于语言自身中。这极大提升了高频、数据密集型交易的处理效率；对程序员而言体验也非常顺畅，不必在写业务逻辑的中途切换去编写 SQL 查询，一切都在同一套语言体系内完成。对于主要功能就是“围绕数据库做封装、不断进行读写”的系统来说，这种架构再合适不过。 - **专为多用户并发设计**：在医院场景下，不同科室可能需要同时更新同一份病历，与此同时后台行政人员也在同步维护该记录（核对哪些已向支付方报销、哪些被拒赔或申诉）。这种高并发绝不能出现写入冲突导致数据丢失，毕竟“这关乎病人的生命安全”。当时在 MGH 领导开发 COSTAR 与 MUMPS 的核心程序员是麻省理工学院（MIT）的新晋毕业生尼尔·帕帕拉多（Neil Pappalardo）。1968 年，年轻有为的尼尔选择走出象牙塔独立创业，成立了一家名为 Medical Information Technology（缩写正好是 MIT）的公司来商业化这套软件。不久后，大概是在正牌麻省理工学院的交涉下，公司更名为 MEDITECH。业内人士对 MEDITECH 绝不陌生——历经数十年风雨，它至今依然活跃在市场上，是仅次于 Epic 和 Cerner（甲骨文旗下）的全美第三大电子病历（EMR）巨头。 --- ### [32:58](https://www.youtube.com/watch?v=FNRU9ZNqcls&t=1978s) 客厅里的顿悟：Judy 创造 Chronicles 数据库内核与此同时，在威斯康星大学麦迪逊分校，朱迪·福克纳（Judy Faulkner）正与沃纳·斯莱克（Dr. Warner Slack）博士合作推进各项医疗应用软件开发。后来斯莱克博士前往哈佛大学任职，自然结识了尼尔并开始与 MEDITECH 合作，但他与朱迪始终保持着紧密联系。当时全美在医疗系统中做软件的计算机圈子很小，大家的目标高度一致：威斯康星想要的东西，与哈佛大学、MGH 梦寐以求的完全一样——一套能够贯穿整个照护连续体（Continuum of care）、形成纵向完整病历并可直接用于医保账单结算的集成系统，堪称该领域的“圣杯”（The Holy Grail）。通过这层人脉，朱迪接触并深入学习了 MUMPS 语言。20 世纪 70 年代中期的某一天，传奇时刻降临了。当时朱迪正坐在自家的客厅里，脑海中突然灵光一闪，悟出了如何打造这套单一集成数据库的系统架构。正如她后来回忆所言： > “阳光正好照进来，我当时正有些心不在焉，就那么坐着，脑子里突然豁然开朗：‘集成系统应该这样做！’我记得自己一路飞奔冲进厨房，抓起一个便签本，立刻在纸上疯狂地写代码，一行接一行。那段代码后来变成了 Chronicles。” Chronicles 意为患者照护全过程的“编年史”。尽管朱迪后来开玩笑说自己当年手写的代码早已不在现网生产环境中运行，但 Chronicles 奠定的数据库内核，时至今日依然是整个 Epic 系统的绝对根基与最核心的护城河： - **单一核心数据库驱动全系统**：Epic 旗下的数百个应用模块——无论是临床端的 EMR、负责计费的 Resolute 模块，还是大数据分析平台 Cosmos、针对特定专科的各种模块——全都直连且仅依赖 Chronicles 这一套统一的数据库。没有任何中间数据孤岛，所有应用天然实时打通。在 20 世纪 70 年代，所谓“软件产品”的形态还极其简陋。那依然是大型机（Mainframe）与 Unix 终端的时代，根本没有现代意义上的图形界面（GUI）。朱迪写好的 Chronicles 数据库部署在医院的大型机上，威斯康星大学医疗中心的不同科室只需自行针对该数据库编写查询界面，就可以在各自诊室或办公室的绿色荧光屏、80 列字符宽的纯文本终端上实时调取与录入病人数据。 --- ### [36:14](https://www.youtube.com/watch?v=FNRU9ZNqcls&t=2174s) 口碑爆炸与 Human Services Computing 的创立起初，朱迪和 MGH 团队一样，只打算把 Chronicles 做成威斯康星大学内部使用的定制工具，自己也只是一名普通的大学雇员程序员。然而，威斯康星大学的医生们在频繁奔赴全美各地参加学术会议时，开始四处向同行炫耀：“我们院里有个厉害的程序员叫朱迪·福克纳，她给我们做了一套无比强大的系统叫 Chronicles！” 这种口口相传带来了病毒式的扩散效应。威斯康星大学的电话被全国各地的医疗系统打爆，随后所有的电话都转到了朱迪的办公桌前：“嘿，你能帮我们也写一套一模一样的系统吗？”起初朱迪多次被劝说自立门户开公司，但她每次都果断拒绝，坚持认为：“不，我只是给本院做了一个特定版本，我们没必要为此专门开家公司。” 但登门求购的机构越来越多，朱迪终于招架不住举手投降：“行吧，我开公司还不行吗？我就兼职干干，拉几个身边的同事一起兼职搞，就当成一个小打小闹的副业。”1979 年，这家公司正式登记注册，名字起得极其朴素甚至略显平庸——Human Services Computing（人类服务计算公司）。而这正是日后统治全美医疗软件版图的 Epic Systems 的前身。 --- ### [37:32](https://www.youtube.com/watch?v=FNRU9ZNqcls&t=2252s) 赞助商插播：JPMorgan Payments 与医疗支付的复杂生态在进入 Human Services Computing 如何正式蝶变为 Epic 之前，节目穿插介绍了本期节目的呈现合作伙伴摩根大通支付（JPMorgan Payments）。节目讨论了现代支付基础设施如何成为支撑各行各业商业创新的命脉： - 现代产品形态如网约车、创作者经济、B2B SaaS 交易市场以及复杂的供应链管理，若没有如今高度成熟的支付技术，在十年前根本不可能成立。 - 摩根大通支付每天处理的交易体量高达 **10 万亿美元**（$10 trillion a day），拥有跨国巨头与初创企业无法超越的承载能力。 - 医疗健康领域的支付流程极其繁琐摩擦：涉及患者、医疗机构、医保网络、专科医生以及健康监测服务商等多方主体，涵盖极其错综复杂的结算账期、支付条款以及严苛的数据隐私合规要求。JPMorgan 凭借包括 InstaMed 在内的专利云端医疗支付解决方案，致力于实现医疗支付与数据流动的无缝打通。 ### [39:58](https://www.youtube.com/watch?v=FNRU9ZNqcls&t=2398s) 创立“人类服务计算”与七万美元的小型机离开上一份工作后，朱迪·福克纳（Judy Faulkner）创立了一家名为“人类服务计算”（Human Services Computing）的公司——这个名字念起来可一点都不顺口。在此之前，朱迪从未计划成为一家公司的创始人，然而在接下来的 47 年里，这成了她一生的事业。在开展任何实际业务之前，朱迪首先面临的是资金问题，最迫切的开销就是买一台能跑程序的计算机。在那个年代，买一台计算机可不像现在这么简单，他们盯上的是一台 Data General Eclipse 16 位小型机（Minicomputer），售价高达 7 万美元。这台机器体积庞大，像一套洗衣机和烘干机组合，只能塞在地下室里。当时的技术正处于大型机与微型计算机交替的尴尬过渡期：微处理器尚未普及，个人电脑（PC）时代还未真正降临，像“上网本前身”一般的小型机成了填补空白的折中产物。消费者买不起，但小企业或大公司内部的小部门却能负担得起——比起动辄天价的 IBM 大型机，7 万美元的小型机成了初创公司触手可及的生产工具。 ```mermaid flowchart TD accTitle: Epic Systems 创立初期的资金与硬件结构 accDescr: 展示朱迪·福克纳如何通过贷款与早期股权融资筹得14万美元并购买关键的小型机设备。 subgraph Capital["启动资金 (共 14 万美元)"] Debt["银行贷款 $70,000 债务"] Equity["亲友与同事投资 $70,000 股权"] end subgraph Valuation["公司估值"] Pre["投前估值: $70,000"] Post["投后估值: $140,000 (稀释 50%)"] end subgraph Asset["核心资产投入"] Computer["Data General Eclipse 16位小型机 (耗资 $70,000 / 洗衣机大小)"] end Debt --> Valuation Equity --> Valuation Valuation --> Asset ``` 为了筹集这笔巨款并启动公司，朱迪先从银行拿下了 7 万美元贷款，专门用于采购那台 Data General Eclipse。随后，她动员了身边的亲朋好友，以及在威斯康星大学系统内合作过的医生和程序员同事，凑齐了另一笔 7 万美元的股权投资。当时公司的投前估值定为 7 万美元，加上这笔新注入的 7 万美元，投后估值达到 14 万美元，朱迪为此付出了 50% 的股权稀释。 - 这 7 万美元的股权投资与 7 万美元的银行贷款，构成了 Epic 历史上仅有的外部原始融资。 - 如今 Epic 年营收超过 57 亿美元并称霸全美医疗软件市场，但总融资额就只有这 7 万美元股权和 7 万美元债务，堪比宜家（IKEA）般的内生增长奇迹。 - 早期投资人的股权后来大部分被公司回购，但仍有少数散落在外；到了 2000 年代前后，红杉资本（Sequoia Capital）通过某种途径辗转获得了一大笔早期股份，这成了 Epic 这家远离硅谷的公司在整个历史上几乎唯一一次与硅谷产生交集的秘密插曲。这之所以行得通，很大程度上是因为朱迪具备划时代的编程天才。软件开发在当时并非轻资产运作，但朱迪本身就是公司不可替代的“垄断资源”（Cornered Resource），就像早期微软拥有比尔·盖茨（Bill Gates）和保罗·艾伦（Paul Allen）一样。她一个人就能写出极其优秀的软件架构，不需要雇佣成群的程序员，因而公司在成立之初一直极其精简——坊间传闻最初仅仅是“地下室里的三个半职员工”，且这种极简规模维持了很多年。 ### [44:27](https://www.youtube.com/watch?v=FNRU9ZNqcls&t=2667s) 大学大道 2020 号的地下室与美国女孩娃娃朱迪为公司租下的第一处办公地点，位于威斯康星州麦迪逊市大学大道 2020 号（2020 University Avenue）的一栋公寓地下室。大卫在节目里试图用一道冷知识考考本：在这处地下室里，同时孕育了另一家日后家喻户晓的美国传奇本土公司，两家公司甚至曾在同一处共享空间里重叠办公。本立刻接话拆穿：“大卫，关于 Epic 的公开资料就那么多，我们俩肯定都能查到——那是‘美国女孩娃娃’（American Girl Doll Company）！”两位主持人笑谈背后的插曲：大卫本来想在调研通话里保守这个秘密，结果本也早就知道了。甚至当美国女孩娃娃搬走时，Epic 还顺手买下了他们留下的一部分二手办公家具。由普莱森特·罗兰（Pleasant Roland）在麦迪逊创立的美国女孩娃娃公司后来风靡全美，不仅让两位主持人的姐妹和女儿们为之痴迷（大卫感叹自己三岁的女儿现在每周都翻着目录杂志指着标价 250 美元的娃娃喊着要当生日礼物），最终在 90 年代末被美泰（Mattel）高价收购。两家从同一个地下室走出的企业，各自以截然不同的轨迹改写了美国的商业版图。 ### [46:08](https://www.youtube.com/watch?v=FNRU9ZNqcls&t=2768s) 命运的转折：师从 MEDITECH 与程序员治国搞定了设备和资金，朱迪面临的真正短板是商业运作：她是一名优秀的软件工程师，却完全不懂如何开办和运营一家企业。这时，导师沃纳·斯拉克（Warner Slack）为她指明了一条道路。他安排朱迪前往波士顿，去拜访 MEDITECH 的创始人尼尔·帕帕拉多（Neil Pappalardo），向这位医疗信息技术先驱学习如何创办公司。这是一个“历史在刀尖上扭转”的关键节点。斯拉克没有把朱迪介绍给传统的商界人士，而是把她推给了一位同为顶级程序员的创业者。在 1970 年代，标准的商业剧本是“引入穿西装的成年人”——让商学院 MBA 毕业生接管管理层，把创始人踢出局；朱迪之所以日后对风险投资极度警惕，正是因为亲眼目睹了那个年代拿了 VC 资金的创始人基本都被资本解雇、换上了商业经理人。而尼尔·帕帕拉多来自麻省理工学院（MIT），是 MUMPS 语言原始开发团队的核心成员，他完全是用软件开发者的逻辑来治理 MEDITECH。在波士顿的三天里，朱迪接受了一场关于公司运营的密集速成培训。尼尔把整家公司的管理流程完全模块化、标准化，就像打造了一套硬核的 API： - 一切流程皆有详尽的操作手册和文档支持； - 他把 MEDITECH 的全部制度手册毫无保留地分享给朱迪，手把手教她如何搭建人力资源系统、薪酬福利体系、大学校园招聘流程以及内部晋升通道； - 招聘策略上，MEDITECH 绝不在市场上高薪挖成熟程序员，而是直接从波士顿本地高校吸纳应届大学毕业生。朱迪将这套机制原汁原味地移植到了 Epic，最初面向威斯康星大学招聘，随后拓展至全美中西部名校。Epic 不看重医学背景，专挑理工科背景、高能动性的年轻毕业生，坚信“医疗业务知识完全可以在入职后学会”。这一做法在 1979 年堪称颠覆——当时甚至连微软都才刚刚从阿尔伯克基迁往华盛顿州，DOS 尚未面世，校园招聘和放手赋能年轻毕业生的模式在全行业还根本未成气候。时至今日，Epic 的骨子里依然保留着这种纯粹的工程师基因：全公司充斥着程序员、逻辑分析师和具有开发者思维的实施工程师，却几乎找不到传统意义上的商业管理与销售人员。 ### [50:04](https://www.youtube.com/watch?v=FNRU9ZNqcls&t=3004s) 进军高端学术医疗机构与定位分化在 20 世纪 70 年代末和 80 年代初，个人电脑（PC）行业几乎还不存在，微软当时的核心产品依然是 BASIC 解释器。对于朱迪·福克纳（Judy Faulkner）和她初创的 Human Services Computing 而言，市场环境极其受限。虽然大型学术型教学医院和拥有大学计算基础设施的医疗系统有需求、且拥有可供利用的大型机（Mainframe），但全美绝大多数普通的社区医院、诊所或门诊部根本买不起小型机或大型机。面对这种硬件条件的约束，朱迪做出了关键的市场定位决策。出于对尼尔·帕帕拉多（Neil Pappalardo）及其创办的 MEDITECH 的敬意与避让，她没有跟随 MEDITECH 去做小型医院市场，而是选择向上切入最高端、最复杂的机构：学术型培训医院，以及日后被称为综合医疗交付网络（IDN，Integrated Delivery Networks）的巨型医疗集团和儿童医院。这些机构有着最繁杂的业务流程和最深度的软件需求，同时也是唯一买得起昂贵硬件并搭建了 Unix 基础设施的客户。 ```mermaid flowchart TD accTitle: Epic Early Market Architecture accDescr: 展现 Epic 与 MEDITECH 的市场定位差异以及软硬件适配路径 subgraph Market_Segmentation["市场客群分化"] direction TB MEDITECH["MEDITECH 策略 专注中小型医院市场"] EPIC["Judy / Epic 策略 主攻顶尖学术医院与复杂 IDN"] end subgraph Infrastructure["计算基础设施支撑"] direction TB UNIX["Unix 大型机 / 小型机 （高昂成本，仅顶尖医学院与大学附属医院具备）"] DB["Chronicles 底层数据库 （自建核心架构）"] end EPIC --> UNIX UNIX --> DB ``` ### [51:31](https://www.youtube.com/watch?v=FNRU9ZNqcls&t=3091s) 从“人道服务计算”到 Epic：病患人生即是一部史诗 1979 年公司创立时仅有 4 家初始客户。到了 1983 年，朱迪决定将公司正式更名为 Epic Systems。这一命名的背后再次展现了朱迪独特的思考方式与人文气质——就像之前给核心数据库取名“Chronicles”（编年史）一样，朱迪联想到的是希腊史诗（Greek Epic）的概念，指代宏大漫长、具有纵向历史维度的传奇叙事。在她看来，病患一生的健康档案与生命历程，正是一部壮丽的“史诗”（Epic）。尽管名字气势磅礴，但当时这绝不是一家有巨大野心的扩张型巨头，而是一个发展极其缓慢的本地小型软件公司： - 1979 年：4 家初始客户。 - 1983 年（更名时）：仅有 9 家客户。 - 1988 年（成立满十年）：客户数量才达到 24 家，年营收仅 150 万美元，员工只有寥寥数人。与创立后迅速爆发的微软不同，Epic 在第一个十年里虽然稳步增长，但“绝没有震撼世界”，仅仅相当于威斯康星州麦迪逊市一家典型而滋润的本土软件小作坊。 ### [53:00](https://www.youtube.com/watch?v=FNRU9ZNqcls&t=3180s) 破局点 Resolute：账单系统与单一数据库的护城河 Epic 前十年扩张缓慢除了受制于医院缺乏计算硬件基础设施外，还有一个核心致命伤：当时的软件仅处理临床病历（Clinical Medical Records），完全没有触碰医院最痛的命根子——计费与账单系统（Billing System）。直到 1987 年，Epic 才推出了名为 Resolute 的计费模块。 Resolute 的诞生彻底扭转了公司的命运。它直接构建在 Chronicles 单一核心数据库之上，至今仍是 Epic 面向所有医院的“收入周期”（Revenue Cycle）拳头产品。这也奠定了 Epic 碾压所有竞争对手的两大杀手锏： 1. **极致的系统可靠性**：行业内其他竞争对手（除 MEDITECH 外）在漫长的资本浪潮中，几乎都通过疯狂并购（M&A）、被私募股权（PE）杠杆收购私有化或反复上市，最终沦为将 30 多家不同公司的系统硬缝在一起的“拼装怪兽”。而 Epic 几十年来始终保持自研与统一架构。医院采购 Epic 时，承诺的实施周期和预算几乎都能兑现，“上线后它就能稳定运行”，这在医疗 IT 行业堪称奇迹。 2. **临床与财务的无缝闭环**：临床记录和计费模块在同一套代码库中编写、在同一个数据库上运行，数据天衣无缝。如果两个系统相互割裂且数据传输失准，后果不仅是医院收不到钱，或者因提交虚假错误凭证而构成联邦重罪欺诈，更严重的是如果账单错误反向污染临床数据，将直接导致医疗事故甚至危及病患生命。优质医疗科技通讯《Health API Guy》对此曾精辟总结：“Epic 能够成为大型企业决策者的必然选择，恰恰是因为其高度一体化的系统架构。买方不再需要管理多个供应商和杂乱系统，而是获得了一个拥有单一数据库、统一工作流以及开箱即用互操作性（通过 Care Everywhere 模块实现）的综合平台。” ### [57:27](https://www.youtube.com/watch?v=FNRU9ZNqcls&t=3447s) PC 浪潮、EpicCare 与门诊业务的飞跃进入 20 世纪 80 年代末至 90 年代初，PC 时代正式降临。医院和医疗机构采购电脑的成本从原先的数万美元瞬间骤降至数千美元，市场需求全面井喷。 1992 年，Epic 顺势推出了颠覆性的产品——EpicCare。这是整个医疗行业公认的第一款基于 Windows 图形用户界面（GUI）的电子病历（EMR）系统： - **终结黑白终端时代**：在此之前，医生想要访问基于 Unix 的 Chronicles 数据库，只能使用繁琐简陋的命令行终端；而 EpicCare 让任何普通医护人员都能在自己的 Windows PC 上通过点击图形界面来处理病患诊疗。 - **门诊“三驾马车”齐备**：EpicCare 最初专为门诊（Ambulatory/Outpatient，即非住院不过夜场景）设计，与排程模块 Cadence、账单模块 Resolute 完美结合，为医院管理者打造了梦寐以求的运营闭环。这套杀手级组合彻底改变了 Epic 的增长曲线，使其彻底摆脱了麦迪逊本地小公司的范畴： - 1988 年，公司营收仅为 150 万美元。 - 1995 年，公司营收飙升至 1800 万美元，7 年间实现了超过 10 倍的爆发式跨越。尽管与 1975 年成立、此时早已统治 PC 软件帝国并在 1995 年迎来 Windows 95 巅峰的微软相比，同处技术早期的 Epic 步伐显得慢而扎实，但其作为医疗软件巨头的基石已然无可撼动。 ### [01:00:17](https://www.youtube.com/watch?v=FNRU9ZNqcls&t=3617s) 垂直还是横向：当医疗软件面对占美国 GDP 18% 的庞大市场节目在这里做了一个关键的对比与澄清。在 1995 年前后，微软已经上市、推出了划时代的 Windows 95，年营收突破 60 亿美元；而 Epic 此时依然是一家深耕细分行业的未上市企业。两位主持人特别提到，虽然两者的成长轨迹完全不同——一个是在构建横跨所有行业的未来通用计算平台（Horizontal Platform），另一个则是深扎医疗健康赛道的垂直软件（Vertical Software）——但必须纠正一种把垂直软件等同于“小生意”的偏见。垂直软件固然受限于单一行业的边界，但 Epic 所身处的这单一垂直领域，却吞吐着整个美国高达 18% 的国内生产总值（GDP）。这意味着，即便它永远不会成为像微软、谷歌或甲骨文那样的通用平台，医疗健康这一条赛道的惊人容量，也足以让深耕其中的胜利者触碰到超乎寻常的商业天花板。 --- ### [01:01:14](https://www.youtube.com/watch?v=FNRU9ZNqcls&t=3674s) 圣杯降临：EpicCare Inpatient 与“大一统”战略 2001 年，Epic 正式推出了住院系统 EpicCare Inpatient，专门针对医院的夜间留观与住院治疗流程。这一举措补齐了 Epic 产品版图中最核心的拼图，使整个体系达成了医疗信息化领域的“圣杯”状态：底层由 Chronicles 单一数据库统领，上层同时串联起负责门诊诊所的 EpicCare Ambulatory、负责留院救治的 EpicCare Inpatient，以及底层的综合计费系统 Resolute。住院业务在任何大型医院系统中都是绝对的收入命脉，因为最复杂、费用最高昂的救治操作全部发生在此处，正如 Medicare 之所以是医院最关键的支付方，也是因为老年群体往往伴随着最复杂沉重的医疗负担。对于医院高管和信息主管来说，所有模块长在同一个数据库、同一套代码架构之上的体验堪称梦寐以求。也正是在这个时期，Epic 全面确立了自身的核心信仰——医院客户厌恶向数十家供应商东拼西凑零散的插件方案，他们迫切渴望从单一厂商手中采购全面集成、端到端的完整解决方案。 ```mermaid flowchart TD accTitle: Epic 统一技术架构图 accDescr: 展示以 Chronicles 统一数据库为底座，整合门诊、住院及计费系统的产品架构 subgraph Core["底层统一数据引擎"] DB["Chronicles 统一数据库"] end subgraph Products["上层全业务应用套件"] AMB["EpicCare Ambulatory （门诊诊所系统）"] INP["EpicCare Inpatient （留院救治系统）"] BILL["Resolute （医疗计费结算系统）"] end DB --> AMB DB --> INP DB --> BILL ``` - **单一数据源优势**：门诊、住院与计费三大系统完全告别跨数据库接口调用，患者跨科室流动时数据零割裂。 - **战略重心转移**：放弃单点极致工具的打法，将公司未来全面押注在系统广度（Breadth）与全流程极致整合（Tight Integration）上。 --- ### [01:02:40](https://www.youtube.com/watch?v=FNRU9ZNqcls&t=3760s) 超前时代的突破：MyChart 的极客起源与患者革命早于 EpicCare Inpatient 发布的 2001 年，Epic 在 2000 年互联网泡沫顶峰时期悄然推出了面向大众的在线患者门户 MyChart。这是一个令现代科技从业者难以置信的里程碑：在 HIPAA 医疗隐私法案严格监管、公众对网络数据安全极度敏感的 2000 年，让患者能在浏览器端直接读取自己的病历，其前沿程度堪比医疗界的亚马逊，彻底颠覆了行业认知。 MyChart 的诞生源于 1997 年一个名为“EpicWeb”的原型项目，最初是为了让医生半夜在家里醒来时能远程调取病人的病历记录——这也讽刺性地预告了日后医生们不得不把大量下班时间消耗在填系统上的现状。当时一位刚大学毕业不久的年轻程序员苏米特·拉纳（Sumit Rana，现任 Epic 总裁）直截了当地走向当时的总裁卡尔·德沃夏克（Carl Dvorak），表示手头的事情做着有点无聊：“你能给我安排个更难、更前沿的项目做做吗？”这种刚入职的新人能直接找最高管理者挑战难题的工程师文化，催生了 1998 年立项、2000 年上线的 MyChart，它一跃成为全美首个全集成的患者门户系统。这种体验就像房地产领域的 Zillow 一样具有不可逆性：一旦患者能够坐在家里直接用电脑查看病历报告、无需到医院柜台排队复印或苦等传真，世界就再也回不去了。此外，MyChart 还在两个层面彻底引爆了价值： 1. **家庭照护与患者端裂变**：儿女可以远程绑定并协同管理异地年迈父母的诊疗事务，带来了极强的自然口碑传播。 2. **重塑医院运营利润率**：后续上线的在线自主挂号（Self-scheduling）与自动候补机制（Waitlist），直接拯救了医院此前因患者爽约（No-show）而造成的日常收入漏洞，将曾经损失的门诊时间切片重新自动化填满。 - **用户规模**：发展至今，MyChart 拥有超过 1.91 亿去重后的高活跃独立真实用户。 - **破除纸质传真**：终结了以往患者更换居住城市或转院时，医疗记录必须在传真机两端反复往返发送的低效噩梦。 --- ### [01:07:45](https://www.youtube.com/watch?v=FNRU9ZNqcls&t=4065s) 挺进顶级联赛：破局前夜与凯撒医疗的来电凭借着 2000 年 MyChart 与 2001 年 EpicCare Inpatient 的双剑合璧，Epic 的年营收顺利突破 5000 万美元大关。此时公司已经稳健打下了全美 88 家大型医疗系统，通过一家家医院攻坚谈判、逐步替换竞争对手的旧系统，Epic 证明了自己完全具备进入顶级联赛的资质。就在 2003 年，一通来自全美医疗界当之无愧的超级巨头——加利福尼亚州凯撒医疗（Kaiser Permanente）的电话打进了 Epic 的办公室，一场即将彻底蜕变 Epic 命运、重塑全美医疗格局的世纪战役正式拉开序幕。随后，节目切入了对长期赞助伙伴 Fundrise 的感谢与商业模式梳理： - **颠覆私募投资门槛**：Fundrise 起步于面向散户投资者的美国最大房产投资平台，在观察到科技公司选择长期保持私有化、普通投资者被挡在 Pre-IPO 财富增值门外的痛点后，于 2022 年将众筹民主化模式拓展到风险投资领域。 - **明星投资组合**：平台已让普通人直接参与到 Databricks、Canva、Anduril、Ramp、Vanta、Anthropic 以及刚于去年 12 月上市的 ServiceTitan 等顶尖科技独角兽的早期投资中。 ### [01:09:50](https://www.youtube.com/watch?v=FNRU9ZNqcls&t=4190s) 巨无霸敲门：走近凯撒医疗集团（Kaiser Permanente）在结束了赞助商插播后，本和戴维将视角转向了 Epic 发展史上具有里程碑意义的超级大单——2003 年，美国医疗界真正的“庞然大物”凯撒医疗集团（Kaiser Permanente）找上门来了。凯撒是一个总部位于加州的非凡机构，它属于高度整合的“管理式医疗联合体”（Managed Care Consortium）。简单来说，凯撒既是健康保险公司，又是实体医院系统，两者深度捆绑融为一体。投保的会员只能在凯撒自营的医疗机构内看病，凯撒完全控制了就诊、开药、报销、住院的全流程闭环。这是美国本土最接近于“单一支付方”（Single Payer）全民医保的形态，同时它还要与联邦医疗保险（Medicare）对接，为年长客户提供凯撒自家的 Medicare Advantage 计划。 2003 年的凯撒医疗规模堪称巨兽：它是当时全美规模最大的单一医院系统，旗下坐拥 30 家医院、400 多家门诊诊所、1.1 万名医生以及 850 万会员患者。就在这一年，凯撒决定针对全集团范围发布一份 RFP（需求建议书），招标采购一套全新的、全集团统一部署的电子病历（EMR）系统。正是从这一刻起，大众需要彻底颠覆对 EMR 的固有认知——它绝对不是简单的病历记录、账单录入或排班工具，而是整个医疗帝国乃至整个行业的“中枢神经系统”与底层操作系统。 ```mermaid flowchart TD accTitle: Kaiser Permanente Managed Care Model accDescr: Diagram showing Kaiser's integrated healthcare and insurance structure. subgraph Kaiser["凯撒医疗集团 (Kaiser Permanente) 闭环体系"] Ins["健康险 / 医疗计划 (同时对接 Medicare Advantage)"] Hosp["自营医疗实体 (30家医院 / 400+诊所 / 1.1万名医生)"] Ins <-->|"完全一体化运作"| Hosp end Pat["850 万会员患者"] -->|"投保缴费并接受专属诊疗"| Kaiser EMR["全新集团级 EMR (整个组织的底层中枢神经系统)"] -.->|"串联支撑全部业务"| Kaiser ``` - **凯撒医疗 2003 年体量**：30 家综合医院、400 多所诊所、11,000 名在册医生、服务 850 万名患者会员。 - **业务模式特征**：健康险提供方与医疗服务机构完全合一，自负盈亏，打造排他性的闭环医疗网络。 --- ### [01:12:03](https://www.youtube.com/watch?v=FNRU9ZNqcls&t=4323s) 不只是软件：医疗帝国的“中枢神经系统”与 40 亿美元天价大单有人曾向戴维生动地形容，这套系统就是现代医疗体系的神经中枢。面对几十家大型医院、成千上万的医生护士和后勤行政人员，需要一套触手遍布各个角落、复杂至极的庞大软件将一切紧紧编织在一起。系统中包含数以千计的独立界面、严苛复杂的授权鉴权机制、精细入微的角色权限矩阵。称其为“操作系统”甚至都低估了它，它更像是底层操作系统、顶级企业资源规划（ERP）系统以及各类上层专业应用的合体。甚至可以说，传统的 EMR 记录功能，不过是这套庞然大物面向医生这一特定群体时的一个小组件而已。在这场世纪招标中，Epic 最终击败一众对手摘得桂冠。对于当时规模尚小的 Epic 而言，这笔交易彻底改变了公司的命运。就在签约前几年，Epic 的年收入才不过 5,000 万美元左右；而拿下凯撒项目后，Epic 的年化营收直接飙升至 1.62 亿美元，实现了质的飞跃。然而，当年新闻头条纷纷轰动报道的“三年 40 亿美元世纪大单”，在财务口径上却颇耐人寻味。这 40 亿美元其实是凯撒自身的整体项目总预算，其中真正流向 Epic 软件许可与实施服务的金额大概在 4 亿美元左右。 - **40 亿美元总盘子的构成**： - **Epic 实际所得**：约 4 亿美元（前期实施交付费用 + 软件长期使用许可）。 - **医院内部巨额成本**：数万名医院员工直接参与上线与运维的薪资支出。 - **隐形生产力折损**：全系统医生与医护人员在漫长换机培训期内因效率下降造成的实际经济损失。 - **营收倍增效应**：从 5,000 万美元量级飙升至 1.62 亿美元，几乎让 Epic 的年收入在一夜之间翻了数倍。 --- ### [01:15:14](https://www.youtube.com/watch?v=FNRU9ZNqcls&t=4514s) 加州南北割据、波特兰“偷跑”与 IBM 的折戟凯撒当年拿下这套系统的过程充满戏剧性。彼时凯撒内部有两大重心：北加州分部与南加州分部。这两大派系如同同一屋檐下的两家独立公司，各自管理、各自为政，各自拥有不同的系统并试图自建专有 EMR。当时医疗行业正流行一种不切实际的幻觉：医院认为 EMR 应当是自研的核心知识产权，甚至指望未来能包装商用卖给其他同行。于是南北加州两派各自豪掷重金，雇佣埃森哲（Accenture）等顶尖咨询机构帮自己量身定制系统，双方明争暗斗、互不相让。然而，凯撒在太平洋西北区（以俄勒冈州波特兰为基地）还有个毫不起眼的小型分支。这处边缘机构此前悄悄采购了 Epic 的系统，专门用来管理门诊诊所。当时凯撒内部的医生经常发生跨区域调动，不少在波特兰工作的医生轮岗调去加州后，被当地自研系统的糟糕体验震惊了。他们忍不住吐槽加州同僚：“你们花大把冤枉钱请埃森哲搞自研到底图什么？我们在波特兰用的 Epic 哪怕还没上住院部，也比你们这套破玩意儿好用千百倍！”一线医生的口碑迅速发酵，终于促使南北两派握手言和，宣布停战并决定对外公开招标。 ```mermaid flowchart TD accTitle: Kaiser EMR Selection Process accDescr: Flowchart of Kaiser's journey from internal conflict to IBM failure to choosing Epic. subgraph Conflict["内部割据时期"] North["北加州分部 (聘请咨询机构搞自研)"] South["南加州分部 (聘请咨询机构搞自研)"] PNW["波特兰/西北小分部 (门诊悄悄采用 Epic)"] end PNW -->|"医生轮岗调任，口碑反哺加州： 'Epic 体验远胜你们自研' "| Conflict Conflict -->|"内耗终止，达成停火协议"| RFP1["开启全集团统一 EMR 招标"] RFP1 -->|"信奉'买 IBM 绝不会被开除'原则"| IBM["选定 IBM 推进 40 亿美元项目"] IBM -->|"落地惨败，系统无法交付使用"| Fail["项目彻底告吹 / 损失惨重"] Fail -->|"痛定思痛，寻找真正能打硬仗的伙伴"| Epic["转投 Epic，成就行业标杆"] ``` 耐人寻味的是，凯撒最初选择的中标者并不是 Epic，而是老牌巨头 IBM。在企业 IT 领域，长久以来信奉着一句话：“没有人会因为选了 IBM 而被解雇（Nobody gets fired for buying IBM）”。然而，IBM 进场后的表现却演变成了一场灾难——这套庞大项目彻底宣告失败。这种大型 ERP 或 EMR 实施翻车在商业史上屡见不鲜，许多巨头因此白白打水漂数亿乃至数十亿美元。而这恰恰印证了 Epic 绝无仅有的核心生存法则：Epic 赌上了整家公司的声誉，向业界许下“在我们手里绝不出现实施失败”的诺言。这正是 Epic 在医院高管层中建立起极高忠诚度的主因。本和戴维特别强调，当外界提到“客户深爱 Epic”时，这里的“客户”主要指的是医院的 CEO、CIO（首席信息官）和 CFO。虽然一线医护人员的体验极为重要，但真正掏钱买单并决定机构生死的掌舵人，正是这批最害怕系统宕机瘫痪的医院管理层管理者。 ### [01:19:37](https://www.youtube.com/watch?v=FNRU9ZNqcls&t=4777s) 医院的内部权力博弈与 Kaiser 的重新招标在经历了 IBM 主导的系统实施惨败之后，医院遭受了巨大的性能瓶颈与损失，其中受害最深的莫过于一线医生。医生和护士等临床医务人员构成了医院的核心价值与核心竞争力，他们极度渴望系统的稳定运行。对于医院管理层而言，如果不能让医生满意，管理职位本身就岌岌可危，因为医生在组织内部天然拥有极高的议价能力。然而，软件销售的现实却存在某种悖论：销售人员并不会直接向医生兜售系统。医院的运作机制类似于一家传媒巨头——医生、护士和临床专家是镜头前的“台前明星”（on-air talent），但像迪士尼或环球影业这样的公司，真正的商业和采购决策永远由幕后的管理层敲定，医务人员虽参与其中，却非最终决策者。在 IBM 项目宣告流产后，Kaiser Permanente 必须推倒重来、重新发起方案竞标（RFP）。此时，先前在波特兰试点中见识过 Epic 卓越表现的医生们纷纷南下加州，极力向管理层为 Epic 站台唱赞歌，促使 Kaiser 决定认真审视这家来自威斯康星州麦迪逊的“小微企业”。 ```mermaid flowchart TD accTitle: Hospital Purchasing Decision Structure accDescr: Diagram showing how clinical feedback influences administration purchasing decisions. subgraph Talent["临床一线（台前明星）"] MD["医生与护士 提供核心竞争力"] -->|"要求稳定性 / 施加强大内部杠杆"| Admin end subgraph Decision["决策幕后（执行层）"] Admin["医院管理层 / CIO / CEO 真正的采购决策人"] -->|"发布重新招标 (RFP)"| Vendors["软件选型竞标 Cerner vs. Epic"] end ``` --- ### [01:20:45](https://www.youtube.com/watch?v=FNRU9ZNqcls&t=4845s) 大卫与歌利亚：Cerner 与 Epic 的两条发展道路最终进入 Kaiser 决选名单的，只剩下 Epic 及其宿敌 Cerner。Cerner（如今已被甲骨文 Oracle 收购）与 Epic 之间有着惊人的历史巧合：两家公司几乎同时成立于 1979 年，Cerner 由尼尔·帕特森（Neal Patterson）在密苏里州堪萨斯城创立。然而，两家公司走上了截然相反的发展轨道： - **Epic 的路径**：创始人朱迪·福克纳（Judy Faulkner）坚持零风投资金（no VC）、绝不发起并购（no acquisitions）、依托单一底层架构平台（single platform）、永远保持私有化不上市。 - **Cerner 的路径**：走典型的资本扩张路线——大举外部融资、频繁并购整合、公开挂牌上市、迅速做大规模，并拓展至全球市场。在被 Oracle 收购之前，Cerner 前期就已是 24 家不同公司拼凑融合的产物（外加后来整合西门子医疗的 12 家公司），表面套上一层统一的包装。到 2003 年 Kaiser 重新招标时，两者的体量对比悬殊至极：Cerner 是一家年营收近 10 亿美元、经验深厚的跨国上市公司；而 Epic 当时的年营收仅在 5000 万美元出头，完全是一个无名小卒。从常规逻辑来看，选择 Cerner 几乎是 Kaiser 理所当然、无懈可击的安全决策。 --- ### [01:21:55](https://www.youtube.com/watch?v=FNRU9ZNqcls&t=4915s) “所罗门裁决”与朱迪的豪赌：拒绝妥协面对这两家公司，Kaiser 起初试图采取折中方案，如同古代的“所罗门分婴”： - **住院系统（Inpatient）**：交给 Cerner 负责。因为住院部是医院的重中之重与核心基本盘，Cerner 深耕多年，备受信任；而 Epic 刚刚在 2001 年才推出住院模块，市场底蕴尚浅。 - **门诊系统（Outpatient / Ambulatory）**：交给 Epic 负责。Epic 在门诊系统领域首屈一指，且在波特兰门诊诊所的实际落地中给 Kaiser 留下了深刻印象。然而，朱迪·福克纳毫不客气地拒绝了这种拼凑式的妥协。她直言不讳地告知 Kaiser，这种拆分是极度错误的决定：“无论是选我们还是选他们，只要为了你们的患者、整体系统、顺畅的账单结算，以及门诊与住院部之间无缝的数据流转，你们就必须从我们两家之中只选一家做全套系统。” 尽管这种强硬姿态带有一定的战略算计——因为 Cerner 的门诊系统极弱，逼迫对方单挑反而放大了 Cerner 的软肋；但这也是一次风险极高的豪赌。Epic 的住院产品尚缺乏全行业的充分信任，且体量与 Cerner 相比如同蜉蝣撼树。 --- ### [01:23:33](https://www.youtube.com/watch?v=FNRU9ZNqcls&t=5013s) 决定命运的技术尽调：通宵制作的 Excel 算力模型在方案评审进入最胶着的第 11 个小时阶段，Kaiser 安排了一场极具戏剧性的技术尽职调查。双方团队被带入同一座办公楼内相邻的两个会议室，Kaiser 的技术评估小组整天在两个房间之间来回穿梭，每次与一方详谈一小时，并将从另一家听到的方案提炼成尖锐的问题，转头去质询对方。两家团队甚至可以通过 Kaiser 突然抛出的新问题，反向推测隔壁房间正在讲解什么。尽调的核心难题是：Kaiser 拥有 850 万名参保患者、1.1 万名在职医生，系统必须证明自身能承载海量高并发数据的实时流转。Epic 原本的备战方案是一套偏学院派的纯理论架构阐述，论证其单一技术栈在理论极限上可以承受的最高负载。就在评审前夜，Epic 总裁卡尔·德沃夏克（Carl Dvorak）乘夜班飞机飞抵加利福尼亚。在获悉演示方案后，卡尔果断叫停：“不行，不能讲空洞的理论。我们必须在 Excel 里精准模拟出 Kaiser 在全天各个时间段的实际交易峰值流量，演示我们的系统将如何逐一处理这些吞吐，拿出无可辩驳的证据证明我们的带宽和处理能力远超其负荷，绝对不会宕机崩溃。” ```mermaid flowchart LR accTitle: The Technical Due Diligence Showdown accDescr: Step-by-step process of how Epic won the technical bake-off over Cerner. subgraph NightBefore["演示前夜：战略转向"] OldPlan["理论架构推导 （原定方案）"] -->|"卡尔·德沃夏克否决"| AllNighter["通宵搭建 Excel 全天真实并发交易模型"] end subgraph DayOfPitch["尽调当天：交叉问询"] AllNighter --> EpicRoom["Epic 会议室 拿出量化模型数据证言"] CernerRoom["Cerner 会议室 无法提供深度量化证明"] KaiserEval["Kaiser 评估小组 穿梭于两个会议室"] <--> EpicRoom KaiserEval <--> CernerRoom end ``` 卡尔带领团队在酒店通宵达旦地搭建出了精密的动态数学模型。第二天，面对 Kaiser 的高频技术盘问，Epic 拿出了翔实的量化数据与并发推演，而隔壁的 Cerner 显然未做此等深度的细致建模，无法拿出硬核证据证明系统如何扛住交易洪峰。这一轮硬核的技术拼杀，成为整场竞标战局扭转的决定性时刻。 --- ### [01:26:12](https://www.youtube.com/watch?v=FNRU9ZNqcls&t=5172s) 股权诱惑与绝对原则：Epic 荣登行业“T型车” 即便技术评估占据上风，这笔交易仍未尘埃落定。在最终拍板前夕，Kaiser 摆出了超级客户的居高临下姿态，向两家竞标方提出附带条件：要求获得竞标公司的认股权证（warrants/equity）。Kaiser 的逻辑十分霸道：“我们是全美最大的医疗健康集团，这是全行业有史以来最大的订单，给你们这么大的合同，我们理应拿到你们的部分股权。” - **Cerner 的回应**：为了拿下这单惊天交易，Cerner 选择妥协，据传甚至愿意向 Kaiser 出让多达 10% 的公司股权。 - **Epic 的回应**：朱迪·福克纳严词拒绝：“绝不可能。我们不会为你破例，也不会为任何人破例。如果破了这个先例，我们就必须给所有大客户分股权，最终对你们也是灾难。” 面对 Kaiser 索要股权的重压，朱迪和 Epic 团队全程不卑不亢，寸步不让。结合此前在系统架构对决中建立的技术压制，以及坚持门诊、住院与计费三位一体系统架构的连续性，Kaiser 最终在多重疑虑与阻力下，孤注一掷全面选择了 Epic。拿下 Kaiser 合同让 Epic 的年营收一夜之间实现了翻倍乃至翻三倍的跃升。《洛杉矶时报》（Los Angeles Times）在报道此事件时给出了极具分量的评价：“鉴于其庞大的规模，Kaiser 与 Epic 联手打造的这套系统，极有可能成为整个医疗健康软件行业的‘福特T型车’——它或许不是同类产品中诞生的第一辆，但绝对是第一辆走向大众普及的工业标杆。”正是借助 Kaiser 这一全美最大医疗系统的成功背书，Epic 正式跃升为全美医疗信息化系统（EMR）的新一代黄金标准。从此，全美任何一家大型医疗机构的 CEO 或 CIO 在选型时都明白：连规模最大、要求最严苛的 Kaiser 都顶着各种阻力选择了 Epic，Epic 便是行业最不容置疑的选择。 ### [01:28:54](https://www.youtube.com/watch?v=FNRU9ZNqcls&t=5334s) 凯撒项目落地与飞速扩张凯撒医疗（Kaiser Permanente）的成功上线成为了 Epic Systems 发展史上最关键的转折点。起初大家可能只是抱着观望态度，但几年后整套系统顺利投入运营，不仅没有“翻车”，反而运转得极其出色。在 2006 至 2008 年间，全美医疗机构的大门彻底向 Epic 敞开。到 2007 年，Epic 的年收入已飙升至 5 亿美元，是签下凯撒订单时规模的 3 倍，更是拿下凯撒之前的近 10 倍。曾经的小作坊彻底蜕变成了一家庞然大物。一位 Epic 前员工在回忆这一转折点时提到，在赢得凯撒大单之前，公司每个月大概只从大学招收 10 名应届毕业生；而在这之后，每个月涌入公司大门的毕业生达到上百人之多，唯有如此庞大的人才规模才能支撑起业务的爆炸式扩张。而这种招募规模的急剧扩大，也直接促成了后来闻名业界的维罗纳（Verona）园区建设。 - **营收跃升**：2007 年收入突破 5 亿美元，相较拿下凯撒医疗之前增长近 10 倍。 - **人员扩张**：校招节奏由原先的“每月 10 名应届生”激增至“每月数百人涌入”。 --- ### [01:29:59](https://www.youtube.com/watch?v=FNRU9ZNqcls&t=5399s) 飞利浦合作败局：闭源独立路线的确立在 2000 年代初期，Epic 经历了一段鲜为人知却深刻影响其发展轨迹的插曲。当时荷兰巨头飞利浦（Philips）主动找上门来，提议联合研发：由 Epic 承担核心软件开发，飞利浦提供客户关系与分销渠道，共同主攻放射学（Radiology）领域。不仅如此，飞利浦还希望获得 Epic 现有整个 IT 系统的授权，贴上“Philips Externity Enterprise”的自有品牌对外销售。对于渴望打入欧洲等国际市场的 Epic 来说，这种由硬件巨头带路的分销模式看似极具诱惑力，如同当年惠普贴牌销售苹果 iPod（HP iPod）一般。然而，这趟看似美好的合作最终演变成了一场昂贵且痛苦的弯路。Epic 甚至在荷兰招兵买马组建团队，耗费数年心血完成开发与产品上线，但整套合作方案在发布一两年后便彻底夭折。这次惨败在 Epic 内部留下了极其深刻的“伤疤组织”（scar tissue），让团队对一切外部合作模式产生了本能的戒心：在他们看来，“合作伙伴”意味着无法控制的外部变量、不可承受的风险，以及历史上从未成功过的先例。 ```mermaid flowchart TD accTitle: Epic与飞利浦合作及战略转变 accDescr: 展示Epic与飞利浦合作失败后形成伤疤组织并转向全面自主掌控的流程 subgraph Collab["飞利浦合作模式（失败教训）"] A["飞利浦提出贴牌合作 （Philips Externity Enterprise）"] B["Epic负责底层开发 荷兰团队与重金投入"] C["产品上线1-2年后宣告夭折 极其昂贵的弯路"] A --> B --> C end subgraph Transformation["战略反思与文化沉淀"] D["形成合作伤疤组织 （不可控、高风险、依赖外部）"] E["确立自研与直销战略 （拒绝贴牌、坚守定价、闭环生态）"] C --> D D --> E end ``` 这次失败直接塑造了 Epic 后续几十年饱受争议却无比坚固的商业基因：拒绝任何华而不实的第三方集成，坚决剔除不可控的外部依赖，坚持 100% 自研、直销客户、明码标价绝不妥协。虽然外界常常抨击 Epic “封闭、缺乏互操作性”，但这一行事风格正是当年与飞利浦那段失败合作所烙下的深深印记。 - **贴牌计划**：飞利浦计划将 Epic 的整套医疗软件冠以“Philips Externity Enterprise”品牌销售。 - **沉重代价**：在荷兰设立分支机构、耗费数年研发，项目推向市场一两年后全面溃败。 - **核心教训**：Epic 自此坚守“只做自己能完全掌控的事”，在长达 15 到 20 年的时间里对外部合资与贴牌合作敬而远之。 --- ### [01:33:34](https://www.youtube.com/watch?v=FNRU9ZNqcls&t=5614s) 微软雷德蒙德启发与千亩农场选址手里握着数亿美元的充沛现金流，Epic 决定重金押注未来，而落脚点正是其位于威斯康星州维罗纳（Verona）的奇幻总部园区。关于维罗纳园区的诞生，有着两个相辅相成的起源故事。第一个故事与微软有关：当时朱迪·福克纳（Judy Faulkner）的儿子正在微软雷德蒙德（Redmond）园区担任软件工程师，而朱迪与丈夫卡尔（Carl）一直对微软的运营哲学颇为推崇。趁着探望儿子的契机，朱迪顺便参观了雷德蒙德园区。在当时那个“Googleplex”尚未声名大噪的时代，微软园区就像是科技界的童话乌托邦。开阔的大学校园式氛围、低矮的二三层小楼、大片绿地步道，以及“每位工程师都享有独立办公室”的设计，给朱迪留下了极深震撼。此前 Epic 挤在一所改造后的老校舍中办公，早已无法承载高速扩张的需求。朱迪回到威斯康星后，几乎全盘借鉴了微软的园区逻辑，一口气在麦迪逊郊区半小时车程的维罗纳买下了整整 1000 英亩农田，正如微软塑造了雷德蒙德一样，Epic 从此在精神与地理上彻底成为了维罗纳的主宰。 - **灵感源头**：朱迪参观儿子任职的微软雷德蒙德园区，深受其大学校园般低密度的单人办公室风格启发。 - **大手笔拿地**：在麦迪逊郊外的维罗纳一口气购入 1000 英亩（约 404 公顷）农田建设新总部。 --- ### [01:35:50](https://www.youtube.com/watch?v=FNRU9ZNqcls&t=5750s) 会议室壁炉传说与“迪斯尼级”奇幻园区的诞生第二个故事则源于老校舍时代的一场意外装修。当时设计师在改建麦迪逊老校舍的主会议室时，为了营造温馨居家感，特意装了一座壁炉。朱迪和团队灵机一动，干脆将错就错，往屋里挂上了雪鞋、皮草和斧头，硬生生将其装潢成了一间“威斯康星湖畔小木屋”（Wisconsin lakes lodge）。令人意想不到的是，这间小木屋成了整栋楼里最受欢迎的地方，所有造访麦迪逊的医院客户都指名要在这里开会。这一奇效让朱迪大受启发：既然客户喜欢这种沉浸感，那在新园区里就应该把这种风格放大十倍、彻底做绝。在规划维罗纳园区时，Epic 双管齐下：他们不仅聘请了操刀微软雷德蒙德园区的设计公司，更请来了在 2008 年负责加州迪士尼加州冒险乐园翻新的专业团队。最终呈现在世人面前的，是一个融汇了《爱丽丝梦游仙境》、《哈利·波特》和《绿野仙踪》翡翠城的超现实梦幻王国。这片看似荒诞离奇的园区，对刚出校门的年轻毕业生具有无与伦比的吸引力——它让工作环境延续了象牙塔般的趣味与活力。而这正折射出 Epic 极其独特的一体两面企业文化：表面上是古灵精怪、充满童话色彩的烂漫乐园，骨子里却是雷厉风行、不惜一切代价追求胜利的残酷战舰。这种看似矛盾却高度自洽的 DNA，恰恰就是创始人朱迪·福克纳本人的真实写照。 - **设计团队**：同时聘用微软雷德蒙德园区的主力建筑事务所，以及迪士尼加州冒险乐园（2008 年翻新工程）的操刀团队。 - **园区主题**：深度还原《哈利·波特》、《绿野仙踪》、《爱丽丝梦游仙境》等经典奇幻 IP 场景。 - **双重基因**：外表天真烂漫的“童话乐园”与内在极度硬核、追求卓越的“竞争机器”深度融合。 ### [01:38:25](https://www.youtube.com/watch?v=FNRU9ZNqcls&t=5905s) 威斯康星农场上的科技乐园：Deep Space 地下大礼堂为了吸引最聪明、年轻且充满冲劲的应届大学毕业生前往威斯康星州维罗纳（Verona, Wisconsin），Judy Faulkner 和 Epic 选择为他们打造一个世外桃源般的园区。整个园区总占地达 1,700 英亩，其中 410 英亩为建筑核心区，其余土地仍保留为农田，园区内总计分布着 89 栋建筑。园区拥有四个室内大型礼堂，总计容纳 18,000 个座位。其中最震撼的是名为“深空”（Deep Space）的地下礼堂——这是全球最大的地下礼堂，拥有惊人的 11,400 个座位，深入地下达 74 英尺。主播们惊叹道，这个体量相当于把两个无线电城音乐厅（Radio City Music Hall）的人流塞进地下，其规模甚至更接近金州勇士队的大通中心（Chase Center）等专业 NBA 篮球馆，远超普通企业礼堂的概念。 ```mermaid flowchart TD accTitle: Deep Space Auditorium Capacity Comparison accDescr: 比较 Deep Space 地下大礼堂与著名公共场馆的座位规模 subgraph Epic ["Epic 维罗纳园区"] DS["Deep Space 地下大礼堂 地下 74 英尺 / 11,400 座"] end subgraph Arenas ["规模对比参考"] RCM["Radio City Music Hall 约 6,000 座 × 2"] Chase["Chase Center 篮球馆 约 18,000 座"] end DS -. 规模相当.-> RCM DS -. 接近职业球馆.-> Chase ``` Epic 当初决定建造 Deep Space 的初衷非常质朴且务实：早期公司每个月租用一家当地电影院来召开全员大会（All-hands Meeting），搬到新园区后，管理层希望能在自己的地盘上开会。当时他们乐观地认为“我们员工人数绝不可能突破 11,400 人”，建造如此庞大的空间不仅能满足公司内部全员大会的终极需求，还能接待所有客户和整个生态系统的来访。然而颇具戏剧性的是，随着 Epic 后续业务的迅猛扩张，员工人数迅速超出了当初的设想，如今连 Epic 自己的员工都无法全部塞进这个原本以为“永远坐不满”的地下大厅了。 - 园区总占地 1,700 英亩（410 英亩园区办公建筑，其余为农田），共计 89 栋建筑。 - 4 个室内礼堂总计提供 18,000 个座位。 - Deep Space 礼堂拥有 11,400 个坐席，向下挖掘深度达到地下 74 英尺。 --- ### [01:39:58](https://www.youtube.com/watch?v=FNRU9ZNqcls&t=5998s) “软件工厂”的运营哲学与洗手间里的十诫在理解 Epic 的企业文化之前，必须首先明白 Judy Faulkner 对公司的定位：它不是一家普通的科技公司，而是一座极其严密的“软件工厂”（Software Factory）。在 Judy 看来，Epic 从不收购外部公司，在内部也摒弃了一切与“构建软件”无关的边缘职能；公司的唯一使命就是招募开发者，然后如同精密车间一般，源源不断地把脑力转化为交付给医疗行业的代码。主播打趣道，当今英伟达（NVIDIA）和戴尔（Dell）常挂在嘴边的“AI 工厂”，其实 Judy 早在几十年前就已经把这种纯粹的“软件工厂”模式付诸实践了。更具传奇色彩的是 Epic 园区内无处不在的“摩西十诫”（The Epic Ten Commandments）。这份纲领被醒目地张贴在园区内每一个洗手间和茶水间的墙壁上，向员工与所有外部访客公开展出： 1. **不可上市**（Do not go public） 2. **不可收购其他公司，亦不可被收购**（Do not acquire or be acquired） 3. **软件必须好用/有效**（Software must work） 4. **现实必须符合预期**（Reality equals expectations） 5. **信守承诺，即使是心照不宣的承诺**（Keep commitments, even the unspoken ones） 6. **专注胜任能力，绝不容忍平庸**（Focus on competency. Do not tolerate mediocrity） 7. **坚守标准，公正对待所有人**（Have standards. Be fair to all） 8. **要有勇气；你所容忍的，就是你所立足的**（Have courage. What you put up with is what you stand for） 9. **传授哲学与文化**（Teach philosophy and culture） 10. **保持节俭，绝不通过负债进行日常运营**（Be frugal. Do not take on debt for operations）主播们重点指出，前两条“不上市”与“不并购”本是只有 CEO 本人才能决定的最高决策，但 Judy 执意把它们贴进卫生间，不仅是为了将自主独立、永不妥协的DNA深植于每位员工的骨髓，更是向每一位来访医院客户给出的定心丸——昭示着 Epic 永远是一套统一、连贯且不受资本市场操控的系统。耐人寻味的是，在这十条诫命里，竟然没有任何一条直接提及“患者”、“挽救生命”或“医疗”。这纯粹是 Judy 关于“如何运营一家顶级软件工厂”的终极治企宪章。除此之外，园区里还有一个独特的仪式：每当 Epic 与新客户签约时，全园区都会鸣响结婚钟声（Wedding Bells），象征着他们与客户从此结下一生相守、永不背弃的“契约婚姻”。 --- ### [01:43:09](https://www.youtube.com/watch?v=FNRU9ZNqcls&t=6189s) 极致标准化的入职熔炉与高信任低管理机制对于刚走出大学校园的高材生来说，Epic 是一座近乎严酷但也极其强大的训练营。外界常戏称 Epic 的文化带有“邪教倾向”（cult-like），但从企业效率角度看，这恰恰成就了其非凡的执行力。新员工入职后，公司会从零开始格式化并规范他们的所有工作习惯：甚至连如何用黄色便签本（yellow legal pad）记笔记、如何撰写一封符合规范的电子邮件，都有极其严格且经过数十年数据迭代验证的“Epic 标准流程”。这种方法甚至可以追溯到当年 Neil Pappalardo 与 MEDITECH 的渊源。通过将每个人标准化，Epic 打造了一个高度确定、彼此无缝互信的运转机器。 ```mermaid flowchart LR accTitle: Epic Internal High Trust Operational Loop accDescr: 展示 Epic 内部由高度流程规范支撑的扁平化与高信任运作逻辑 A["应届毕业生入职"] --> B["严苛标准化培训 (笔记/邮件/工作流)"] B --> C["高度一致的行为模式 与极致互信"] C --> D["极少中层管理 无年度部门预算限制"] D --> E["全员临床轮岗沉浸 直接对接真实业务"] ``` 得益于这种深入骨髓的共识与信任，Epic 几乎不存在臃肿的中层管理层（middle managers），也根本没有传统大公司的部门年度预算（budgets）。Judy 过去拜访医院客户时，经常听到客户推脱“这件事是对的，但我今年没预算了，只能推迟到明年”，或者听到“快到年底了，我得赶紧突击花光预算免得被砍”。Judy 痛斥这种预算体制是“愚蠢透顶的”，因此在 Epic 内部断然废除了硬性预算机制。在职位体系上，Epic 极其扁平与淡化头衔，无论是工作了20年的资深员工还是入职仅6个月的新人，名片上往往都只简简单单印着“实施顾问”（Implementation）。此外，Epic 贯彻着极致的客户沉浸机制：所有员工，包括核心软件工程师，都必须亲赴临床一线，甚至深入手术室实地观摩医护人员的操作流程。新员工入职阶段必须完成 5 次临床轮岗，此后每年也必须定期回访医院，将 Y Combinator 所推崇的“深入客户现场”做到了极致。 --- ### [01:45:40](https://www.youtube.com/watch?v=FNRU9ZNqcls&t=6340s) 零容忍的零 Bug 纪律：为生命安全与医院营收护航在软件工程层面，Epic 实行的是一种极为严格、以“消除 Bug”为唯一导向的开发规范。新工程师加入后要接受长达 6 个月的密集技术培训。Epic 系统的核心诉求是：尽可能缩短“一行代码被写出”到“该代码被测试”之间的时间间隔。一旦在测试中发现 Bug，编写该代码的原始开发人员必须立刻放下手头所有事情去修复它。Epic 严格遵循“谁写的代码谁修 Bug”（Fix your own bugs）的铁律，绝不允许 Bug 拖延数周数月，从而在开发人员对上下文记忆犹新时就掐灭隐患，避免错误如滚雪球般在底层架构中蔓延。这种开发模式并不追求以最快速度交付花哨的功能，也不追求采用行业最炫酷的技术栈，而是把系统稳定性提升到了压倒一切的地位。主播们总结，这种对“零 Bug 环境”近乎偏执的追求背后有着两大致命原因： 1. **生死攸关（Lives are on the line）**：在医疗核心系统中，开药剂量的哪怕一个细小软件漏洞，都直接关乎患者性命，软件出故障就是致命事故。 2. **复杂的营收周期与计费合规**：医院计费与收入结算系统极度错综复杂。一旦软件在计费环节发生逻辑错误，医院轻则损失天文数字的营业收入，重则可能直接触犯联邦医保欺诈罪名（Medical Fraud）而招致联邦刑事诉讼。因此，Epic 打造的软件工厂从基因上就必须做到万无一失。 ### [01:47:17](https://www.youtube.com/watch?v=FNRU9ZNqcls&t=6437s) 严苛的实施前线：高压、极度敬业与“赢”的文化在 Epic Systems 担任实施工程师（Implementation）所面对的工作强度简直令人咋舌，其复杂度堪比军事级的后勤统筹。正如现代管理学之父彼得·德鲁克（Peter Drucker）的名言所指出的，医院是人类历史上尝试管理的最复杂的组织形式。每个刚走出校门的年轻人都需要同时处理多个医院客户，每天梳理跟踪成百上千项错综复杂的依赖关系与执行状态。这是一项极其高压、高肾上腺素但又担负极高责任的领导岗位。年轻员工每天工作十到十二个小时是家常便饭，但许多人对此乐此不疲，因为他们刚刚毕业就能站在最前沿做大事，并且身边全是同等聪明的伙伴。Epic 倾向于招聘高智商、高情商且性格温和务实的“中西部好孩子”（Midwestern kids）来直接面对客户。这种扁平的组织架构让年轻员工能直接与公司最高管理层并肩作战。比如年轻程序员 Sumit，其日常搭档便是总裁 Carl（Carl 本人也是一线程序员兼团队带头人）。对于极具野心、专注于职业成长的人而言，能与同样优秀的精英团队在极高赌注的环境中持续打胜仗，本身就是无与伦比的乐趣。这种“打胜仗的快感”加上园区童话般的奇思妙想（whimsy），构成了独特的 Epic 文化。结果就是：这份工作成了员工生活的绝对核心。加上园区坐落在威斯康星州相对偏僻的乡村（离麦迪逊市还有一定距离），员工几乎没有接触其他跳槽机会的干扰，Epic 顺理成章地接管了你的全部生活。节目主播将这一氛围与两个著名企业相类比： - **80至90年代早期的微软（Microsoft）**：同样位于相对隔离的园区，充满极度聪明、敬业甚至狂热的年轻技术人员，生活与工作完全融为一体。 - **Palantir 的前线部署模式**：将训练有素、武装完备且意气风发的顶尖年轻人直接“空投”进全球最严苛复杂的客户现场，用强大的流程与工具去扫清一切障碍。当然，硬币的另一面是残酷的“不进则退”（Up or out）末位淘汰机制。Epic 会以极高强度考验新员工，果断淘汰表现不达标的人。因为招募大量新人来筛选胜任者，比长期保留冗员更符合成本效益。前几年的高流失率确保了最终留下来的老员工各个都是绝对值得信赖的中流砥柱。 ```mermaid flowchart TD accTitle: Epic 实施模式与文化生态 accDescr: 展示 Epic 如何通过年轻高素质团队、末位淘汰以及孤岛式园区构建高凝聚力组织 subgraph Talent["人才与前线"] A["高智商/高情商年轻毕业生"] --> B["军事级现场实施"] B --> C["管理全球最复杂组织（医院）"] end subgraph Culture["文化与组织机制"] D["与高管并肩作战（扁平化）"] E["残酷淘汰机制（Up or Out）"] F["威斯康星乡村园区（沉浸式生活）"] end B --- D B --- E Talent --> F ``` --- ### [01:50:11](https://www.youtube.com/watch?v=FNRU9ZNqcls&t=6611s) 传奇的招募测试：初中生编写的代码考卷与去面试化 Epic 的招聘流程在科技界别具一格：早在几十年前，他们就开创了完全依赖逻辑与编程测试的筛选机制。这段故事源于创始人朱迪·福克纳（Judy Faulkner）的家庭轶事。在 20 世纪 80 年代末，朱迪在面试软件开发人员时发现，传统的面对面交谈根本无法预测一个人能否成为优秀的程序员。与此同时，她还在上初一（约七年级）的儿子经常参加全州的编程竞赛并屡获大奖。朱迪发现这类编程测试对程序员的实际能力有着惊人的预测效度，于是便问儿子：“你能不能帮我编写一套这种类型的测试题，我拿来在 Epic 招程序员时测测看他们到底行不行？”令人称奇的是，这套由一个初中生捣鼓出来的编程测试，竟然在随后的 **18 年里** 成了 Epic 选拔软件工程师的核心试卷（直到后来答案在互联网上传播开才停用）。不仅如此，Epic 的招聘高度依赖测试成绩： - **测试成绩达标即发 Offer**：在多数情况下，应聘者只要在笔试中拿到高分，几乎不需要通过层层面试就能直接收到录取通知（虽然后续会邀请前往园区参观）。 - **全员逻辑测试**：测试体系后来演进出 Rembrandt 测试等多种标准化体系，公司所有岗位入职前都必须通过特定测试，甚至连后勤餐饮团队（Culinary Team）都要在入职前参加入门逻辑测试。 - **领先行业数十年的实践**：在 80 年代，这种基于硬核逻辑测试去面试化的招聘流程，远远早于后来 Google 赖以成名的算法测试体系。 --- ### [01:52:17](https://www.youtube.com/watch?v=FNRU9ZNqcls&t=6737s) 极致以客户为中心：没有销售部的软件巨头在外部业务运营上，Epic 展现出了近乎偏执的“以客户为中心”（Customer-obsessed）。芝加哥拉什大学医疗系统（Rush University System for Health）首席信息官杰夫·戈特尼（Jeff Gotney）曾留下经典评价：“虽然 Epic 绝对不便宜，但你花出去的每一分钱，100% 都能得到实打实的回报。”在行业中，Epic 以极其可靠、从不画饼虚标、交付完全一体化而闻名。正如曾经科技界的行业共识“买 IBM 绝不会被炒”一样，如今在医疗行业，“选择 Epic 也绝不会犯错”。 Epic 与客户之间的互动流程极其独特： - **客户投票决定产品路线图**：每年全体客户代表飞抵威斯康星园区参加年度大会，Epic 会现场征集想法并组织客户直接投票，由此决定下一阶段开发哪些功能模块。 - **全公司仅三大核心岗位**： 1. **软件开发人员（Software Developers）**：负责底层与功能构建； 2. **项目经理（Project Managers / Implementation Managers）**：负责新客户的现场部署上线； 3. **技术专家（Technical Specialists）**：负责既有客户的持续运维支持。 - **全公司没有专门的销售或营销部门**：全公司仅有大约 8 名所谓的“销售人员”，他们从不进行主动地外部电销或拉客，只负责响应打进来的入站咨询（Inbound requests），且这 8 个人全都是从项目经理或技术专家晋升上来的老手。 --- ### [01:53:47](https://www.youtube.com/watch?v=FNRU9ZNqcls&t=6827s) 专属“BFF”、年度成绩单与极其强势的标准打包在所有岗位中，**技术专家（Technical Specialists）**是 Epic 内部规模最庞大的群体。针对全美 600 多家医院客户系统，Epic 为每家客户在他们使用的**每一个产品模块**上都配备了专门的技术专家支持团队（从核心的 EpicCare EMR、MyChart，到计费模块 Resolute、数据研究网络 Cosmos 等，各模块均有专属对接班底）。更绝的是 Epic 为每家客户指派的 **“BFF”（Best Friend Forever）** 机制： - 每个医院客户都会分到一位 Epic 内部的唯一接口人，其唯一职责就是对该客户使用 Epic 软件的最终成功全权负责。 - **每年向 C-Suite 发送评分成绩单**：BFF 会参照全美同等体量医疗机构的基准表现，每年给该医院的 CEO、CIO 和 CFO 分别寄送独立评估报告。报告在多个维度上给客户打出 1 到 5 分的成绩，并附上其与同行竞品的详细横向对比数据。这种深入骨髓的服务伴随着极高的议价权与强势立场。尽管 Epic 终极导向是让客户满意，但他们绝不任由医院随意定制。在 2025 年的当下，Epic 是高度“有态度、重标准”（opinionated）的系统： - **推行标准套件（Standard Package）**：竞争对手 Cerner 的每次部署往往千奇百怪、高度碎片化；而 Epic 则极力迫使医院采纳标准化模块，以确保底层互操作性（Interoperability）顺畅无阻，并能一键推送全网系统更新。 - **宁缺毋滥的拒单底气**：如果医院非要坚持按自己的错误习惯魔改系统，Epic 拥有极强的议价底牌。由于他们每年全美仅接收 10 至 30 家新客户，如果院方拒绝遵循最佳实践，Epic 会直截了当地告诉对方：“我们认为你们尚未准备好成为 Epic 的客户，今年我们先服务别人，等你们彻底准备好了我们再合作。” ### [01:56:21](https://www.youtube.com/watch?v=FNRU9ZNqcls&t=6981s) 标准化实施的定价杠杆与“利他资本主义” 在深入拆解 Epic 的商业策略时，主持人指出，医院想要在 Epic 那里拿到价格折扣，几乎只有一种途径：采用完全或绝大部分的“标准化实施”（standard implementations）。Epic 内部很可能设有严格的分级体系，客户的系统配置越接近标准模版，折扣力度就越大；反之，一旦医院偏离标准进行定制，需要支付的费用就会呈指数级上升。此外，只要客户积极配合系统与数据库的日常维护、按时完成版本升级，Epic 也会给予相应的费用减免。这种策略表面上看是 Epic 为自己不断聚拢控制力，但其背后还流淌着一种与客户深度绑定的“利他主义”色彩。这种独特的风格，很大程度上源于创始人朱迪·福克纳（Judy Faulkner）的个人价值观——一种被主持人形容为“极度好胜、追求价值最大化，却又带有道德利他色彩的资本主义”（ultra-competitive, ultra value-maximizing, altruistic capitalism）。尽管 Epic 的软件造价极其高昂，但只要机构签约成为正式客户，Epic 几乎从不主动大幅涨价。其软件服务年费的平均涨幅常年维持在 2% 左右，甚至低于常规通胀率，这与硅谷其他企业级 SaaS 动辄大幅提价的做法截然不同。坊间甚至一直流传着一个说法：Epic 旗下的患者门户 MyChart 在整个历史中从未调整过价格。虽然从客户最终拿到的整体账单来看，单项未涨价可能具有一定的迷惑性，但不可否认的是，Epic 在骨子里维持着惊人的“客户至上（customer-obsessed）”执念。 - **实施折扣的核心法则**：只有严格采用标准化模版部署、积极跟随版本升级和数据库维护的客户，才能换取价格上的减免；个性化定制越多，账单越昂贵。 - **克制的提价策略**：在入场门槛极高的前提下，签约客户的年平均合同涨幅仅约 2%，以低于通胀的速度保持长期合作。 --- ### [01:58:01](https://www.youtube.com/watch?v=FNRU9ZNqcls&t=7081s) 客户至上的本质：通过免费捆绑构建绝对统治这种利他姿态与商业利益并不矛盾，而是 Epic 谋求长期霸权的最优解。对医院和大型医疗系统的 CEO、CIO 而言，采购 Epic 的根本目的，是在合规与风险最低的前提下，将每一次与患者的接触最大程度地转化为系统的合法收益。而 Epic 的目标则是拿下合同后长盛不衰、永久扎根，最终实现“一统医疗江湖”。为了达成这一目的，Epic 常常做出一些看似替客户着想、实则深谋远虑的商业举措。最典型的例子莫过于第三方软件的引入与自研捆绑。当医疗行业出现某种新兴需求时（例如 2020 年新冠疫情爆发初期急剧膨胀的远程医疗 Telehealth），Epic 不会强行阻拦，而是主动向医院推荐合规的第三方解决方案（如符合 HIPAA 隐私标准的 Zoom 方案），甚至大方地表示：“你们现在可以放心先用 Zoom。顺便提一句，我们自己的远程医疗模块正在研发中，马上就会上线；而且等它上线时，作为老客户的你们可以**完全免费**使用。” 这一策略直接打消了客户去市场上大肆采购专业独立竞品的念头。客户很清楚，Epic 新出的模块刚开始可能功能简陋，但只要等它迭代得足够好用，医院就会毫不犹豫地换回 Epic 自家的免费模块，将其无缝整合到现有的企业级框架协议中。这种巧妙的产品捆绑与蚕食策略，手法酷似当年的微软。正如杰夫·贝索斯（Jeff Bezos）早在 1999 年就阐述过的商业哲学——“从长远来看，客户想要的与公司想要的长远目标完全一致”。Epic 全力满足医院管理层的诉求，最终也将自己铸造成了医院体内无法被剥离的“中央神经系统”。 ```mermaid flowchart TD accTitle: Epic 免费功能捆绑与生态防御机制 accDescr: 展现 Epic 如何面对外部新兴技术需求，先推荐第三方过渡，再推出自研免费模块替换竞品，实现终身锁定的闭环。 subgraph Phase1["第一阶段：需求爆发与过渡"] Demand["医疗行业出现新兴功能需求 (例如：2020年远程医疗爆发)"] --> Recommend["Epic 主动向客户推荐合规第三方方案 (例如：HIPAA 兼容的 Zoom)"] end subgraph Phase2["第二阶段：自研与免费捆绑"] Recommend --> Build["Epic 内部快速开发同类模块"] Build --> FreeBundle["自研模块向现有客户『免费赠送』 纳入既有企业级框架协议"] end subgraph Phase3["第三阶段：生态替换与绝对锁定"] FreeBundle --> Evolve["自研功能逐步迭代至『足够好用』"] Evolve --> ChurnOut["医院淘汰外部独立第三方软件 全面回归 Epic 统一系统"] end ``` --- ### [02:00:23](https://www.youtube.com/watch?v=FNRU9ZNqcls&t=7223s) 赞助商插播：ServiceNow 与统一平台理念在深入讨论改变整个医疗信息技术行业的宏观政策转折点之前，主持人顺势接入了节目长期赞助商 ServiceNow 的商业动态。他们强调，外界通常将 ServiceNow 视为企业级 AI 操作系统，却往往忽略了其麾下悄然孕育出的一笔规模超 10 亿美元的 CRM 业务。与数十年前那些纯粹为了“记录数据和汇报工作”而设计的传统 CRM 不同，ServiceNow 的核心逻辑是将销售、客服与底层企业运营完全打通。如果企业各部门系统相互割裂，就会迫使员工充当所谓的“人肉中间件”（human middleware），在各个孤岛系统之间来回手动倒腾数据；而基于单一真相来源（single source of truth）的统一平台，才能真正让 AI Agent 在跨部门调用库存、物流与售后数据时直接解决问题。主持人笑着打趣道，这种用底层单一平台穿透所有前端业务的底层架构，与本期节目贯穿始终的 Epic 哲学简直如出一辙。 --- ### [02:03:21](https://www.youtube.com/watch?v=FNRU9ZNqcls&t=7401s) 被误读的历史起点：小布什时代的医疗 IT 序幕回顾完 Epic 的园区文化与商业版图，时间线推进到了 2000 年代中叶的关键转折点——2006 年。外界普遍存在一种根深蒂固的历史叙事，认为全美医疗机构大范围采纳电子病历（EMR）及后续严苛的“有效使用”（Meaningful Use）法案，完全是由奥巴马政府和《平价医疗法案》（Obamacare）一手推动的产物。然而翻开真实的历史档案会发现，这一彻底颠覆医疗行业格局的运动，实际上具有极其深厚的两党共识，且早在小布什总统执政时期便已吹响号角。在 2006 年的国情咨文演讲中，乔治·W·布什（George W. Bush）向全美明确提出： > “为了全体美国人民，我们必须直面不断飞涨的医疗成本，巩固医患关系，并帮助人们承担得起他们所需的医疗保险。我们将更广泛地利用电子病历（electronic records）与其他卫生信息技术，以此帮助控制成本并减少致命的医疗差错。” 紧接着，布什还在演讲中呼吁让医疗保险具备“便携性”（portable coverage），让劳动者在换工作时不必担忧断保——在仅仅四年后奥巴马医改引发的激烈党派撕裂中，这种主张由一名共和党总统如此高调地提出，显得耐人寻味。这揭示了一个更深层的社会背景：在 2006 年的美国，医疗体系的千疮百孔已成举国公认的顽疾。疯狂攀升的“成本病”（cost disease）侵蚀着巨额 GDP，整个系统庞大而又低效至极。从白宫到两党议员，所有人都迫切希望能找到一把手术刀来革除弊病，而拥抱电子医疗记录，被当时的高层视为了拯救美国医疗系统最核心的灵丹妙药。 - **小布什 2006 国情咨文的定调**：正式确立国家级战略，寄望通过普及“电子医疗记录与卫生 IT 技术”来压降医疗开支、降低医疗事故。 - **两党共识的起源**：电子病历现代化的浪潮绝非始于奥巴马，而是由小布什政府铺垫、并贯穿后续金融危机经济刺激政策的历史必然。 ### [02:05:32](https://www.youtube.com/watch?v=FNRU9ZNqcls&t=7532s) 医疗数字化的承诺与“互操作性”的三大层级在2000年代中期，美国医疗系统面临着深层的机制性问题，虽然大家在具体根治方案上莫衷一是，但“全面推行数字化并激励采用电子病历（EMR/EHR）”被普遍视作走出困境的关键一步。当时全美仅有大约 13% 的医疗机构配备了某种形式的 EHR 系统——尽管业内后来有人辩称这一数字偏低是因为当时对 EHR 的定义发生过改变，但毫无疑问，彼时的医疗依然是一个极度缺乏数字化的传统行业。在深入具体的立法变革之前，必须厘清围绕在 Epic 及所有 EHR 厂商周围的核心命题：互操作性（Interoperability）。这个概念主要可以划分为三个层级： ```mermaid flowchart TD accTitle: 医疗数据互操作性的三个层级 accDescr: 展示 Epic 在面对不同系统和机构时互操作性的三种不同场景与难度 subgraph L1["第一层级：同构互通"] A["Epic 系统 (医院 A)"] <-->|"Care Everywhere / 相同架构"| B["Epic 系统 (医院 B)"] end subgraph L2["第二层级：跨厂商互通"] A <-->|"跨厂商标准缺失 / 竞争壁垒"| C["其他 EMR (如 Cerner / MEDITECH)"] end subgraph L3["第三层级：生态扩展"] A -.->|"严格受控的 API / 数据安全考量"| D["第三方应用程序与开发者"] end ``` 第一类是最容易实现的同构互操作，即不同医院之间的 Epic 系统相互连接。因为采用完全相同的技术架构与底层设计，这种记录迁移在技术上顺理成章。第二类则是不同医院的不同 EMR 之间的互联，例如从 Epic 传输到 Cerner、MEDITECH、Allscripts 或各机构自行研发的专有系统。这在技术上面临着巨大鸿沟：各厂商底层架构各异，除了 MEDITECH 等少数系统外多数并非单一数据库，且医疗行业的标准统一化程度远远落后于通用型 B2B SaaS 软件。更关键的是背后的利益博弈：医院之间存在激烈的商业竞争，彼此缺乏向对手开放患者数据的动力；而 Epic 作为商业实体，除了“为了患者福祉”这一崇高伦理外，在商业利益上也完全没有动力去便利对手系统。第三类互操作性则发生在 Epic 与试图基于其数据开发、搭建在其生态之上的第三方应用程序之间。多年来 Epic 对此极其审慎与克制。Epic 官方曾公开强调，正是由于他们对向第三方开发者共享数据采取了极度严密的审核防范，才从来没有发生过类似 Facebook 那样骇人听闻的“剑桥分析”（Cambridge Analytica）式数据泄露丑闻。 - **全美数字化现状**：在2000年代中期，全美仅约 13% 的医疗机构上线了 EHR 系统。 - **HIPAA 合规与护城河**：保护涉及 HIPAA（健康保险可移植与责任法案）的隐私数据是合理的合规借口，但在客观上也构成了 Epic 极具战略价值的商业防壁，使得外部开发者接入 Epic 的难度远高于其他软件领域。 --- ### [02:08:49](https://www.youtube.com/watch?v=FNRU9ZNqcls&t=7729s) Care Everywhere 的误打误撞与医疗资本的内在冲突在 Epic 体系内部的“Epic 到 Epic”数据交换上，他们打造的平台 Care Everywhere 取得了惊人的落地成效。如今 Care Everywhere 每天在全美处理并交换超过 2,000 万份患者病历。节目主持人提到，自己在西雅图的三家不同医疗系统开设了 MyChart 账户，筹备本期节目时将它们一键关联合并，体验极其顺畅。然而，这一杀手级功能的普及却源于一个充满戏剧性的历史意外：Judy Faulkner 当时亲自拍板，决定 Care Everywhere 绝不能成为一个让客户挑肥拣瘦、选择性开启的附加项；只要接入，就必须对所有 Epic 客户全面互通，哪怕对方是位于同一街区、针锋相对的本地竞争医院。在早期推广该功能时，一家合作医院系统在不知不觉中签署了包含该互操作性条款的新版软件协议。该医院系统的 CEO 后来公开坦言，如果当时在合同中注意到了这项互通条款，他绝对会断然拒绝签字。Judy 将这次侥幸的突围归结为“纯粹的运气”。以此为突破口，Epic 顺势将该功能列为全体客户的强制性标配，并对旧版本系统进行了全面改造，使全美每一家使用 Epic 的医疗机构都能无缝共享数据。这成为了 Epic 面对外界垄断批评时的王牌公关筹码——他们可以理直气壮地宣称自己是全美规模最大的病历数据共享枢纽。然而，在整个行业范围内，跨系统的互操作性依然被医院管理者讽刺为“滑稽的天方夜谭”。一位医院首席信息官（CIO）曾用生动的类比道出了背后的资本现实：“虽然医疗数据关乎患者福祉，大家情感上觉得必须无条件共享，但如果换成航空业，联合航空会仅仅因为‘对旅客体验友好’而主动把客户数据拱手分享给达美航空吗？当然绝不可能。那我凭什么要高高兴兴地把自己病人的数据送给街对面的竞争对手医院？”这种矛盾恰恰揭示了医疗体制的撕裂感：它一方面承载着救死扶伤、关乎每个人切身健康的崇高社会职能，另一方面本质上却是一个高度营利性、受制于资本回报率与严格监管的商业战场。 - **病历交换体量**：Care Everywhere 当前每日承载的患者数据交互量达 2,000 万条。 - **强制互通策略**：利用早期一位医院 CEO 疏忽签字的契机，Epic 将原本可能被各大医院抵制的互操作功能彻底固化为全生态的强制性底层标准。 --- ### [02:12:24](https://www.youtube.com/watch?v=FNRU9ZNqcls&t=7944s) 2008年金融海啸与“即开工”的电子病历刺激计划时间推进到2008年，席卷全球的次贷危机全面爆发。为了避免整个国家乃至全球经济彻底停摆，美国联邦政府启动了前所未有的救市与经济刺激方案。货币政策层面，美联储迅速将基准利率降至零并开动印钞机；但在财政政策层面，光靠降息还远远不够，政府急需通过财政手段将天量资金像“直升机撒钱”一样注入实体经济。向全体国民直接邮寄支票固然是一种办法（正如后来的新冠疫情期间所做的那样），但更具生产力、能够传承自罗斯福新政时期“民间资源保护队”（Civilian Conservation Corps）的成熟做法，是寻找全社会公认急需推进的“即开工项目”（shovel-ready projects）。政府的核心逻辑在于达成“双重红利”：一方面通过大举拨款向市场注入流动性、创造就业岗位并带动经济飞轮重新运转；另一方面，顺势将长期以来全美高度共识却因成本高昂而搁置的大型公共工程一举攻克。电子医疗病历（EMR/EHR）在此刻进入了决策层的视野，成为了堪称完美的“即开工项目”。政策制定者决定以此为契机，通过立法给予医疗机构极其丰厚的真金白银补贴，用真金白银的财政杠杆彻底撬动全美医疗体系的全面信息化。这项宏大举措被确立为三大核心目标： 1. **财政刺激**：在金融危机后向全美经济实体注入强心针并稳定就业。 2. **推动普及**：通过巨额补贴激励医疗机构彻底告别纸质病历，采购并上线电子健康档案系统。 3. **推动有效使用（Meaningful Use）**：强调不仅仅是花钱采购系统装装样子，而是必须在临床和管理实践中真正“有意义地”深度使用。这一关键的历史拐点，为日后彻底改写美国医疗软件产业格局、将 Epic 推向行业霸主地位埋下了决定性的伏笔。 ### [02:14:28](https://www.youtube.com/watch?v=FNRU9ZNqcls&t=8068s) 《高科技法案》与 360 亿美元的“胡萝卜与大棒” 2009 年，奥巴马政府签署的《美国复苏与再投资法案》（ARRA）中，包含了一部对美国医疗信息化产生深远影响的子法案——《经济与临床健康信息技术法案》（Health Information Technology for Economic and Clinical Health Act，简称 HITECH Act，《高科技法案》）。该法案的核心目标不仅仅是让医院和医生把系统买回去闲置，而是要求他们产生“有意义的使用”（Meaningful Use，业内常简称为 MU），并且推动建立互操作标准，以便全国范围内的医疗机构能够共享患者数据。这笔公共支出的规模极其惊人：法案最初直接拨出 270 亿美元作为奖励资金，向实施电子病历（EMR/EHR）的医院发放补贴；如果算上其他医疗 IT 项目、数据交换平台建设以及医护培训等综合激励，总注资金额高达 360 亿美元。两位主持人打趣道，对于一家软件初创公司来说，这简直是梦幻般的天降横财——政府不仅替你的潜在客户付软件采购费，还会因为客户持续、深入地使用你的软件而继续给他们发钱。这就好比政府给每家医院发了一张专用信用卡，限定这笔钱只能用来买软件，并且还附带了苛刻的约束条件：必须真正用起来；如果 5 年过渡期结束后医院仍未使用符合标准的系统，政府非但不再发钱，还会通过医保报销扣减等手段施加严厉的经济惩罚。 ```mermaid flowchart TD accTitle: HITECH Act 激励与惩罚机制 accDescr: 展示美国高科技法案如何通过前期的资金奖励与后期的罚款倒逼医院采纳电子病历系统。 subgraph Policy["政策驱动机制 (HITECH Act)"] A["360 亿美元总资金池"] --> B["前5年胡萝卜：激励补贴 每位医生 4.4万 - 6.4万美元"] A --> C["5年后大棒：财务惩罚 未达标削减 Medicare 报销"] end subgraph Hospital["医院端决策"] B --> D["消除预算阻力 倾向选择最昂贵、最可靠的系统"] C --> D D --> E["全美渗透率狂飙 2009年 9% ➔ 2014年 95%"] end subgraph Epic["Epic 的竞争优势"] D --> F["零风险偏好强化 Epic 地位 （爱马仕铂金包效应）"] E --> G["系统深度绑定与壁垒成型"] end ``` 具体落实到账面上，医院每推动一名在岗医生有效使用 EHR 系统，就可以在数年内累计申领 44,000 至 64,000 美元的奖励。对于像 Kaiser Permanente 这样拥有约 11,000 名医生的庞大医疗集团而言，这意味着数以亿计的巨额资金。因此，“Meaningful Use”这个词在整个医疗行业立刻成了一个能触发所有人神经的敏感暗号。 - 2009 年通过的《高科技法案》（HITECH Act）被纳入近万亿美元规模的《美国复苏与再投资法案》（ARRA）。 - 政策包含 270 亿美元的直接使用奖励，广义医疗 IT 项目总投资达到 360 亿美元。 - 医院按医生人头获得补贴，每位医生可折合 4.4 万至 6.4 万美元奖励。 - 前 5 年以资金激励为主（胡萝卜），刺激期结束后对未达标机构施加实质性罚款与报销核减（大棒）。 --- ### [02:17:00](https://www.youtube.com/watch?v=FNRU9ZNqcls&t=8220s) 繁重的合规考核与“免费的爱马仕铂金包” 非营利媒体 Kaiser Health News（KFF）在深度报道《点鼠标累死的医疗》（*Death by a Thousand Clicks*）中生动刻画了当时的行业状态：当时整个 EHR 行业还只是个规模约 20 亿美元的草根市场，面对政府列出的连篇累牍的严苛功能清单，软件供应商们虽然怨声载道，但面对 360 亿美元的政策红利诱惑，没有任何人愿意掉队。正如 Epic 的主要竞争对手之一 NextGen Healthcare 时任 CEO 拉斯蒂·弗朗茨（Rusty Frantz）所言：“整个行业的心态就是，一张巨额支票就挂在我眼前晃来晃去，我只要把表格上的每一项勾选上就能拿到，那我当然拼了命也要去勾选。” 对于 Epic Systems 来说，这固然是天大的行业红利，但为了迎合各项考核指标，公司内部也不得不抽调大批研发资源去应对各种繁琐的流程与合规标准。然而，当全美所有医院都在政府资金驱动下必须采购系统时，市场的竞争法则发生了根本性倾斜： 1. **可靠性压倒创新性**：医院为了拿到补贴，绝对不能承担软件实施失败或中途崩盘的风险，因此创新与否不再关键，按时上线并不出故障成了唯一硬指标。 2. **“铂金包”效应**：以前医院不买 Epic 是因为其部署成本极其昂贵；如今政府全额补贴甚至倒贴，价格不再是阻碍。用主持人的生动比喻来说，这就像政府给你发购物券让你去买包，货架上有普通的 Target 帆布包，也有昂贵的爱马仕铂金包（Birkin bag），在两者都能报销的情况下，所有人都会毫不犹豫地挑选最贵、声誉最高的那一个。Epic 凭借其在交付成功率和高净值客户中积累的极高声誉，成为了这场资金盛宴中最大的赢家。 - EHR 软件行业在法案出台前年产值仅约 20 亿美元，面对 360 亿美元的政府资金注入迅速全员跟进。 - 行业考核指标过于细碎，迫使所有供应商重构产品路线图与开发优先级以满足验收条件。 - 医院的避险心态直接转化为对市场龙头 Epic 的青睐，高定价反而不再构成竞争劣势。 --- ### [02:19:20](https://www.youtube.com/watch?v=FNRU9ZNqcls&t=8360s) 普及率奇迹、患者红利与“过度诊疗”争议《高科技法案》最初设定的三大愿景，在落地后交出了截然不同的答卷。首先在**普及率与数字化**方面，法案取得了堪称奇迹的成果：全美医院的电子病历渗透率从 2009 年的区区 9%，在短短 5 年内飙升至 2014 年的 95%。这种在全行业内从几乎为零到全民普及的剧变，放眼整个科技史，大概只有新冠疫情暴发初期的 Zoom 等远程办公软件能够与之媲美。对普通患者而言，这带来了颠覆性的体验飞跃——通过 MyChart 等应用，患者第一次能够直接给医生在线发消息、随时查阅化验报告、自主预约挂号并管理整个家庭的就医记录。但在**降低医疗成本**这一目标上，实际效果却极具争议。虽然有研究声称引入 EHR 让医院运行成本下降了约 10%，但更多行业批评指出，软件系统反而助长了不必要的过度检查与“向上编码”（Upcoding，即套用报销额度更高的诊疗代码）： - **检查项目泛滥**：诊室内的医生在使用电脑系统时，勾选附加检查和化验项目的门槛变得极低，导致单次问诊产生的收费项目大幅增加。 - **医院创收机制的激励**：即便多数医院挂着非营利或大学附属机构的名头，但其本质依然是由追求财务表现的管理层和高收入医护群体驱动的实体。《美国展望》（*The American Prospect*）在一篇题为《一场 Epic 式的乌托邦反面》（*An Epic Dystopia*）的报道中披露，有医生反映医院主管会频繁私下施压：“刚才那个问诊我看评级只是 2 级，难道就不能算作 3 级吗？”在计算机系统的辅助下，只要勾选了某些关联动作，就能合规地套用更高赔付标准的保险代码，导致整体医疗账单不降反升。 - 全美医院 EHR 普及率在 2009 至 2014 年间从 9% 飙升至 95%。 - 患者端通过 MyChart 享受到了移动互联时代的数字化便利，包括病例调阅与家属档案协同。 - 成本控制未达预期：EHR 的模板化输入与代码推荐间接诱导了过度开单与提高报销等级的创收冲动。 --- ### [02:22:25](https://www.youtube.com/watch?v=FNRU9ZNqcls&t=8545s) 搁浅的互操作性与朱迪的顾问角色争议《高科技法案》最未能实现的承诺是**数据互操作性（Interoperability）**。法案条文对“有意义的使用”规定得极其严格且有直接罚款挂钩，但对各系统间的数据共通标准却语焉不详，更没有设立任何直接挂钩的经济惩罚与激励机制。资本与机构永远逐利而行，既然没有考核硬指标，全行业便普遍缺乏打破壁垒、开放数据接口的动力。这种情况在客观上对 Epic 极为有利。Epic 走的是自主研发、全自研模块紧密耦合的单一架构路线，它根本不需要与第三方软件系统打通就能在一家医院甚至一个大型医联体内实现完整运转；相反，一个相互孤立的封闭生态反而筑牢了 Epic 的竞争护城河，让已经采纳它的医院客户更难迁出。这一结果也让 Epic 及其创始人朱迪·福克纳（Judy Faulkner）面临长期的利益冲突质疑——在法案制定与推行期间，朱迪本人始终担任奥巴马政府健康信息技术政策委员会（Health IT Policy Committee）的顾问成员，直接参与了法案落地细节的建言献策。不过节目主持人也明确指出，指责 Epic 单方面操纵政策是有失公允的，因为在那个委员会以及相关政府决策机构中，Cerner、Allscripts 等主要竞争对手的高管同样位列其中，各方都在试图影响顶层规则的走向。 - 法案对数据接口与跨厂商互操作缺乏实质性激励或惩罚措施，导致跨院数据自由流动的构想基本落空。 - Epic 的一体化单体架构在缺乏外部兼容要求的环境中获得竞争优势，强化了其对医院客户的绑定深度。 - 朱迪·福克纳在奥巴马政府的健康 IT 政策委员会任职引发过政商利益纠葛的争议，但实际上 Cerner 等主要竞品高管同样参与了该委员会的工作。 ### [02:23:36](https://www.youtube.com/watch?v=FNRU9ZNqcls&t=8616s) 规则扭曲市场：“有意义使用”带来的两大系统性恶果在回顾《高技术法案》（HITECH Act）与“有意义使用”（Meaningful Use）政策时，不可回避的争议是数据互操作性（Interoperability）并未如预期般迅速实现。虽然在十五年后的今天互操作性有了显著进展，但事实证明，“没有系统的可操作性（Operability），就根本谈不上互操作性”。政府介入并扭曲市场、直接监管产品研发与使用流程的做法，不可避免地引发了混乱。国会不仅在法案中界定了何为“有意义使用”，实质上还越俎代庖，替科技行业设计了软件功能规格，更间接告诉了医生“应该如何看病”。这一干预带来了两个灾难性的副产品：首先，软件功能被立法绑架。医生如今为了完成一次诊疗，必须在系统里连续点击九个此前完全不必要的选项。这其中一部分源于医院自身糟糕的实施方案，一部分源于系统本身的极端复杂性，但归根结底，是为了满足法案要求的合规指标，以免医院面临沉重的财政处罚。其次，它永久性地加重了全行业行医的监管负担与运营成本。即便最初的财政补贴消耗殆尽，强制性字段、繁琐表单和僵化流程依然永远留在了医疗系统中。医院优化的目标悄然从“帮助医生和患者获得更好疗效”异化为“在法律层面上精准达标”。这直接推高了医院的经营成本，倒逼各大医疗机构加速并购整合，向医疗体系中注入了大量无谓的内耗与浪费。 ```mermaid flowchart TD accTitle: HITECH法案与有意义使用的连锁反应 accDescr: 展示国会立法如何通过行政合规指标扭曲医疗软件并加剧系统内耗 subgraph Legislation["国会立法介入"] A["HITECH 法案出台"] --> B["硬性定义'有意义使用' (Meaningful Use)"] end subgraph Impacts["双重负面影响"] B --> C["产品功能被立法绑定 (强制点击与合规表单)"] B --> D["全行业合规成本剧增 (未达标将面临财政惩罚)"] end subgraph Consequences["医疗系统异化"] C --> E["医生陷入无休止的数据录入 诊疗时间被大幅挤占"] D --> F["医院运营开销暴涨 倒逼医疗机构加速兼并重组"] end ``` ### [02:26:10](https://www.youtube.com/watch?v=FNRU9ZNqcls&t=8770s) 刺激经济还是制造内耗？一场关于“就业工程”的争论尽管合规成本极高，但回顾过去，数字化依然是不可逆转的进步——毕竟谁也不愿意倒退回混乱低效的纸质病历时代。如果跳出医疗行业微观视角、从最宏观的政策目标来审视：这项耗资巨大的政策是否实质上充当了一场国家级的“以工代赈”与经济刺激计划？如果答案是肯定的，那么这项政策拿到了“A+”的成绩单。医疗保健及相关服务目前占据了美国 GDP 的约 18%。因数字化转型而诞生了无数无法被进口取代、无需依赖传统制造业的本土高薪岗位，例如医院的 IT 管理人员、Epic 系统认证管理员等，为大量美国人创造了稳定的职业晋升路径。不过，两位主持人对此观点产生了激烈的交锋：一方认为动用纳税人资金长期支撑某个行业的就业绝非健康常态，“如果纯粹为了刺激经济，直接向全民寄送支票同样能达到效果”；但另一方指出，在金融危机背景下，这项财政刺激至少为全美附赠了一张“全面数字化病历”的兑换券——哪怕它不够完美、互操作性差强人意，也总好过一无所获。 ### [02:28:05](https://www.youtube.com/watch?v=FNRU9ZNqcls&t=8885s) 360 亿美元锁死赛道：Epic 的监管顺风与竞争终结对 Epic 而言，“有意义使用”的推行并非直接的监管俘获（Regulatory Capture），而是一场巨大的“监管顺风”（Regulatory Tailwind）。在赢得凯撒医疗（Kaiser Permanente）的历史性大单后，Epic 本就展现出横扫市场的统治势头。然而，联邦政府砸下的 360 亿美元补贴相当于按下了快进键，将未来十几年才会逐步发生的采购需求，强行压缩到了同一个时间窗口内。这一做法产生了深远的行业后果： - **竞争大门被彻底锁死**：医院面临补贴申领的紧迫倒计时，只能选择当时市场上最现成、能立即部署的成熟方案（Shovel-ready）。这直接终结了未来潜在创新型竞争对手入场的可能，因为医疗系统的替换成本极高，一旦选定就会绑定数十年之久。 - **产品被强行复杂化**：Epic 自身团队也未必乐见这种推行方式。原本按照自身节奏自然占领市场的路径，被迫变成了缝合政策条款的“弗兰肯斯坦式怪物”。每次患者就诊，医生必须在屏幕上勾选多达 57 个合规选项。软件的易用性在层层行政法规的重压下遭到了严重破坏。 ### [02:30:49](https://www.youtube.com/watch?v=FNRU9ZNqcls&t=9049s) 数字化刚性与医护倦怠：被剥夺的弹性与消失的时间立法主导产品开发的恶果，最终完全转嫁到了临床医生身上。海量的交互界面、复杂的录入要求和系统消息不仅拖垮了软件体验，更大量吞噬了医生直接面向患者的诊疗时间。一项 2016 年的权威研究表明，医生每为患者提供 1 小时的面对面诊疗，就必须在电子病历（EHR）系统中耗费整整 2 小时进行数据录入。许多医生被迫日均工作 11 至 12 个小时，在下班后还要持续处理系统收件箱中堆积如山的信息。除了时间被挤占，数字化还带来了一种隐藏的负面效应——剥夺了医院原有的“制度弹性”（Slack）： - **过往的非正式分工**：在纸质时代，虽然规章要求由主治医生亲自下达药房处方、检查和扫描医嘱，但面对每日 20 位以上的接诊量，医生往往口头交代医疗助手（MA）代为录入，自己事后集中签字确认，系统在默契中保持高效运转。 - **系统痕迹与监管刚性**：全面数字化后，系统对每一次点击和权限留下了不可磨灭的审计日志。规章制度被迫以最严苛的形式执行，医生必须事必躬亲地逐条下达医嘱。 - **职业倦怠的根源**：助手无法分担日常琐事，原有的缓冲空间荡然无存，成为近年来医生职业倦怠（Doctor Burnout）爆发的关键诱因。不过，历史数据也提醒人们不能美化过去：早在 1970 年的一项研究就曾发现，在纯纸质时代，仅仅是管理和调阅实体病历等沟通类事务，就吞噬了医院整体运营成本的 35% 至 39%。 ### [02:32:44](https://www.youtube.com/watch?v=FNRU9ZNqcls&t=9164s) 数字化与数字化转型：奥巴马与医疗行业的幻灭尽管从纸质病历转向电子病历在很大程度上是将繁重的人力成本转嫁并重新分配，但由于《有意义使用》（Meaningful Use）带来的繁冗监管条文，医护人员的工作负担反而不减反增。甚至连当年强力推行这项政策的当事人也深感失望。2017 年，奥巴马在接受 Vox 采访时坦言，《高科技法案》（HITECH Act）远未达到他的预期："事实上，依然有堆积如山的文书工作，医生们依然不得不手动录入大量内容，护士们的所有时间也都被这些行政事务填满了。我们砸进了一大笔钱，试图鼓励所有人实现数字化以赶上世界其他国家的步伐，但这比我们想象的要艰难得多。" 对这番反思，Ben 和 David 指出，美国在医疗数字化领域原本就处于领先地位，并不存在“赶超世界”的问题；当时大家真正奢望的，是通过数字化大幅削减开支，使美国的医疗成本降低到与其他发达国家相当的水平，但这一愿景彻底落空了。一位接受调研的医院首席信息官（CIO）对此作出了极为精辟的总结：Meaningful Use 和 HITECH 法案在“行业的数字化（digitization）”上取得了空前狂热的成功，但在“行业的数字化转型（digital transformation）”上却一事无成。不过，将一切全面转化为数字数据并非毫无价值，它至少为下一阶段铺平了道路——在统一的数据底座之上，如今兴起的人工智能与环境感知 AI（ambient AI）终于有望接管这些令医护人员苦不堪言的录入重担。 ### [02:34:28](https://www.youtube.com/watch?v=FNRU9ZNqcls&t=9268s) Epic 营收破十亿，对手却深陷并购泥潭与系统分裂在全行业急速数字化的浪潮中，Epic 展现出了无与伦比的统治力。2011 年，Epic 年营收正式突破 10 亿美元大关，并在全美各大顶级医疗系统的公开招标（RFP）中攻城略地，接连斩获约翰斯·霍普金斯（Johns Hopkins）、西达赛奈（Cedars-Sinai）以及加州大学旧金山分校医疗中心（UCSF）等标杆大单。与此同时，面对 Meaningful Use 的合规压力，Epic 的一众竞争对手为了迅速凑齐产品资质，纷纷开启了急功近利的并购整合狂潮。然而，这种拼盘式的吞并不仅没有削弱 Epic，反而进一步放大了 Epic 单一代码库套件的架构优势。在对手阵营中，频繁的拼凑造成了系统难以弥合的技术裂痕： - 2008 年，Allscripts 与 Misys 换股合并；2010 年，重组后的 Allscripts 又吞并了 Eclipsys，并在随后持续吃进中小型医疗软件商（该公司如今更名为 Veradigm）。 - 2011 年，MEDITECH 出资收购了长期合作伙伴 LSS Data Systems。 - 2014 年 8 月，行业巨头 Cerner 斥资 13 亿美元收购西门子医疗 IT 业务（Siemens Health Services）。正是收购西门子这一决定，将 Cerner 推向了下坡路的开端。即便到了 2025 年，大量老客户在心理和系统层面上依然清晰地将自己划分为“西门子客户”或“Cerner 客户”，双方底层架构始终未能深度融合。后来甲骨文（Oracle）以溢价收购 Cerner 后试图对其进行全盘重写，同样面临着进展艰难的技术沼泽。 ```mermaid flowchart TD accTitle: Competitor Consolidation vs Epic Architecture accDescr: 竞争对手不断并购造成代码撕裂，反衬出 Epic 单一代码库的巨大整合优势。 subgraph Competitors ["竞争对手：拼凑式并购陷阱"] A1["Allscripts + Misys + Eclipsys"] --> A2["多重并购集成困难 （现更名为 Veradigm）"] B1["Cerner 13亿美元收购 Siemens Health Services (2014)"] --> B2["系统长期分裂、重写泥潭 （客户至2025年仍无法统一）"] end subgraph EpicStrategy ["Epic：单一架构战略"] C1["自研一体化平台 单一代码库 (Single Database)"] --> C2["连克大单：Johns Hopkins / Cedars-Sinai / UCSF 2011年营收突破10亿美元"] end ``` ### [02:36:27](https://www.youtube.com/watch?v=FNRU9ZNqcls&t=9387s) 节目赞助插播：全栈 AI 算力基建先锋 Crusoe 在切入更加沉重的政企外包话题前，节目迎来了老朋友 Crusoe 的赞助插播。Crusoe 并非单纯的 AI 云服务商，而是一家从物理底层重新构想 AI 基础设施的垂直一体化公司。不同于传统数据中心，Crusoe 早期开创性地利用偏远地区的闲置与弃用能源（stranded energy）直接驱动 GPU 集群，随后更是将能力拓展为从能源获取、物理数据中心建设一直到顶层 AI 云服务（AI Cloud）的全栈垂直整合能力。为了说明其恐怖的建设速度与规模，节目列举了一组震撼数据： - **阿比林超大型 AI 工厂**：Crusoe 正在德克萨斯州阿比林（Abilene, Texas）兴建一座装机容量高达 1.2 吉瓦（GW）的巨型 AI 数据中心。该项目于 2024 年 6 月在荒地上破土动工，不到一年时间便开始陆续点亮交付首批建筑，施工速度远超传统行业周期。 - **体量对比参照**：承载了全球约 70% 互联网流量的北弗吉尼亚州（Northern Virginia）数据中心走廊，其全部算力集群总容量目前约为 4.5 吉瓦。这意味着仅阿比林这一个 Crusoe 基地，规模就超过了北弗吉尼亚总容量的四分之一。 - **最新开发者服务**：Crusoe 近期推出了两项托管级新功能——解放服务器运维烦恼的 Managed Inference，以及消除分布式大模型训练繁琐调配的 AutoCluster。Ben 与 David 亦透露，他们已与英伟达（Nvidia）、Founders Fund 等机构共同参与了 Crusoe 的新一轮融资。 ### [02:39:00](https://www.youtube.com/watch?v=FNRU9ZNqcls&t=9540s) 国防部与退伍军人事务部天价合同：纳税人的灾难话题转回 2015 年爆发的“美国国防部（DoD）招标惨案”。美国国防部决定统一全军各军种医疗系统的电子健康档案，启动面向现役军人全球医疗机构的 EHR 全球招标。这项 2015 年的合同总标的额高达 43 亿美元，打破了早前 Kaiser 创下的历史纪录，成为当时史上最大的医疗 IT 单项采购案；而在两年后，面向退伍军人事务部（VA）的医疗合同更是飙升到了惊人的 100 亿美元。对于医疗 IT 厂商而言，这是维持几十年饭碗的无上肥肉；但对于纳税人来说，它却成了美国政府公帑浪费的象征。受限于美国联邦采购复杂的制度壁垒，软件商无法直接向政府独立竞标，必须依附于拥有成熟军工资质的主承包商组队参与。Epic 与 Cerner 各自挑选盟友激烈博杀，最终 Cerner 作为软件分包商大获全胜，不可思议地同时吞下了国防部（DoD）与退伍军人事务部（VA）两张百亿级超级大单。然而，这两项巨额项目随即陷入了灾难性的超支与严重延期： - **国防部（DoD）项目**：2015 年正式发标，整套面向现役部队的系统一直拖延到 2024 年底才基本实现完全上线（Go-Live），实施交付周期长达 9 年之久。 - **退伍军人事务部（VA）项目**：惨烈程度远超国防部。该项目自 2017 年发标，截至目前仅在极少数试点医院勉强运转。根据 VA 官方给出的最新预测，全美所有退伍军人医疗站点全面上线的时点"最早也要等到 2031 年"（as early as 2031）。即便按这一近乎不可能达成的乐观预期推算，整个工程交付周期也高达 14 年。面对这起离谱的预算黑洞，业界无奈感叹起政企采购中残酷的逆向激励潜规则："在政府外包合同里，你在失败项目上能捞到的钱，往往比在成功项目上赚得还多得多。"这种按工时与变更单不断加码预算的结构，彻底将利国利民的医疗系统改造演变成了消耗公共财政的无底深渊。 ### [02:42:32](https://www.youtube.com/watch?v=FNRU9ZNqcls&t=9752s) 政府大单的泥潭：Cerner 的梦魇与 Epic 的“因祸得福” 主持人深入剖析了美国国防部（DoD）与退伍军人事务部（VA）大型政府采购案背后的扭曲激励机制。在政府合同体系中，只要项目实施没有彻底终结，总承包商与分包商就能不断发现新的成本并持续获得付款。将这一烂摊子完全归咎于 Cerner 是有失公允的——正如在讨论洛克希德·马丁（Lockheed Martin）那一期节目中所指出的，Cerner 只是大型政府总承包商（GC / Primes）底下的分包商，这是典型军工复合体体系的运作方式。多层官僚主义的重叠——联邦政府、军方以及本就错综复杂的 EMR（电子病历）系统——直接构成了一团官僚泥潭。数十年过去，数百亿美元的纳税人资金被源源不断地砸入其中，系统却迟迟无法完全上线，局面堪称惨烈。面对“Epic 是否庆幸当年没有拿下这个大单”的疑问，主持人指出 Epic 当年确实提交了 RFP 竞标，且当时肯定极度渴望中标——毕竟合同最初的 100 亿美元标价（尚未计入后续超支）已是如今 Epic 全年收入的两倍，这本可能像当年 Kaiser Permanente 项目一样，彻底重塑一家公司的命运。然而，在与 Epic 的现有客户和医疗集团 CIO 交流时，所有人都在跪谢 Epic 当年“痛失”这笔大单。正因为没有中标，Epic 才得以避开将 Cerner 彻底拽入泥潭的无尽消耗战。 ```mermaid flowchart TD accTitle: DoD/VA 合同对 Cerner 与 Epic 的分流影响 accDescr: 对比 Cerner 陷入政府合同泥潭与 Epic 专注于核心客户的两种不同走向 subgraph Cerner["Cerner 泥潭"] A["赢得 DoD/VA 庞大订单"] --> B["陷入多层官僚与项目超支泥潭"] B --> C["创始人去世及并购整合包袱"] C --> D["营收停滞 / 2021年被 Oracle 收购"] end subgraph Epic["Epic 专注策略"] E["错失 DoD/VA 竞标"] --> F["避开政府巨额官僚损耗"] F --> G["保持对核心代码与客户的极致专注"] G --> H["横扫顶级学术医院与国际市场"] end ``` ### [02:44:06](https://www.youtube.com/watch?v=FNRU9ZNqcls&t=9846s) Cerner 的动荡与 Epic 对顶级医疗机构的横扫与此同时，Cerner 内部正面临一系列复合危机。在收购西门子医疗 IT 业务后，Cerner 本身就是由多达 36 家被并购公司拼凑起来的庞然大物。恰在此关键时刻，其联合创始人兼长期领袖尼尔·帕特森（Neal Patterson）不幸罹患癌症，并于 2017 年英年早逝。在极其复杂的国防部项目推进中失去灵魂人物，导致 Cerner 在随后的几年内高频更替领导层。反观 Epic，彻底从军方和 VA 的混乱局面中解脱出来，凭借创始人朱迪·福克纳（Judy Faulkner）高度克制、坚守原则的管理哲学，赢下了一个又一个大型医疗系统客户： - **波士顿 Partners HealthCare**（现 Mass General Brigham，整合了哈佛大学医学院与麻省总医院 MGH 等顶级研究机构）； - **梅奥诊所（Mayo Clinic）**旗下的所有医院，成为除国防部项目之外规模最大的民用医疗合同之一； - **英国剑桥大学医院**（Cambridge University Hospitals NHS Foundation Trust）； - **Intermountain Health**； - 近期攻下的 **CommonSpirit Health**。到 2018 年，Epic 的年营业额达到 27 亿美元，并正式宣布包揽了《美国新闻与世界报道》（U.S. News & World Report）全美排名前 20 的所有顶级学术医院。这一里程碑极具战略意义：医学院与教学医院虽可能采用分开购买的教育与临床软件许可，但在这些顶尖机构受训的医护人员直接在临床中使用 Epic。如今全美超过 90% 的新毕业医学生在日常训练中使用的均是 Epic，为其建立了不可动摇的用户习惯与人才护城河。 ### [02:46:34](https://www.youtube.com/watch?v=FNRU9ZNqcls&t=9994s) Oracle 收购 Cerner 与 Judy 的长期主义胜利 2019 年，Epic 营收跃升至 3.2 亿美元。到了 2021 年 12 月，身陷困境的 Cerner 最终以 280 亿美元被甲骨文（Oracle）全盘收购。Cerner 虽依然是一艘拥有国际业务与政府大单、年收入约 55 亿美元的巨轮，但自 2018 年起其营收就已经陷入停滞甚至下滑。在被收购后，Oracle Health 频繁进行裁员，甚至不再单独披露 Cerner 的独立财务数据，反而公开称其为拖累公司整体增长与利润率的“逆风因素”（headwind）。分析师普遍认为，这笔交易已成为 Oracle 历史上表现最糟糕的收购案之一。这种行业惨状反而彻底印证了 Judy Faulkner 创立 Epic 时所坚持的信条： - 打造一套统一的代码与底层平台架构，而非拼凑并购而来的废铁； - 对软件工程与产品架构保持硬核要求，100% 聚焦于客户交付； - 法律层面上绝不被收购、绝不上市、绝不发起外部收购。 Epic 在前二十年的发展看似平淡无奇，甚至略显迟缓，但正是这种拒绝走捷径、拒绝人为拔苗助长的“笨功夫”，为它在二三十年后积累了坚不可摧的爆发力。关于《Meaningful Use》（有意义使用）法案对公司的实际助益，Epic 内部一直持有极为复杂的态度。尽管政策确实催化了普及速度，但从历史曲线来看，Epic 本就凭借更优异的产品与客户服务走在全面胜出的轨道上；其文化深处天然反感这种人造的增长激素，宁愿选择以“极慢的姿态走向胜利”，因为只有依靠内生力量慢赢，最终才能形成压倒性的持久统领地位。 ### [02:50:08](https://www.youtube.com/watch?v=FNRU9ZNqcls&t=10208s) 国际扩张与医疗并购潮带来的隐秘东风如今的 Epic 迎来了前所未有的广阔前景，海外市场业务已逐步扩大，贡献了约 10% 至 15% 的营业收入。2023 年，Epic 在英国伦敦的盖伊和圣托马斯国家医疗信托（Guy's and St Thomas' NHS Foundation Trust）成功上线，创下其单次实施规模的历史之最，英国正迅速成为 Epic 的核心国际重镇。与此同时，美国本土医疗市场的结构性转变更为 Epic 带来了强劲且特殊的顺风——大型医疗集团对地方诊所和小医疗机构的疯狂并购与整合（Hospital Roll-ups）： - 《平价医疗法案》（Affordable Care Act）以及《高科技法案》（HITECH Act）大幅收紧了对违反《健康保险可携性与责任法案》（HIPAA）的惩罚机制； - 医疗合规与行政负担急剧膨胀，小型私人诊所或社区诊所根本无法单独承担昂贵的合规与技术开销，被迫选择并入超大型医疗网络； - 整个医疗系统的话语权向巨型集团集中，而每次并购后，巨头往往会借重新招标的机会将原有零散软件替换为同一套核心系统； - Epic 三十年前就认准的策略——不碰小型分散客户，只专注拿下最庞大、最复杂的顶级医疗卫生系统——在当下被证明具有惊人的前瞻性，因为如今市场上真正具备雄厚财力的大型存活者，已全数变为 Epic 的原生根据地。 ### [02:52:03](https://www.youtube.com/watch?v=FNRU9ZNqcls&t=10323s) 医疗兼并潮与医院的“规模生存战” 过去数年间，随着美国私营医疗保险公司在整个医疗生态中积累了极大的议价权，医院为了在艰难的谈判中生存下来，也被迫走向高度整合，形成了体量巨大的区域性“类垄断”医疗集团。两位主播坦言，在录制本期节目之前，原本以为医院是坐收暴利的“反派角色”，但深入研究后才发现，医院实际上几乎不怎么赚钱，大部分机构都在盈亏平衡线上苦苦挣扎，这也正是它们频频抱团合并的核心诱因。对于极少数经营优秀的医院而言，赚到的利润也必须全部重新投入扩张——修建新院区、兼并周边诊所。在这个时代，单体医院如果不把体量做大，根本无法与巨无霸级别的保险公司及大型供应链巨头坐在同一张桌子上博弈。这种全行业对“规模”的疯狂追逐，反过来为 Epic 打开了崭新的增长空间。当 Epic 在医疗提供方（Provider，即医院与诊所）市场建立起不可动摇的庞大基数后，它开始顺理成章地将目光投向医疗生态中更有钱的另外两个超级巨头：支付方（Payers，即商业医疗保险公司）与大型药企（Biotech & Pharma）。 ```mermaid flowchart LR accTitle: Epic 从医院扩展至保险与药企的飞轮 accDescr: 展示医院因与保险公司博弈被迫兼并规模化，促使 Epic 垄断医院端后向支付方与药企横向渗透的路径。 subgraph Consolidation ["医疗生态博弈"] PayerPower["保险巨头议价权膨胀"] --> HospitalMerge["医院被迫抱团整合 / 追求规模"] end subgraph EpicExpansion ["Epic 扩张飞轮"] EpicHospital["统治级渗透医院 IT 系统 (单一供应商策略)"] --> EpicDataGrid["汇聚全美核心临床与就诊数据"] EpicDataGrid --> PayerValue["赋能保险方: 自动化预先授权"] EpicDataGrid --> PharmaValue["赋能药企: 临床试验与罕见病研发"] end HospitalMerge --> EpicHospital ``` --- ### [02:53:01](https://www.youtube.com/watch?v=FNRU9ZNqcls&t=10381s) 扩张新版图：连接网格与预先授权的自动化 Epic 将这一全新战略愿景定义为“互联系统”（System of Connectedness），有时也被内部称为“医疗网格”（Grid of Care）。其核心逻辑在于：利用其在全美医院底层系统所掌控的全面数据，向医疗保险公司、跨国药企、居家护理及康复机构输出高价值软件产品。其中，最具代表性的攻坚场景便是全美医疗体系中人尽皆知的痛点——**预先授权（Prior Authorizations）**。在传统医疗流程中，当医生认为患者需要某种昂贵药物或特殊手术时，保险公司往往不会直接赔付，而是要求医院在事前逐案提交极繁琐的申请并进行资质审查。这一过程耗费医生大量无谓文书时间，延误患者救治，甚至连保险公司自身的审核部门也苦不堪言。而 Epic 手握完整的临床诊疗与病历记录，便能直接找上保险公司：“我们的客户（医院）讨厌它，你们审核起来也痛苦，而我们拥有全部底层病历数据，完全可以利用算法和结构化数据把预先授权流程自动化。” 这一举措完美展现了 Epic 标志性的商业飞轮与产品扩张战术： - **从单一核心做起**：最初只卖医院核心的电子病历（EMR/EHR）系统； - **观察并吞并周边**：进驻医院后，密切关注医院还在采购哪些其他第三方的 IT 工具（排班、计费、影像、耗材等）； - **打包进企业协议（Enterprise Agreement）**：凭借自身统一数据底座的技术壁垒（或仅仅凭借“全家桶单一供应商”的采购便利性），自研出同类功能，直接捆绑销售，逐步淘汰外部竞争对手，成为医院唯一的“单一 IT 供应商”（Single IT Vendor）； - **横向跨圈变现**：当医院端渗透见顶后，转而将手握的数据与网络效应变现，打包卖给保险公司与大型药企。 --- ### [02:55:38](https://www.youtube.com/watch?v=FNRU9ZNqcls&t=10538s) 大数据底座 Cosmos：2.95 亿患者的数据乌托邦在纵向打通产业上下游的同时，Epic 几年前推出了另一项极具前瞻性的产品——**Cosmos**。长久以来，患者的电子病历数据分散在各个医院本地机房、自建私有云、AWS 或 Azure 上，所有权属于医院和患者本身。Epic 做了一件技术创举：将全美各个 EpicCare 实例中高度结构化的临床数据进行全面去标识化（脱敏），汇聚到一个由 Epic 集中托管、支持海量并发查询的超大型数据库中。截至目前，Cosmos 已经汇集了多达 **2.95 亿名患者**的去标识化医疗数据，涵盖了惊人的 **150 亿次独立门诊与住院诊疗记录（Patient Encounters）**。两位主播强调，这真正实现了几十年前布什政府与奥巴马政府在推动医疗数字化改革时设想的“数据乌托邦”： - **医疗质量与患者安全**：数字化不仅仅是让病人在手机上查看化验单，更有据可查地提升了医疗质量。研究表明 45% 的患者认为 EHR 显著提升了护理质量；而在 2023 年，Epic 系统通过用药冲突预警与流程把关，成功在临床上阻止了 6600 万次潜在不良药物相互作用以及 25 万起潜在手术差错。 - **群体宏观流行病学监测**：如果所有病历都是纸质档案，除非有人专门申请科研基金、在一箱箱旧纸堆里手工调档，否则绝不可能发现区域性隐患。就像臭名昭著的密歇根州弗林特水污染事件，正是借助电子化数据库进行横向样本交叉比对才得以被彻底曝光。而 Cosmos 则是把这种能力放大了数万倍。极其聪明的是，Epic 目前对所有自愿贡献数据的客户免费开放 Cosmos。任何加入联盟的医院医生，如果在临床上接诊到极其罕见的怪病，可以直接在 Cosmos 中检索全美历史病例。系统会瞬间匹配出匿名化的历史“相似患者（Lookalikes）”，并展示过往的治疗方案及主治医生信息，甚至能直接赋能新药研发与临床试验招募。这种海量、干净、结构化的高质量真实世界医疗数据，也为未来与 AI 的深度结合埋下了无可比拟的伏笔。 --- ### [03:00:32](https://www.youtube.com/watch?v=FNRU9ZNqcls&t=10832s) Epic 当前业务核心体量盘点主播整理了 Epic 当下的经营指标，呈现出一幅近乎无懈可击的企业软件帝国版图： - **客户基数与黏性**：拥有 607 家大型医疗集团客户，覆盖全美 3,200 家医院。最为逆天的是，在其数十年发展历史中，**客户流失率为 0**（从未流失过任何一家核心医院客户）； - **年新增速率**：每年保持极其稳健的节奏，新增 10 到 25 家大型医疗系统客户； - **终端覆盖规模**：服务全美 590,000 名注册执业医师，覆盖 495,000 张编制病床； - **患者触达面**：全球覆盖 3.25 亿患者，其中美国本土达 2.8 亿人——这意味着绝大多数美国人在一生中的某一时刻，其健康医疗记录都必然流经 Epic 的系统； - **财务表现**：2024 年实现营收 57 亿美元，较上年（49 亿美元）同比增长 16%，过去五年复合年均增长率保持在 13% 左右。主播特别指出一个发人深省的反差：对于一个几乎主宰了全美医疗命脉的基础设施而言，**57 亿美元的年收入其实比预想的要低得多**。若将 57 亿美元简单除以 607 家大客户，平均每家医疗集团每年付给 Epic 的费用仅约 1000 万美元。在企业级软件世界中，对于这样一套掌控整个医院所有挂号、计费、开药、排班与临床的核心操作系统而言，这个定价策略甚至算得上非常克制。 ### [03:01:44](https://www.youtube.com/watch?v=FNRU9ZNqcls&t=10904s) 医疗神经系统与反直觉的“低捕获率” 作为全美各大医院系统的“中枢神经系统”，Epic 几乎承载了医院日常运转的方方面面。部署和维护这套系统的开销极为惊人：一家中型医疗系统仅为了管理 Epic，就需要雇佣上百名全职员工；加上支付给埃森哲（Accenture）、德勤（Deloitte）、IBM、Nordic 等外部咨询公司的实施与优化服务费，以及系统切换停机造成的潜在收入损失，医院的总支出十分庞大。然而令人意外的是，在这巨大的总成本中，Epic 每年实际从单个中型医疗系统拿到的软件费用只有大约 1,000 万美元。这种反差并非源于 Epic 缺乏商业嗅觉，而是一种极为清醒的超长期主义策略。随着其市场地位日益巩固，Epic 理论上拥有巨大的提价空间，但它刻意不把价格压榨到客户承受力的极限。这与杰夫·贝佐斯（Jeff Bezos）早期对 AWS 的战略如出一辙——将使用门槛降到极具吸引力，以至于医院如果选择其他供应商，在商业和运营上会变成一项“既不理性也不负责任的决定（an irrational and irresponsible decision）”。 - **价值捕获与实际成本反差**：一家中型医院系统需配备约 100 名全职技术人员管理 Epic，但 Epic 每年从中抽取的直接软件价值仅约 1,000 万美元。 - **产业规模对比**：Epic 的年营收约为 57 亿美元，而美国联合健康集团（UnitedHealth Group）年营收高达 4,000 亿美元，其产生的 EBITDA（约 350 亿美元）是 Epic 全年总营收的 6 倍以上。Epic 在不占有庞大行业利润池的前提下，成为了掌控整个系统的绝对枢纽。 --- ### [03:05:19](https://www.youtube.com/watch?v=FNRU9ZNqcls&t=11119s) 市场体量、财务指标与真实企业估值从宏观运营数据来看，Epic 目前拥有约 14,000 名员工。虽然按医院机构数量计算其市场份额为 42%，但由于其客户几乎涵盖了全美规模最大、声誉最高、接诊量最多的顶尖医疗中心，其承接的实际诊疗业务比例要高得多；全美 58% 的门诊医生在使用 Epic，全美约 79% 的人口至少通过某种方式间接或直接（如 MyChart）接触过 Epic。根据多方估算，Epic 的 EBITDA 利润率在 30% 至 35% 之间，保守估计对应约 1.7 亿美元的年度 EBITDA。关于 Epic 究竟价值几何，福布斯（Forbes）在 2021 年曾估计朱迪·福克纳（Judy Faulkner）的个人净资产为 76 亿美元，倒推 Epic 的估值仅约 150 亿美元，这在主持人看来是极其荒谬的低估。即使参照甲骨文（Oracle）收购衰退中的塞纳（Cerner）时给出的 30 倍 EBITDA 倍数（或美股软件公司的 9 倍市销率），Epic 的保守估值也至少有 500 亿至 510 亿美元。如果考虑到 Epic 无与伦比的客户黏性和不可替代的护城河，若其进入公开市场，市值完全可以达到 1,000 亿美元级别，比肩 Shopify、ARM、洛克希德·马丁（Lockheed Martin）及星巴克。而在科技巨头的战略并购视角下，仅凭其包含 150 亿次患者交互数据的 Cosmos 医疗数据库和 AI 训练价值，$100B 甚至只是收购底线。 --- ### [03:09:11](https://www.youtube.com/watch?v=FNRU9ZNqcls&t=11351s) 传承蓝图：目的信托与工程师 CEO 铁律现年 81 岁的朱迪·福克纳已经为这家庞大帝国的未来做好了极其严密的制度设计。朱迪持有公司约 50% 的经济权益以及 100% 的投票权。对于其经济权益，朱迪已签署“捐赠誓言”（The Giving Pledge），正逐步将其无投票权股份转入其家族慈善基金会“Roots & Wings”；基金会每年按更新的内部评估价格将股份回售给 Epic，从而持续获取公益资金。 ```mermaid flowchart TD accTitle: Judy Faulkner 继承与目的信托架构 accDescr: 展示朱迪·福克纳所持投票权股份转入目的信托的治理结构及关键章程规定 subgraph Ownership["持股权益分配"] J["朱迪·福克纳 (Judy Faulkner)"] NV["无投票权股份 (约 50% 经济权益)"] V["100% 投票权股份"] F["Roots & Wings 慈善基金会"] PT["目的信托 (Purpose Trust)"] J -->|捐赠| NV -->|回售给公司套现| F J -->|逝世后自动转入| V --> PT end subgraph Governance["目的信托三方管理委员会"] C1["家族代表 (丈夫 Gordon 及 3 名子女)"] C2["内部高管 (5 名 Epic 资深高层)"] C3["核心客户 (3 位大型医疗系统 CEO)"] PT --- C1 PT --- C2 PT --- C3 end subgraph Mandate["铁律章程"] M1["绝对禁止出售或上市"] M2["新任 CEO 必须是 Epic 长期老员工"] M3["新任 CEO 必须是软件工程师背景"] PT --> M1 PT --> M2 PT --> M3 end ``` 而在朱迪逝世后，她所持有的 100% 投票控制权将全部移交给一个“目的信托（Purpose Trust）”。该信托的治理委员会由三大阵营共同组成：第一，其丈夫戈登（Gordon）及三名子女组成的家族代表；第二，Epic 内部的 5 位资深核心高管；第三，3 位来自核心大客户医疗系统的 CEO。信托章程设定了不可篡改的严格条款： 1. **公司永不被出售，且永远不得公开上市**； 2. **未来的继任 CEO 必须同时满足两项硬性要求：必须是 Epic 的长期在职员工，且必须具备软件工程师背景（Software Developer）**。这一要求被主持人高度赞赏为“深得企业经营之道”，正如好市多（Costco）历任 CEO 均从底层业务一线摸爬滚打数十年、耐克（Nike）由球鞋设计出身的马克·帕克（Mark Parker）掌舵时迎来黄金时代一样，牢牢锁定了 Epic 作为一家工程师文化主导的技术公司的根本基因。 ### [03:11:42](https://www.youtube.com/watch?v=FNRU9ZNqcls&t=11502s) 开放生态的演变与进入商业模式分析随着时间推移，第三方与 Epic 的集成方式发生了巨大变化。如今外部开发者如果想接入 Epic，通常需要接受向其分成（Revenue Share）的机制；而在极少数真正的深度战略合作中，甚至需要交出公司的认股权证（Warrants）作为筹码。但退一步讲，无论门槛如何，如今开发者能在 Epic 平台上搭建应用与实现深度集成的空间，已经远比十年前或二十年前广阔得多。伴随这一产品生态的梳理完毕，叙事阶段正式告一段落，节目随之进入招牌分析环节。首先探讨的是来自汉密尔顿·赫尔默（Hamilton Helmer）《7 Powers（七大战略维度）》的商业分析框架，核心问题在于：**究竟是什么核心壁垒，让企业能够长盛不衰地获取显著优于最强竞争对手的超额回报与盈利能力？** 这七种力量分别是： - 反定位（Counter-Positioning） - 规模效应（Scale Economies） - 转换成本（Switching Costs） - 网络效应（Network Economies） - 流程优势（Process Power） - 品牌效应（Branding） - 独占资源（Cornered Resource）在拆解这些力量时，主持人照例将 Epic 的发展划分为早期的“起飞阶段”（Takeoff Phase）与成熟后的“行业垄断期”。 --- ### [03:12:46](https://www.youtube.com/watch?v=FNRU9ZNqcls&t=11566s) 起飞阶段的蓄力：时间作为终极资源回顾 Epic 漫长的前半生，令人意外的是，它在起飞阶段几乎不具备典型的赫尔默七大力量。在很长一段时间里，它根本谈不上真正“起飞”。当时 Epic 最大的产品特征——单一代码库与统一部署架构（Single Platform），在早期并未展现出决定性的竞争优势。彼时绝大多数竞争对手才刚刚起步，软件系统的功能堆叠尚浅，医疗机构对计算机化的接纳程度更是处于蛮荒时期。 Epic 能够在萌芽期的电子病历市场中活下来并不断壮大，主要在于其极其独特的资本和运营节奏： - 依靠销售足够小巧、轻便的产品，维持自负盈亏并保有健康利润。 - 绝大多数创业者受制于风投资本结构的时间限制，被迫在瞬息万变的市场中仓促推出平台让外部开发应用，或者附庸在他人平台上单打独斗； - Epic 则拥有绝大多数创业者根本无法奢求的奢侈品——**时间**。由于没有外部资本逼迫快速变现，朱迪·福克纳（Judy Faulkner）甘愿在漫长的岁月里慢慢打磨系统，熬到行业逐渐成熟，直至客户对集成度高、功能无所不包的超大套件产生井喷式需求。 --- ### [03:14:14](https://www.youtube.com/watch?v=FNRU9ZNqcls&t=11654s) 成熟期的绝对护城河：从天量转换成本到网络效应当讨论如今身为行业霸主的 Epic 拥有哪些护城河时，答案变成了压倒性的“几乎全占”： ```mermaid flowchart TD accTitle: Epic Systems Core Moats accDescr: The four interconnected moats sustaining Epic Systems' dominance in healthcare IT subgraph Moats ["Epic 核心壁垒"] SC["转换成本 医疗命脉，替换风险极高"] SE["规模效应 海量研发分摊与产品捆绑"] NE["双重网络效应 跨院病历打通 + 医生人才标准"] BR["品牌信任 买 IBM 永远不会被开除"] end SC --> NE SE --> BR ``` 1. **不可撼动的转换成本（Switching Costs）：** Epic 堪称“有史以来转换成本最高的软件产品”。它不仅是一套 ERP，更是医疗机构全身神经系统的中枢命脉。如果系统宕机或数据出错，甚至会导致人员丧命。医院实施一套系统往往要消耗数亿乃至数十亿美元的实际费用与机会成本。相较于日常替换 ERP 或 CRM，彻底拔除并更换核心 EMR 的概率几乎为零。 2. **庞大的规模效应（Scale Economies）：** 开发满足全美各专科医院所需的功能广度，需要极其惊人的资金消耗，唯有 Epic 当前的客户体量才能分摊这种研发成本。规模效应还让 Epic 可以随意玩弄“打包赠送”功能模块的策略。如今新入局者如果要重写一套同等规模的系统，前期代码量门槛已是天文数字。 3. **双重网络效应（Network Economies）：** Epic 经历了跨越奇点后的质变： - **患者与数据层面的网络效应：** 通过 Care Everywhere 协议，跨院之间互调病历极其流畅。主持人提到自己在西雅图带孩子看食物过敏时，唯独一家坚持使用 Cerner 的医院让跨院看诊非常痛苦，所有人都巴不得它赶紧换成 Epic。 - **医护人才与劳动力市场的网络效应：** 全美顶尖医学院毕业的医生、护士早在受训时就习惯了 Epic。医院为了招募新医生并应对退休断档，必然优先采购全行业通用的标准系统。 4. **品牌信任（Branding）：** Epic 树立起了顶尖医疗机构最信赖的金字招牌。其品牌地位类似于科技界的古老谚语——“没有人会因为采购了 IBM 的产品而被解雇”。在同台竞标时，只要标的相当，客户就会毫不犹豫地选择 Epic。位于威斯康星州农场中那个能容纳上万人的宏伟地下礼堂，也是 Epic 定期汇聚全行业生态、强化其品牌宗教感的重要舞台。 5. **流程优势（Process Power）：** Epic 大量校招刚毕业的年轻工程师，让他们在由 InterSystems 授权的 MUMPS / Caché 专有开发架构下闭门写代码。这种“Epic Way”深度塑造了团队工作流，同时也构成了极强的防守机制：工程师掌握的技术栈很难平移到使用现代通用语言的竞争对手（如 Cerner），极大地降低了核心技术人员的跳槽与流失率。 --- ### [03:19:31](https://www.youtube.com/watch?v=FNRU9ZNqcls&t=11971s) 商业打法：极端的内研基因与 35% 研发费用率进入商业模式与经营策略（Playbook）探讨，最吸睛的数据莫过于其非同寻常的研发投入比例。世界合伙人公司（Worldly Partners）的阿文·纳瓦拉特南（Arvin Navaratnam）在对 Epic 的长线跟踪报告中指出，坚持系统全部自研、坚决不买任何外部公司的 Epic，将高达 **35% 的运营支出（OpEx）** 投入到了研发中。横向对比该数据在医疗软件与科技巨头间的表现： - **Athenahealth**：研发支出占比仅约 10% - **Oracle（整体）**：研发支出占比约 23% - **Amazon**：研发支出占比约 28% - **Epic Systems**：研发支出占比约 35% - **Apple**：研发支出占比约 36% - **Google**：研发支出占比约 45% 35% 的研发费率不仅轻松碾压所有传统医疗软件同行，甚至与顶尖硅谷硬件及科技巨头苹果平起平坐，仅次于大举押注未来前沿技术的 Google。造成这种惊人财务结构的深层逻辑其实非常明晰：由于 Epic 从不进行对外并购，通用与行政支出（G&A）压得极低，更重要的是其在销售与市场营销（Sales & Marketing）上的预算常年维持在不可思议的 **“零元”** 状态。省去传统企业最庞大的销售雇佣兵团队后，Epic 能够将赚到的绝大部分现金直接反哺进内部工程研发，进一步筑高其产品功能的壁垒。 ### [03:20:48](https://www.youtube.com/watch?v=FNRU9ZNqcls&t=12048s) 研发投入的复利效应与 Paul Graham 的“爬坡理论” 在极低销售与市场营销支出的背景下，Epic 的商业模式展现出惊人的长期威力。即使将公司内部负责销售事务的约八个人全部归入销售人员编制，其占总支出的比重依然微乎其微。如果一家软件公司能在早期不依赖昂贵的销售和营销费用而赢得客户，并将几乎全部资金压注在研发（R&D）上，这些技术资产便会随着时间推移产生巨大的复利效应，最终构建起其他依靠营销驱动的竞品难以企及的壁垒。这种稳健的扩张逻辑在朱迪·福克纳（Judy Faulkner）对业务增长的理解中体现得淋漓尽致。她曾将自己的经营方式形容为“登山”——她不会试图一眼看清整座高山的全貌，而是把所有注意力集中在眼前要翻越的下一座小山丘。这一哲学恰好与保罗·格雷厄姆（Paul Graham）在 2012 年发表的经典文章《创业等于增长》（Startups = Growth）形成了惊人的暗合。格雷厄姆在文中写道，尽管理论上这种“爬山算法”（hill climbing）可能会让初创公司陷入局部最优解，但现实中几乎从未发生过；每周盯紧增长数字迫使创始人不断行动，而行动本身就是成功的决定性因素。大多数时候，坐而论道只是拖延的变体，而且创业想法的最优点往往并非孤立存在，好的想法旁边通常紧挨着更好的想法。朱迪大概率从未读过保罗·格雷厄姆的文章，却在威斯康星州的农场里凭借直觉践行了完全相同的理念。两位主持人感叹，朱迪作为程序员转型为企业家的轨迹，与 Y Combinator 所倡导的黑客精神本质相通：凭借计算机科学底层思维与反直觉的常识来运营企业。如果 YC 当初不是走向风险投资模式，而是号召创业者彻底摒弃外部融资、完全依靠业务自负盈亏，或许世界上就会出现更多像 Epic 这样的巨头。 ```mermaid flowchart TD accTitle: Epic Growth and R&D Compounding accDescr: 朱迪·福克纳的研发复利驱动与爬坡式成长模型 subgraph CapitalStrategy["资金投入与成长逻辑"] A["极简销售与营销 (仅约8名销售相关人员)"] --> B["绝大部分资源聚焦 R&D"] B --> C["软件资产长期复利 构建深厚技术壁垒"] end subgraph HillClimbing["爬山理论 (Hill Climbing)"] D["不预设整座大山全貌"] --> E["专注翻越眼前的小山丘"] E --> F["发现相邻的更大更具价值的机遇"] end CapitalStrategy --> HillClimbing ``` --- ### [03:24:01](https://www.youtube.com/watch?v=FNRU9ZNqcls&t=12241s) 强硬的法律防线：知识产权、竞业协议与反垄断诉讼如果说朱迪在管理上呈现出反传统的工程师风格，那么在捍卫公司资产方面，Epic 则表现得极度强势且毫不退让。Epic 拥有极具进攻性的诉讼策略：在知识产权上寸步不让，坚决打击专利流氓；在员工雇佣法领域，公司不仅通过严格的竞业禁止协议防范核心机密外泄，甚至曾一路打到美国最高法院，赢得了一场确立雇主有权要求员工签署强制仲裁协议的里程碑式诉讼。在行业生态层面，Epic 同样对任何可能违反其服务条款的医疗 IT 公司施加严密遏制，引发了多起法律冲突。其中最具代表性的是初创公司 Particle Health 对 Epic 提起的诉讼，指控 Epic 涉嫌违反《谢尔曼反垄断法》（Sherman Antitrust Act），目前该案件仍处于极为关键的博弈阶段。Epic 始终毫不掩饰地利用法律武器来守护自身认定的资产边界，不断加固其护城河。然而与这种强硬手腕形成极其鲜明对比的是，Epic 创业数十年间从未对任何一位客户提起诉讼，也从未与客户陷入任何法律调解程序。在朱迪的文化理念中，客户关系享有近乎神圣不可侵犯的地位，这种“一切为了客户”的信条构成了 Epic 企业价值观不可动摇的一体两面。 - **知识产权与专利防护**：积极应对并痛击各类专利流氓（Patent Trolls），严防技术资产外溢。 - **劳工法律与竞业限制**：推行严格的保密与竞业条款，曾在最高法院赢得关于员工强制仲裁的标杆性判决。 - **行业反垄断纠纷**：正面临 Particle Health 依据《谢尔曼反垄断法》发起的反垄断指控。 - **不可逾越的底线**：自创立至今，Epic 从未与任何一家客户对簿公堂或参与争议调解。 --- ### [03:25:26](https://www.youtube.com/watch?v=FNRU9ZNqcls&t=12326s) 超强版企业软件销售：医疗监管如何扼杀自下而上的 PLG 普通的企服软件销售本就充满挑战：漫长的销售周期、采购决策者与终端使用者严重脱节、庞杂的功能导致界面充斥着各类繁琐的下拉菜单与按钮，以及定制需求多、实施周期长等通病。但在医疗健康领域，这些难题还要面临“加了类固醇”般的超级放大——HIPAA（健康保险可携性与责任法案）的严苛合规要求、患者生命安危的连带风险，以及极度敏感的数据安全红线。这直接导致企服领域屡试不爽的“产品驱动增长”（PLG，即 Product-Led Growth）在此完全失灵。在日常 SaaS 领域，初创团队往往可以通过类似 Slack、Figma 或 Notion 的模式，引导一线员工在部门内部刷信用卡先行使用，形成自下而上的渗透。但在医疗体系中，根据 HIPAA 规定，处理受保护健康信息必须签署商业伙伴协议（Business Associate Agreement, BAA），而签署该协议的人必须拥有能合法代表医院整体组织的法定授权。这意味着任何单独的科室、医生或诊所护士都无权私自引入一套未经机构批准的医疗软件。每一笔采购必须从顶层穿透，由医院首席信息官（CIO）亲自审批。正是这种制度性的合规天堑，让自下而上的打法几乎无法诞生下一个现代电子健康档案（EHR）巨头，也进一步巩固了 Epic 在顶层直销体制下的绝对统治地位。 --- ### [03:27:27](https://www.youtube.com/watch?v=FNRU9ZNqcls&t=12447s) 特立独行的女性创始人与百亿美元估值讨论在两位主播眼中，朱迪·福克纳身上洋溢着与沃伦·巴菲特、英瓦尔·坎普拉德（宜家创始人）及杰夫·贝佐斯同质的“古怪与极致”。她曾公开表示自己“更有可能工作到死而非退休”，其将一生奉献给 Epic 的专注度犹如坎普拉德之于宜家；她对客户反馈的敏锐与急速响应堪比贝佐斯；而她对“死磕技术的极客最终必将胜出”的坚信，也在医疗信息化市场得到了验证。回顾商业史，朱迪几乎是全球公认白手起家最成功的女性创始人。在潜在人选对比中： - 雅诗兰黛（Estée Lauder）虽然在疫情期间市值曾短暂冲破 1000 亿美元，但随后滑落至约 200 亿美元； - 泰勒·斯威夫特（Taylor Swift）与奥普拉·温弗瑞（Oprah Winfrey）的商业版图庞大，但估算其整体商业估值分别约为 110 亿美元及更低水平； - 《福布斯》榜单上常年位居全美白手起家女首富的戴安·亨德里克斯（Diane Hendricks），其创办的屋顶建筑材料供应巨头 ABC Supply 年营收虽达到约 200 亿美元，但其利润率远无法与 Epic 的纯软件业务相提并论。一个有趣的巧合是，全美最成功的两位白手起家女性创始人竟然全都扎根在威斯康星州。对于 Epic 的身价，两位主持人一致认为其公允估值至少在 1000 亿美元以上。在被问及“如果现在有人以 1000 亿美元估值向你转让 Epic 股票，你会买吗？”时，主播毫不犹豫地给出肯定答复：只要 Epic 能够持续提供丰厚的分红派息，拥有这样一家具备极深护城河与现金流能力的公司股票，绝对是值得终身持有的绝佳资产。 ### [03:29:57](https://www.youtube.com/watch?v=FNRU9ZNqcls&t=12597s) 估值思考与创始人的长期执掌力量针对 Epic 的潜在估值，主持人探讨了投资心态。如果纯粹为了博取流动性事件（如上市套现），持有 Epic 的股票显然是不切实际的，因为创始人早已立法规避了上市或被并购；但从个人持有优质资产的角度来看，以 1000 亿美元的估值买入 Epic 股份是完全让人心动的——“掏出钱包来检验，你是否愿意在这个价格接盘？”虽然外界实际上永远无法买到其股份，但其强大的基本面让其成为一个极具吸引力的假想投资标的。这种稀缺资产的背后，最令人震撼的核心驱动力之一是“创始人连续性”（founder continuity）。Epic 由创始人 Judy Faulkner 独自执掌了长达 47 年。横向对比科技与商业巨头：黄仁勋（Jensen Huang）领导英伟达已 31 年，扎克伯格（Mark Zuckerberg）掌舵 Meta（Facebook）约 21 年，唯一在跨度上超越 Judy 的唯有巴菲特（Warren Buffett）管理伯克希尔·哈撒韦（Berkshire Hathaway）超过 50 年的例外。Judy 历经 47 年，将个人的性格与价值观深深烙印在整个组织的肌理之中，陪伴它从起步成长到如今掌控全美医疗基础设施的庞大规模，并且至今依然保持着极高的业务参与度，这种长周期的深耕极为罕见且异常强大。 ### [03:31:17](https://www.youtube.com/watch?v=FNRU9ZNqcls&t=12677s) Epic 的空头逻辑：国际化受限、反垄断诉讼与数据法案在讨论 Epic 的做空论点（Bear Case）时，主持人列举了几个潜在的隐患和结构性挑战： - **国际化拓展的付费意愿差距**：Epic 过去依赖于体量庞大且医疗支出占 GDP 比例极高的美国本土市场，客户预算充裕；然而，国际市场的医疗体系与行政预算远不及美国宽裕，为医疗管理系统付费的意愿普遍大幅降低，因此海外扩张的回报可能会大打折扣。 - **Particle Health 反垄断诉讼风险**：若 Particle 对 Epic 发起的诉讼进一步演变成触犯《谢尔曼反垄断法》（Sherman Act）的实质性反垄断调查或惩处，这将是一次动摇公司根基的致命事件。 - **《21世纪治愈法案》（21st Century Cures Act）反信息封锁规定**：该法案在法律上禁止医疗系统或电子健康记录（EHR）供应商封锁数据访问。即使 Epic 过去出于数据隐私、安全及患者保护等合理理由严守围墙，现在屏幕抓取（screen scraping）、Chrome 插件或机器人流程自动化（RPA）提取 Epic 数据在法律上已获许可。 - **被“最佳单点方案（Best-of-Breed）”蚕食的理论可能**：理论上，若有人开发出一款覆盖 70% 客户的精美前端 Chrome 插件，用户可能沉浸在优质 UI 中，最终尝试用其他后台替代 Epic；或者互操作性普及后单点方案崛起。但实际上，这种可能概率近乎于零，因为医院 CIO 依然倾向于使用 Epic 的一揽子企业授权（Microsoft-style bundle），既省钱又免去维护多个单点供应商的复杂成本。 - **范式转移（AI 与以价值为导向的医疗）**：AI 是否会彻底重塑医疗系统的需求？或者以价值为基础的护理（value-based care）是否会彻底改变就医模式，使得大部分传统诊疗和就诊记录模式过时？但目前来看，这些论点过于宽泛（handwavy），且 Epic 本身也在紧密布局这些变化。 ### [03:35:16](https://www.youtube.com/watch?v=FNRU9ZNqcls&t=12916s) 多头逻辑：横向扩张与医疗系统的巨额低效浪费多头论点（Bull Case）则显得极为坚实。Epic 成功从最早的门诊记录加计费系统，横向扩张到各类临床专科模块，并持续向同一批医院客户交叉销售更多产品。更重要的是，Epic 正在利用其在医疗系统的绝对统治地位，将客户群体跨界拓展至药企（Pharma）、保险支付方（Payers）和科研机构（Researchers）。以**预先授权（Prior Authorization）**业务为例，这是 Epic 极具确定性的新增长点。医院本身并非利润丰厚的客户，而商业保险公司才是坐拥巨额资金的金主。谈到这里，主持人对美国医疗保健体系的畸形效率提出了严厉批评： - **万亿级的系统浪费**：研究表明，美国占 GDP 18% 的医疗总支出中，有高达 30% 纯属无谓浪费；另一份量化研究显示该浪费金额高达 8000 亿美元，相当于浪费了瑞士整整一年的国家 GDP。 - **个人感知的账单脱节**：普通家庭一年为医保支付 2.5 万至 3 万美元的保费，然而在真实的就诊需求中，消费者几乎不可能实际消耗掉对等价值的服务。从标价、协议价格到报销额度，整套医疗账单数字全充斥着虚高与造假，导致普通消费者承担着极其昂贵的成本却得不到相应的优质体验。 ```mermaid flowchart TD accTitle: Epic Data Ecosystem accDescr: Epic leverages core hospital clinical data to cross-sell to outside entities while integrating AI scribes. subgraph CorePlatform ["Epic 核心平台"] EMR["电子病历 (EMR)"] Billing["计费系统 (Billing)"] Modules["各临床专科模块"] end subgraph Expansion ["横向商业拓展领域"] Payers["保险支付方 (Prior Auth 预授权)"] Pharma["制药企业 (临床数据与试验)"] Researchers["科研机构 (大规模医学研究)"] end subgraph AIScribes ["环境智能收听与记录 (Ambient AI)"] Nuance["Microsoft / Nuance (Dragon)"] Abridge["Abridge"] end EMR --> Modules Billing --> Modules CorePlatform --> Expansion AIScribes -.->|"结构化就诊数据"| CorePlatform ``` ### [03:38:05](https://www.youtube.com/watch?v=FNRU9ZNqcls&t=13085s) AI 赋能的突破口：环境智能收听（Ambient Listening）在所有 AI 医疗应用中，最具革命性且正在落地的场景便是**环境智能收听（Ambient Listening）**。长期以来，医护人员对 EHR 系统最大的抱怨在于“整天黏在电脑前”，医生不得不分心应对繁琐的下拉框、弹窗警报并手动敲打病历，有时甚至必须配备随行记录员（medical scribe）。而大语言模型与环境收听 AI 书记员的结合，使得系统能在医生与患者自然交谈的同时，全自动抓取、分类并结构化生成病历记录，极大地将医生从屏幕前解放出来。这项技术不仅已经真实落地且效果显著。Epic 目前正以开放生态的方式与行业头部玩家深度合作整合，其中包括微软（Microsoft）收购的医疗语音老牌霸主 Nuance（其前身为 Dragon Dictation），以及包括 Abridge 在内的新一代 AI 医疗初创公司。 ### [03:39:21](https://www.youtube.com/watch?v=FNRU9ZNqcls&t=13161s) 环境声 AI 录入与医疗创新的“咽喉要道” 在探讨 Epic 未来增长潜力时，主持人提到了像匹兹堡的 Abridge 和 Suki 这样近年来大放异彩的环境声 AI（Ambient AI）产品。这些应用能够直接接入 Epic，充当临床医生的“AI 书记员”（AI Scribe）。实际使用这些工具的医生对它们赞不绝口：接诊时医生再也不必全程盯着电脑屏幕敲键盘，而是能将注意力完全集中在患者身上，由 AI 在后台静默记录并自动生成病历结构化文本。对于当前普遍困扰医生群体的职业倦怠、视线疲劳以及繁重屏幕工作而言，这无疑是一剂强心针。这一领域的合作伙伴大多与 Epic 建立了集成关系，部分公司甚至与 Epic 存在股权认股权证（Warrants）或收入分成协议，当然外界也不排除 Epic 未来会亲自下场自研类似产品的可能性。这一趋势折射出一个不可忽视的事实：Epic 已经牢牢占据了整个美国医疗软件生态的“咽喉要道”（Choke Point）。在当前的行业格局下，究竟哪些前沿软件创新能够真正进入医院并交付到临床一线，决定权实际上握在 Epic 手中。对于任何一家试图面向医院医生研发突破性软件的初创公司而言，毫不夸张地说，能否获得 Epic 的首肯与接入许可，直接决定了其产品能否触达目标客户。 ### [03:40:25](https://www.youtube.com/watch?v=FNRU9ZNqcls&t=13225s) 超级多头假设：从电子病历到隐形 AI 操作系统对于环境声 AI 的发展前景，最基础的乐观预期（Bull Case）在于它开辟了一条高确定性的全新软件收入线。然而，借用节目此前讨论 IPL 时的逻辑，还有一种更为宏大的“超级多头假设”（Mega Bull Case）。一位大型医院系统的首席信息官（CIO）曾向主持人分享过一个激进设想：随着环境声 AI 技术的不断精进，未来的电子健康档案（EHR）是否会彻底淡出前台，转而演变成底层的“AI 操作系统”？毕竟从本质上看，EHR 的核心职能无非是两点：一是记录临床诊疗数据，二是服务于后续的医疗计费。计费数据随后被打包发送给医保支付方（政府或商业保险机构），由后者审核、裁定并完成最终赔付结算。如果整套流程都能被 AI 无缝接管并自主闭环，医院是否还真的需要如此庞大且复杂的面向用户的操作界面？在这个略带乌托邦色彩的未来愿景中，长期蚕食医疗体系效率的巨额行政臃肿（Administrative Bloat）和人力开销，或许能被环境声 AI 彻底吸纳化解。尽管主持人对此调侃道“等我亲眼看到才会相信”，因为削减行政成本谈何容易——若真要砍掉大量待遇优渥的医院行政岗位，整个社会都将面临巨大阵痛；但如果 Epic 顺理成章地蜕变为这一套全自动 AI 体系的底层中枢，对 Epic 而言将是不可限量的商业价值。 ```mermaid flowchart TD accTitle: 医疗数据与计费流转演进图 accDescr: 展示传统流程与未来环境声AI整合后的演变路径 subgraph Legacy["传统模式：依赖前台重度人工与界面"] A["医生与患者问诊"] --> B["繁琐的手动病历录入 / 屏幕操作"] B --> C["医院计费与编码团队介入"] C --> D["报销材料递交至支付方（商业保险/政府）"] D --> E["支付方人工与系统审核、结算"] end subgraph Future["超级多头设想：Epic 与环境声 AI 操作系统"] F["医生专注患者交流"] --> G["环境声 AI 静默捕捉全流程数据"] G --> H["后台系统全自动完成临床建档与账单编码"] H --> I["AI 对接支付端自动化智能审定与赔付"] end ``` 除了 AI 带来的增量空间，Epic 的基本盘还拥有多重宽阔的护城河： - **极高的客户黏性**：几乎没有任何大型医院系统会从 Epic 迁出，存量客户的基础收入每年都在稳健增长，且正不断吸引竞争对手的客户转投门下。 - **监管红利与合规壁垒**：作为行业既得利益者，严苛的行业监管往往能把新入局者挡在门外，因为唯有体量庞大、资源丰厚的在位巨头才有能力负担高昂的合规成本。例如面对正在推进的全美跨机构数据互操作性框架《受信任的交换框架和通用协议》（TEFCA），Epic 凭借充沛的技术储备能够率先完成高质量适配，进一步加深对市场的掌控。 - **潜在的提价空间**：虽然 Epic 凭借自身文化不大可能会采取激进手段压榨客户短期价值，但其技术栈在商业上确实享有极高的提价宽容度。 - **生态平台化构想**：尽管 Epic 过去始终坚持自研一切，但如果未来沿着 App Orchard 等尝试进一步开放，成长为支撑第三方应用构建的坚实底层平台，其商业天花板还将进一步抬高。 ### [03:44:07](https://www.youtube.com/watch?v=FNRU9ZNqcls&t=13447s) 精髓探讨：垂直软件制胜法则与医疗行业的“惩罚大棒” 进入“精髓”（Quintessence）讨论环节，主持人们提出了一个核心拷问：为什么在任何其他领域，如果一家初创公司声称要“自己亲手打造底层平台以及所有的垂直应用程序”，投资人都只会斥之为愚蠢并果断拒绝，偏偏在医疗健康领域，Epic 这种将全套应用与平台大包大揽的模式却能大获全胜？答案的本质在于**垂直市场软件（Vertical Market Software）与水平市场软件（Horizontal Market Software）之间截然不同的制胜法则**： - **服务客群的深度 vs 广度**：水平软件（如 Microsoft Office、Windows、Salesforce、Slack 等）面向各行各业的泛化人群，产品经理常言“设计糟糕产品最快的方式就是盲目听取单一客户的意见”，团队必须在众多客群的诉求中抽象共性，时刻警惕在某个特定分支走得太深；而垂直软件只服务单一垂直赛道，产品做得越贴近客户的实际业务运转、解决的问题越深入越好。在垂直领域，倾听客户是至高法则，而 Epic 恰恰将这一点做到了极致——深刻理解医院 CIO 与 CFO 的真正痛点，并为其量身定制每一处细节。 - **医疗行业的“惩罚大棒”（Sticks）**：相比胡萝卜，医疗行业充斥着不能犯错的严苛惩罚。其一是人命关天——软件一旦发生故障或数据脱节，直接后果是“患者可能会死”，因此一家高度整合的一体化单一系统，远比拼凑多个厂商的散落工具要安全得多；其二是法律合规的高压线——稍有不慎触犯 HIPAA 法案发生数据泄露就是灭顶之灾，若无法满足医保规范，更会直接被取消联邦医疗保险（Medicare）和医疗补助（Medicaid）的巨额补贴。在这样一个容错率为零的高风险环境中，选择单一供应商提供的端到端集成系统，成为了医院管理层最理性的决策路径。即便这种重型路线需要像 Epic 那样耗费整整 47 年的漫长时间去雕琢海量代码与打磨协同体系，它也构成了行业最牢固的竞争壁垒。 ### [03:47:20](https://www.youtube.com/watch?v=FNRU9ZNqcls&t=13640s) 全球最具价值的垂直市场软件巨头从商业体量来看，垂直市场软件公司通常很难成长为超大型科技巨头，因为其潜在市场规模（TAM）受制于特定的行业边界。这也是为什么此类节目极少对垂直软件企业做深入长篇报道的原因。然而 Epic 却是一个极其罕见的特例：它所盘踞的美国医疗健康市场规模异常庞大，甚至占到了全美 GDP 的近两成，与每一个人的生老病死息息相关。这使得 Epic 极有可能已经成为了**全球商业价值最高的垂直市场软件公司**。主持人们横向比对了全球科技界可能与其比肩的名字： - **彭博（Bloomberg）**：在金融终端与信息服务领域几乎一统天下，大概是唯一能在估值与行业重要性上与 Epic 争雄的未上市垂直软件公司。 - **Veeva Systems**：同样脱胎于医疗与生命科学垂直领域的 SaaS 上市标杆，市值大约在 200 亿美元左右，但无论在业务体量、医院系统的核心枢纽地位还是整体内在估值上，都完全无法与 Epic 相提并论。面对这片庞大市场，Epic 凭借深耕单一终端领域的战略专注度，缔造了垂直软件领域前所未有的商业奇迹。 ### [03:48:43](https://www.youtube.com/watch?v=FNRU9ZNqcls&t=13723s) 垂直行业软件之王与“资本主义奥运金牌” 在总结 Epic Systems 的地位时，两位主播探讨了整个垂直市场软件（Vertical Market Software, VMS）领域中是否存在同等级别的巨头。唯一能被拉来相提并论的只有 Constellation Software。然而，Constellation 的商业模式本质上是多行业资产的聚合与并购滚雪球（roll-up），即便其旗下体量最大的子公司同样集中在医疗 IT 领域，也与 Epic 这种从单一核心软件生长出来的全系统不可同日而语。主播调侃道，Constellation 创始人马克·莱昂纳德（Mark Leonard）哪怕运营着 Epic 的竞争对手，也必然对朱迪·福克纳（Judy Faulkner）充满敬意，因为朱迪在商业运作上无疑斩获了一枚“资本主义奥运会金牌（gold medal in the capitalism Olympics）”。 --- ### [03:49:25](https://www.youtube.com/watch?v=FNRU9ZNqcls&t=13765s) David 的 Carveouts：传奇拉力车手与备受期待的 Switch 2 在轻松欢快的 Carveouts 分享环节，David 带来了两件推荐。首先是他最近才在朋友推荐下看到的拉力赛车手肯·布洛克（Ken Block）于 2012 年拍摄的旧金山街头漂移视频（Gymkhana 5）。在这段长达 10 分钟的视频中，旧金山多条街道被完全封锁，布洛克驾驶一台 600 马力的拉力赛车在起伏的坡道和弯角极速飞驰，宛如现代版的史蒂夫·麦奎因电影《警网铁金刚》（Bullitt）。David 深入了解后发现，肯·布洛克不仅是顶尖赛车手，居然还是著名滑板鞋品牌 DC Shoes 的联合创始人之一，在将公司出售给 Quiksilver 之后才投身职业赛车。虽然布洛克近期不幸因意外离世，但他留下的街头视觉震撼依旧无与伦比。 David 的第二件是“预先安利（preemptive carveout）”——任天堂即将推出的 Nintendo Switch 2。该主机完全符合播客此前在任天堂专题节目中畅想的一切配置：全面的向后兼容性、保留原有的线上账户系统，并预计将随全新的《马力欧卡丁车》（Mario Kart）一同首发。对于 David 而言，这不仅是属于硬核玩家的兴奋，更承载着一位父亲的心愿——自大女儿出生以来，他就一直盼着能带她一起打游戏，如今女儿终于长到了能握住手柄玩卡丁车的年纪。 - **Ken Block 旧金山漂移视频 (2012)**：封锁旧金山街道实拍，600 马力拉力赛车展现惊人操控；车手曾联合创立滑板品牌 DC Shoes。 - **Nintendo Switch 2**：确认硬件向后兼容、账号互通并配发新作，成为亲子游戏时光的理想期待。 --- ### [03:51:57](https://www.youtube.com/watch?v=FNRU9ZNqcls&t=13917s) Ben 的 Carveouts：爆米花悬疑片、流行神专与《人生切割术》混音集随后，Ben 分享了他的三个心水推荐。第一部是近期重温的经典悬疑推理电影《利刃出鞘》（Knives Out）。影片由丹尼尔·克雷格（Daniel Craig）主演，情节抓人，充满本格推理的“暴风雪山庄/找凶手（Who Done It）”趣味，是一部制作精良、视觉考究且纯粹解压的爆米花佳作。第二项是查莉·XCX（Charli xcx）在 2024 年引爆全球流行文化的专辑《Brat》。Ben 笑称自己是“最晚入坑的千禧一代”，但自从在格莱美颁奖礼上看完现场表演后，就彻底沉迷其中循环播放。他表示整张专辑充满强烈的律动感，“只要戴上耳机，就能瞬间把你从电脑屏幕前的枯燥现实，拽进一个远比现实更酷、更尽兴的异世界派对中”。最后，Ben 推荐了电子乐组合 ODESZA 为 Apple TV+ 热播神剧《人生切割术》（Severance）制作的混音专辑《Music to Refine to》。该专辑将剧中标志性的冷冽配乐重新解构为极具质感的低保真（Lo-fi）电子音乐，非常适合作为日常工作和深度专注时的背景伴奏。 - **《利刃出鞘》（Knives Out）**：由丹尼尔·克雷格领衔的经典群戏悬疑片，视觉华丽且情节高度引人入胜。 - **Charli xcx 专辑《Brat》**：2024 年现象级流行电子专辑，节奏热烈、沉浸感极强。 - **ODESZA《Music to Refine to》**：美剧《人生切割术》官方混音概念伴奏，极佳的工作伴读 Lo-fi 音乐。 --- ### [03:53:27](https://www.youtube.com/watch?v=FNRU9ZNqcls&t=14007s) 致谢与幕后访谈：走进极度低调的医疗软件帝国节目的尾声，两位主播特别感谢了为本期庞大调研提供内幕支持的各界专家与业内人士。由于市面上几乎没有任何关于 Epic Systems 的系统性长篇商业专著，Acquired 团队破例直接向极度低调、鲜少接受公开采访的创始人朱迪·福克纳发送了邮件，出乎意料地得到了朱迪本人的回复与长时间深入交流。同时，Epic 的核心高管团队卡尔·德沃拉克（Carl Dvorak）、苏米特·拉纳（Sumit Rana）与李·利维特（Lee Leavitt）也为节目还原了许多未曾公开的真实创业历史与技术演进节点。 ```mermaid flowchart TD accTitle: Epic Systems 专题调研信息网络 accDescr: 展示为本期 Acquired 节目提供一线深度洞察的四类关键访谈信源 subgraph Core["Epic 核心管理层"] Judy["朱迪·福克纳 (Judy Faulkner, 创始人/CEO)"] Execs["卡尔·德沃拉克、苏米特·拉纳、李·利维特 (Carl Dvorak, Sumit Rana, Lee Leavitt)"] end subgraph Industry["行业分析师与投资人"] Analysts["Arvinder Navaratnam (Worldly Partners) Brendan Keeler (Health API Guy) Brian Lawrence (Oakcliffe Capital) Lan Wang (JSL Health Capital) Patrick Wingo (Elion / 前 Palantir)"] end subgraph Clinicians["医疗系统 CIO 与临床医生"] CIO["Michael Pfeffer 博士 (斯坦福医疗集团 CIO)"] Doctors["Molly Cantor (UCSF 医生) Corinne D'Amato (MGH 医生) Abed Mir 博士 (斯坦福医生 / Epic 股东)"] end subgraph Alumni["资深前员工"] JohnB["John Bertrand (Digital Diagnostics CEO / 前 Epic 员工)"] end Core --> Podcast["Acquired: Epic Systems 深度内幕与商业编年史"] Industry --> Podcast Clinicians --> Podcast Alumni --> Podcast ``` 除了公司高层，调研还覆盖了医疗信息技术领域的多位重磅专家、一线医院首席信息官（CIO）以及长期在临床系统中使用 Epic 系统的执业医生们，共同勾勒出这家掌控全美医疗数据命脉的软件巨头的完整图景。最后，主播预告了将于 7 月 15 日在纽约市举行的线下活动（acquired.fm/nyc），向季度赞助伙伴致谢，并推荐听众回听医疗赛道另一巨头诺和诺德（Novo Nordisk）的往期节目，以及前往社区 Slack 参与深度商业讨论。 - **调研支持学者与投资人**：Worldly Partners 的 Arvinder Navaratnam（撰写了多十年跨度的 Epic 深度研究）、Substack 专栏《Health API Guy》作者 Brendan Keeler、Oakcliffe Capital 创始人 Brian Lawrence、JSL Health Capital 管理合伙人 Lan Wang、Elion 研究主管 Patrick Wingo，以及资深医疗 IT 高管 Scott Gaines。 - **临床实践与文化观察视角**：Digital Diagnostics 首席执行官、8VC 顾问兼 Epic 前员工 John Bertrand；斯坦福医生兼投资人 Abed Mir 博士；斯坦福医疗中心 CIO Michael Pfeffer 博士；旧金山加州大学（UCSF）医生 Molly Cantor 与哈佛麻省总医院（MGH）医生 Corinne D'Amato。 - **本季官方合作伙伴**：J.P. Morgan Payments、ServiceNow、Crusoe 以及 Fundrise。 ## Notable quotes > [09:56](https://www.youtube.com/watch?v=FNRU9ZNqcls&t=596s) 一个小孩子在玩黏土 — Judy Faulkner > [14:42](https://www.youtube.com/watch?v=FNRU9ZNqcls&t=882s) 我们正在接诊的这批患者，医院里的其他科室也在接诊。但是当患者穿梭于不同科室之间时，各科室之间根本无法知道在其他科室都发生了什么。我们极度渴望能有一个单一的数据库，能够在患者于威斯康星大学医学中心接受诊疗的整个纵向病程（longitudinal course of care）中，完整保存每个人的医疗记录。 > [32:58](https://www.youtube.com/watch?v=FNRU9ZNqcls&t=1978s) 阳光正好照进来，我当时正有些心不在焉，就那么坐着，脑子里突然豁然开朗：‘集成系统应该这样做！’我记得自己一路飞奔冲进厨房，抓起一个便签本，立刻在纸上疯狂地写代码，一行接一行。那段代码后来变成了 Chronicles。 — Judy Faulkner > [36:14](https://www.youtube.com/watch?v=FNRU9ZNqcls&t=2174s) 行吧，我开公司还不行吗？我就兼职干干，拉几个身边的同事一起兼职搞，就当成一个小打小闹的副业。 — Judy Faulkner > [01:21:55](https://www.youtube.com/watch?v=FNRU9ZNqcls&t=4915s) 无论是选我们还是选他们，只要为了你们的患者、整体系统、顺畅的账单结算，以及门诊与住院部之间无缝的数据流转，你们就必须从我们两家之中只选一家做全套系统。 — Judy Faulkner > [01:26:12](https://www.youtube.com/watch?v=FNRU9ZNqcls&t=5172s) 鉴于其庞大的规模，Kaiser 与 Epic 联手打造的这套系统，极有可能成为整个医疗健康软件行业的‘福特T型车’——它或许不是同类产品中诞生的第一辆，但绝对是第一辆走向大众普及的工业标杆。 > [01:39:58](https://www.youtube.com/watch?v=FNRU9ZNqcls&t=5998s) 不可上市 — Judy Faulkner > [01:39:58](https://www.youtube.com/watch?v=FNRU9ZNqcls&t=5998s) 不可收购其他公司，亦不可被收购 — Judy Faulkner > [01:52:17](https://www.youtube.com/watch?v=FNRU9ZNqcls&t=6737s) 虽然 Epic 绝对不便宜，但你花出去的每一分钱，100% 都能得到实打实的回报。 — Jeff Gotney > [02:03:21](https://www.youtube.com/watch?v=FNRU9ZNqcls&t=7401s) 为了全体美国人民，我们必须直面不断飞涨的医疗成本，巩固医患关系，并帮助人们承担得起他们所需的医疗保险。我们将更广泛地利用电子病历（electronic records）与其他卫生信息技术，以此帮助控制成本并减少致命的医疗差错。 — George W. Bush > [02:32:44](https://www.youtube.com/watch?v=FNRU9ZNqcls&t=9164s) 事实上，依然有堆积如山的文书工作，医生们依然不得不手动录入大量内容，护士们的所有时间也都被这些行政事务填满了。我们砸进了一大笔钱，试图鼓励所有人实现数字化以赶上世界其他国家的步伐，但这比我们想象的要艰难得多。 — Barack Obama ## Mentioned Epic Systems, MyChart, Judy Faulkner, Lou Greenfield, Dell Greenfield, Lou's Soda Fountain, William Shockley, Bell Labs, Dickinson College, University of Rochester, Stanford University, University of Wisconsin–Madison, Dr. Warner Slack, Massachusetts General Hospital (MGH), American Medical Association (AMA), American Hospital Association (AHA), Association of Health Record Librarians of North America, Rockefeller Foundation, Medicare, Medicaid, Stabilization Act of 1942, National War Labor Board, COSTAR, MUMPS, Caché, Neil Pappalardo, MEDITECH, Chronicles, Human Services Computing, JPMorgan Payments, Data General Eclipse, Sequoia Capital, Pleasant Roland, American Girl Doll, Mattel, Resolute, Cadence, EpicCare, EpicCare Inpatient, Sumit Rana, Carl Dvorak, Kaiser Permanente, Accenture, IBM, Cerner, Oracle, Neal Patterson, Philips, Philips Externity Enterprise, Microsoft, Deep Space, Rembrandt Test, Peter Drucker, Palantir, Care Everywhere, Cosmos, ServiceNow, George W. Bush, Barack Obama, HITECH Act, ARRA, Meaningful Use, NextGen Healthcare, Rusty Frantz, Kaiser Health News, Allscripts, Veradigm, Misys, Eclipsys, Siemens Health Services, Johns Hopkins, Cedars-Sinai, UCSF, Crusoe, Department of Defense (DoD), Department of Veterans Affairs (VA) ## Video description What if we told you that the most important company in US healthcare was run from a farm in rural Wisconsin? And that farm contained the world’s largest subterranean auditorium, as well as Disneyland—style replicas of Hogwarts and the Emerald City? What if we told you that the person who started, runs and owns this establishment has legally ensured that it will never be sold, never go public and never acquire another company? And that this person, Judy Faulkner, is also likely the wealthiest and most successful self-made woman in history? Welcome to the story of Epic Systems, the software company that underpins the majority of the American healthcare system today. Epic isn’t “just” an electronic medical record (the category it’s usually lumped into), or an online patient portal (which is how most of the US population interacts with it via its MyChart application). It’s more akin to a central nervous system for hospitals and health clinics. Almost *everything* in a hospital — from patient interactions to billing, staffing, scheduling, prescriptions and even research — happens on Epic’s platform, and over 90% of American medical schools’ graduating doctors, nurses and health administrative staff are trained on it during their educations. Tune in as we dive into the almost-unbelievable story of how this epic company came to be! *Links:* - Epic’s Verona campus: https://www.epic.com/visiting/ - Worldly Partners’ Multi-Decade Epic Systems Study: https://worldlypartners.com/businesshistory - Episode sources: https://www.acquired.fm/episodes/epic-systems-mychart#sources *Carve Outs:* - Ken Block in San Francisco: https://youtu.be/LuDN2bCIyus?si=rHU8CjPAoeHck0uK - Nintendo Switch 2: https://www.nintendo.com/us/gaming-systems/switch-2/?srsltid=AfmBOoq-kZTYsNm0_if1ONE-8GV0k1x77Nv41aWbkSpgrqSU9zcyQAog - Knives Out: https://www.imdb.com/title/tt8946378/ - Brat by Charli xcx: https://open.spotify.com/album/2lIZef4lzdvZkiiCzvPKj7?si=bdGrTidgRIKMMBNvjHzbWQ - Music To Refine To: A Remix Companion to Severance by ODESZA: https://open.spotify.com/album/7bNj7hUkbRbZzn36MdyvUk?si=tL7uLI15QVasFLXuRaKkZQ *More Acquired:* - Get email updates https://www.acquired.fm/email and vote on future episodes! - Join the Slack http://acquired.fm/slack - Check out the latest swag in the ACQ Merch Store https://www.acquired.fm/store! _Note: Acquired hosts and guests may hold assets discussed in this episode. This podcast is not investment advice, and is intended for informational and entertainment purposes only. You should do your own research and make your own independent decisions when considering any financial transactions._ [Source](https://www.youtube.com/watch?v=FNRU9ZNqcls)

Source quotes

[9:56] Judy Faulkner: 一个小孩子在玩黏土 [14:42] 我们正在接诊的这批患者，医院里的其他科室也在接诊。但是当患者穿梭于不同科室之间时，各科室之间根本无法知道在其他科室都发生了什么。我们极度渴望能有一个单一的数据库，能够在患者于威斯康星大学医学中心接受诊疗的整个纵向病程（longitudinal course of care）中，完整保存每个人的医疗记录。 [32:58] Judy Faulkner: 阳光正好照进来，我当时正有些心不在焉，就那么坐着，脑子里突然豁然开朗：‘集成系统应该这样做！’我记得自己一路飞奔冲进厨房，抓起一个便签本，立刻在纸上疯狂地写代码，一行接一行。那段代码后来变成了 Chronicles。 [36:14] Judy Faulkner: 行吧，我开公司还不行吗？我就兼职干干，拉几个身边的同事一起兼职搞，就当成一个小打小闹的副业。 [81:55] Judy Faulkner: 无论是选我们还是选他们，只要为了你们的患者、整体系统、顺畅的账单结算，以及门诊与住院部之间无缝的数据流转，你们就必须从我们两家之中只选一家做全套系统。 [86:12] 鉴于其庞大的规模，Kaiser 与 Epic 联手打造的这套系统，极有可能成为整个医疗健康软件行业的‘福特T型车’——它或许不是同类产品中诞生的第一辆，但绝对是第一辆走向大众普及的工业标杆。 [99:58] Judy Faulkner: 不可上市 [99:58] Judy Faulkner: 不可收购其他公司，亦不可被收购 [112:17] Jeff Gotney: 虽然 Epic 绝对不便宜，但你花出去的每一分钱，100% 都能得到实打实的回报。 [123:21] George W. Bush: 为了全体美国人民，我们必须直面不断飞涨的医疗成本，巩固医患关系，并帮助人们承担得起他们所需的医疗保险。我们将更广泛地利用电子病历（electronic records）与其他卫生信息技术，以此帮助控制成本并减少致命的医疗差错。 [152:44] Barack Obama: 事实上，依然有堆积如山的文书工作，医生们依然不得不手动录入大量内容，护士们的所有时间也都被这些行政事务填满了。我们砸进了一大笔钱，试图鼓励所有人实现数字化以赶上世界其他国家的步伐，但这比我们想象的要艰难得多。

Source quotes

[33:32] Judy Faulkner: The sun was shining. I was disattentive. I was just sitting there and suddenly it all came to me. Here's how you build it. The integrated system. [35:11] Judy Faulkner: we don't need to build a whole company around this thing. [61:14] Sumit Rana: Hey, you know, this is good that I'm working on this, but I'm kind of bored. Can you give me something hard to do? [75:14] Physicians: Like we've got this thing called Epic up in Portland... and it's way better than the stuff that you're trying to build. [82:40] Carl Dvorak: Guys, no, we need to model out in Excel exactly what Kaiser's transaction flow is going to be [86:12] Judy Faulkner: No, we're not going to do it. We're not going to do it for you. We're not going to do it for anybody. [115:09] Epic Systems representative: I don't think you're ready to be a customer yet [120:18] Hospital CIO: wildly succeeded at digitization of the industry. It did absolutely nothing on digital transformation [210:26] Judy Faulkner: the nerds will prevail in the long run.

Original transcript

[0:00] The answer is somewhere in the middle. [0:02] Well, you texted me last night that [0:03] you've made it to Singapore. I made it [0:05] to Singapore. Yes. Anytime you are [0:07] researching anything in US healthcare, [0:09] you know it is time to stop your [0:11] research process and start the episode [0:12] once you've found [0:15] [Applause] [0:17] [Music] [0:18] Singapore. All right, let's do it. Who [0:21] got the [0:23] truth? Is it you? Is it you? Is it you? [0:26] Who got the truth? [0:28] Now, is it you? Is it you? Is it you? [0:31] Sit me down. Say it straight. Another [0:34] story on the way. [0:37] Welcome to the spring 2025 season of [0:40] Acquired, the podcast about great [0:42] companies and the stories and playbooks [0:43] behind them. I'm Ben Gilbert. I'm David [0:45] Rosenthal. And we are your hosts. [0:48] Listeners, today's episode is about a [0:50] quiet company in rural Wisconsin that [0:52] plays an enormous role in our lives. [0:54] Epic Systems. Indeed. Whether you know [0:57] it or not, yes, you probably know them [0:59] from their medical patient software, My [1:01] Chart, that if you're listening to this, [1:03] you most likely use, Epic is a very [1:06] unusual company in so many ways. They do [1:09] no marketing. They basically don't do [1:11] any sales either. They often say no to [1:13] potential customers who approach them. [1:15] They don't negotiate. They don't [1:16] discount. They never raised any venture [1:18] capital. And they've never done any [1:20] acquisitions in their 47 years of [1:22] existence. They don't work remotely. [1:24] Everyone is in person all the time. They [1:27] notoriously have one gigantic campus on [1:29] a farm with buildings designed to look [1:31] like the Land of Oz, a Wizards Academy, [1:33] a treehouse, a barn, a replica of New [1:35] York's Grand Central Station, and an [1:37] 11,000 seat auditorium underground. They [1:41] have the majority of the US's major [1:43] hospital systems using their software. [1:45] And of their over 600 customers, they [1:48] have never lost a single one. Yeah, that [1:51] is the craziest thing to me about this [1:52] company is 47 years old. They have never [1:56] lost a customer. Actually, we found out [1:58] that's not totally true. They lost one [2:00] customer once for 6 months and then that [2:03] customer came back 6 months later. Yes. [2:05] The company's founder, Judith Faulner, [2:07] is undoubtedly one of the great founders [2:09] of our time. You probably don't know [2:11] much about her or the company because [2:12] the company is still privately held and [2:15] Judy and her family foundation own about [2:17] half of it despite being large. And I [2:20] think at this point they're close to 6 [2:22] billion in revenue and over 14,000 [2:24] employees. They have a stated goal to [2:27] never go public and never be acquired. [2:29] And Judith at age 81 has created a [2:32] succession plan and a trust structure [2:34] for her voting shares to ensure that [2:35] that will stay true forever. Yes, we [2:39] heard all sorts of stories about [2:41] companies sniffing around Epic over the [2:42] years trying to buy the GE, Microsoft, [2:45] Google, you know, everybody you would [2:46] imagine wants to buy this company and [2:49] it's never going to happen. Yep. And [2:51] we'll dig into this at the end of the [2:53] episode when we sort of have all the [2:54] context and all the numbers, but I [2:56] believe that Judith Faulner in starting [2:58] one of the most valuable companies in [2:59] healthcare is the most successful female [3:03] entrepreneur in history. Almost [3:05] undoubtedly. Well, all right then. [3:08] Spoiler alert, listeners. [3:10] We'll discuss that at the end. Yes. So, [3:12] the health care industry, there is so [3:15] much wrong with the American health care [3:16] system. That is an incontrovertible [3:19] fact. There's nobody that's going to [3:20] tell you, "Oh, actually, it's pretty [3:22] good." It's not pretty good. It's a [3:23] disaster. Runaway costs, burdens of [3:25] administration, so much excess and waste [3:28] causing I think healthcare costs are now [3:30] 18% of our GDP. So rather than trying to [3:33] eat that whole elephant today and unpack [3:35] the entire system, today's episode is [3:38] about understanding Epic's role within [3:39] it and how Epic became so dominant. Yep. [3:42] And if you want to understand the [3:43] system, you have to understand Epic. [3:46] Yep. Well, listeners, if you want to [3:48] know every time an episode drops, check [3:49] out our email list. It is the only place [3:51] where we will share a hint at what our [3:53] next episode will be. We'll share [3:55] corrections from previous episodes and [3:57] little tidbits that we learned along the [3:59] way. So that's acquire.fm/email. [4:01] FM/Amail. After this episode, come join [4:03] the Slack. Talk about it with us and the [4:05] entire Acquired community afterwards. I [4:07] bet there's a ton of people in the [4:09] medical ecosystem hanging out in the [4:11] acquired Slack. That's [4:14] acquired.fm/slack. If you want more [4:16] acquired between each monthly episode, [4:18] check out ACQ2, our interview show where [4:20] we talk to founders and CEOs who are [4:22] building businesses in areas that we've [4:24] covered on the show. To go a little bit [4:26] deeper, search ACQ2 in any podcast [4:29] player. [4:30] So, as we announced last episode, we [4:33] have a very fun save the date for you. [4:36] We can't say much yet, but after [4:38] incredible listener demand over the [4:40] years, we are finally coming to New York [4:42] City with our friends at JP Morgan [4:44] Payments. So, July 15th, mark your [4:47] calendars and if you want to be the [4:49] first to find out what we are up to, [4:51] sign up at acquired.fm/nyc. [4:55] This is going to be a night of [4:56] absurdity. An incredible night. Yes. All [4:59] right, listeners, before we dive in, we [5:01] want to briefly thank our presenting [5:02] partner, JP Morgan Payments. Yes. And [5:05] just like how we say every company has a [5:07] story, every company's story is powered [5:09] by payments. And JP Morgan Payments is a [5:11] part of so many of their journeys from [5:12] seed to IPO and beyond. So with that, [5:15] this show is not investment advice. [5:16] David and I may have investments in the [5:18] companies that we discuss, although not [5:20] Epic. And this show is for information [5:22] and entertainment purposes only. David, [5:25] take us in. All right. Well, we start [5:29] our journey in August 1943 when Judy, [5:33] today Faulner, then Judy Greenfield, is [5:37] born in Earlton, New Jersey, which is [5:39] part of Cherry Hill, New Jersey, suburb [5:41] of Philadelphia, right across the [5:43] Delaware River there, not too far from [5:45] where you and I grew up. That's true. [5:47] And Taylor Swift. And Taylor Swift. [5:49] That's right. Fertile ground for [5:50] entrepreneurs there in the uh Delaware [5:53] River. And Dupant. So much great [5:55] American entrepreneurship in that area. [5:56] so much indeed. And of course, I don't [5:59] think Judy really knew about all this at [6:00] the time. But yes, indeed, it is a [6:02] pretty auspicious time and place to be [6:05] born because just about 4 years later [6:08] after Judy is born, just up the road, a [6:10] little ways from Cherry Hill in Murray [6:13] Hill, New Jersey, William Shockley and [6:15] his colleagues would invent the [6:17] transistor at Bell Labs that would [6:20] enable Microsoft and Epic and Intel and [6:24] all of this. And for a long time the [6:26] early pioneers of electronic healthcare [6:28] records were hardware companies. Yes. [6:31] Lockheed G Seamans. Yes. Lockheed was a [6:36] vendor to hospitals. Incredible. But for [6:39] the moment Judy probably didn't know [6:41] anything about this because her family [6:42] is not in the tech industry growing up. [6:44] Her father Lou is a small town [6:47] entrepreneur. He runs a local pharmacy [6:50] and soda fountain there in Earlton [6:52] called LSE Lou Soda Fountain. And [6:54] perhaps this is where Judy would later [6:56] get her own entrepreneurial bent from. [6:58] Could be. So that's Judy's father. Now [7:01] Judy's mother, Dell Greenfield, was an [7:04] absolute freaking dynamo. She graduated [7:07] high school at age 15 and she worked [7:10] first as a secretary and then she worked [7:12] with Lou at the store pharmacy and soda [7:15] fountain and then later got really [7:17] involved in peace advocacy during and [7:20] after the Vietnam War which I assume was [7:22] not typical for her generation that [7:25] lived through World War II. She ends up [7:27] becoming the director of the South [7:28] Jersey Peace Center. And then later in [7:31] life, after the kids were gone, she and [7:34] Lou moved to Portland, Oregon, where [7:36] Dell became the executive director of an [7:38] organization called Oregon Physicians [7:40] for Social Responsibility, which get [7:42] this, in [7:44] 1985, this group in partnership with a [7:47] broader international group called [7:49] Physicians for the Prevention of Nuclear [7:50] Wars the Nobel Peace Prize. Judy [7:54] Faulner's mom was part of a group that [7:56] won the Nobel Peace Prize in 1985. So, [7:59] David, where did you find this? Because [8:01] in everything that I read about Epic, [8:03] and you and I basically read everything [8:05] you possibly can read about Epic that's [8:07] out there on the internet. Nobody knows [8:09] this. This is not cited anywhere. So, [8:11] how did you find your way to it? I was [8:13] curious. I was trying to learn more [8:15] about Judy's family growing up. And I [8:16] was trying to learn about the soda [8:18] fountain and the pharmacy and the impact [8:20] that her dad had on her future [8:22] entrepreneurial career. And so I started [8:24] googling obituaries and I came across [8:27] her mom's obituary where I learned this. [8:29] The company actually confirmed this to [8:30] us. Yes, Judy's mom won a share of the [8:33] Nobel Peace Prize in 1985. Incredible. [8:36] So cool. All right. So you've got [8:38] entrepreneurial DNA. You've got sort of [8:41] peaceoriented divergent thinking DNA. [8:44] Yep. Then you've got Judy. And at this [8:47] point in time growing up, she's mostly [8:49] just interested in math. So she loves to [8:52] tell the story about how when she was in [8:53] seventh grade, a teacher asked the class [8:56] a number theory problem. Of course, [8:59] probably didn't frame it as a number [9:00] theory problem, but asked the class why [9:03] is it that numbers that are divisible by [9:05] three, if you add up all the digits of [9:08] that number, that number that is the sum [9:11] of its digits is also divisible by [9:13] three. M this is like a law of number [9:14] theory. So Judy hears this problem and [9:16] it's just like my future is in [9:20] math. So in 1961, Judy graduates high [9:24] school. She goes to Dickinson College to [9:26] major in math. And while she's there, [9:29] she gets a summer job one year at the [9:32] University of Rochester, just a little [9:33] bit farther north, in their particle [9:35] physics lab. And for the work that she [9:38] had to do over the summer, she needed to [9:40] learn computer programming. This is in [9:42] the early mid60s. And so like, oh, you [9:44] got to learn Forran for the work that [9:46] we're doing here and running these [9:47] experiments over the summer. So they [9:49] give her a book, like a manual on [9:51] Forran. And Judy teaches herself Forran [9:54] in a week and becomes one of the best [9:56] programmers in the lab. If you're not [9:58] getting the picture here, like Judy is a [10:00] genius. She is an incredibly talented [10:02] person. Yeah. And programming at a time [10:06] when programming wasn't a thing. The [10:08] field of software engineering was not a [10:10] field. These are math people taking [10:13] programming languages and using them. [10:15] But there are very few of those people [10:16] in the world. Universities didn't have [10:19] computer science departments until this [10:21] point in time. Judy is intertwined with [10:23] the beginning of computer science as a [10:25] field. Makes sense. So she absolutely [10:29] falls in love with programming that [10:31] summer in Rochester. She'd later say [10:32] that she felt like a kid playing with [10:34] clay and that programming a computer was [10:37] like this amazing combination of math, [10:39] which of course she loved, but it was [10:41] also language and art together with math [10:43] cuz she has an artistic side too for as [10:46] mathy as she is, she also has a hyper [10:49] creative streak. Totally. And obviously [10:53] her story is very different and she is a [10:55] very different person than him. But the [10:57] echoes that you're seeing here in Judy [11:00] in what she's exposed to as a kid, how [11:02] she thinks, how she operates as an [11:04] entrepreneur, you should be getting some [11:06] Bill Gates vibes here. I thought you [11:07] were going to say Steve Jobs with the [11:09] peace time orientation around her [11:12] upbringing and sort of the I don't want [11:14] to say hippie for Judy, but almost [11:16] hippieesque movement, too. Yeah, that's [11:19] actually a good point. Well, Jobs and [11:21] Gates. Well, the reason I said Gates is [11:22] to sort of foreshadow some Microsoftian [11:25] DNA and analogies that come into Epic [11:28] here. And they're contemporaries. She is [11:30] going to build her company in almost the [11:32] exact same time frame that Apple and [11:34] Microsoft were built. Totally. So, when [11:37] Judy gets back to Dickson to college, [11:39] she decides that she's going to apply to [11:41] grad schools in math. So, she applies to [11:43] five PhD programs. She of course gets [11:45] into all of them. And on her CV, she [11:48] lists her Rochester experience and her [11:50] Forran programming experience. So two of [11:54] the schools she applied to, Stamford and [11:56] the University of Wisconsin are just [11:59] starting their computer science [12:00] departments. When she applied, they [12:01] didn't have computer science [12:02] departments. And in the interim, they [12:04] started them. And so they saw Judy's [12:07] application like obviously this is a [12:09] brilliant person who we want here at the [12:11] university. We're starting these CS [12:13] departments. they unilaterally shift her [12:15] applications to their new CS schools. [12:18] And so Judy's like, "Oh, I didn't even [12:21] realize I could go get a PhD in computer [12:22] science." Like, amazing. This is what [12:23] I'm going to do. So, she ends up [12:26] choosing Wisconsin, goes off to Madison, [12:28] Wisconsin to start her PhD in computer [12:30] science, which, as we shall see, she [12:33] never finishes, but she also never [12:35] leaves. What a sliding door moment. If [12:37] Judy had gone to Stamford instead, we [12:39] would probably still be telling this [12:41] story about her, but it would be a very [12:42] different story. For sure. She would [12:44] have been indoctrinated by a very [12:46] different type of DNA and what computers [12:48] are for and what types of companies you [12:52] should be building with computers if she [12:53] was in the Silicon Valley at that point. [12:56] Totally. So, while she's at Wisconsin, [12:59] Judy takes a class called Computers in [13:02] Medicine taught by a faculty member from [13:05] the Wisconsin Medical School named Dr. [13:08] Warner Slack. And probably this is the [13:11] first course of its kind anywhere in the [13:14] world. I mean, computer science [13:16] departments themselves are new. The idea [13:17] of applying computer science and [13:19] computers to the practice of medicine is [13:21] new. Yep. It had to be one of the first [13:24] classes of its kind because mainframes [13:26] were really becoming a thing in the 60s [13:29] period. I think any was only 15 20 years [13:34] before. Yeah. Right. It's a good point. [13:36] This is the first time anybody could use [13:38] computers for anything outside of [13:40] government and the defense industry. [13:42] Right. Yep. So Judy takes this class [13:45] from Dr. Warner Slack and as you might [13:48] expect, she's the star student in the [13:49] class. Afterwards, Dr. Slack asks her to [13:54] work on writing a program for use in the [13:58] hospital in the medical school to [14:00] optimize the on call schedules for [14:02] doctors. You know, doctors have to have [14:03] on call schedules and so optimizing that [14:05] like perfect thing that a computer [14:06] application could do. Y So Judy says, [14:08] "Yes, great." She starts working on it [14:10] and this is, you know, a part-time job [14:11] for her as a grad student. She's getting [14:14] paid $5 an hour for her programming [14:16] time, trading time for money here. And [14:19] apparently the story goes that her [14:21] programming was so good and she was so [14:24] efficient at writing these applications [14:25] because she would go on you know lots of [14:27] departments in the hospital would ask [14:28] her to write you know various [14:29] applications for them that she actually [14:31] didn't make that much money because she [14:33] just wrote them so fast and she was [14:35] getting paid by the [14:37] hour. They gave her a raise at one point [14:39] they doubled her salary to 10 bucks an [14:41] hour and still didn't make that much. [14:42] Anyway, so as she's going around [14:45] throughout the Wisconsin medical [14:47] department, she's working with [14:48] psychiatry and obgyn and rehab and [14:52] inatient in the hospital and the [14:53] intensive care unit. She starts to get [14:56] these requests from all the different [14:59] doctors and the different practices. A [15:02] big problem we have is we're seeing [15:05] these patients, other departments are [15:06] also seeing the same patients, but [15:08] there's no way for us to know what's [15:11] happening to those patients in other [15:13] departments across the hospital as [15:14] they're being seen. Like we really would [15:16] love if there was a single database that [15:19] could keep records on every patient that [15:22] we have across the whole longitudinal [15:25] course of their care at the University [15:28] of Wisconsin Medical Center. you know, [15:30] an electronic sort of health record for [15:32] these patients, you might [15:34] say. And this is the origin of Epic. But [15:38] before we go further on what Judy does [15:40] next, I think it's worth taking a step [15:42] back and talking about what are health [15:44] records, what are medical records, what [15:46] was the state of them in America here in [15:48] the mid60s, and why is this idea of an [15:50] electronic version of them so appealing? [15:53] So patient health records at least here [15:56] in the American medical system were a [15:59] rag tag informal process and they were [16:03] all paperbased of course and I think [16:05] they had dated back to attempts at [16:07] unifying them or creating standards all [16:09] the way in to 1912. Yes. At Mass [16:12] General. So all of the efforts to [16:16] standardize them and you can imagine all [16:18] the reasons why standardizing them is [16:19] important. You know, patients see [16:20] different doctors at different hospitals [16:22] and they move and even within a hospital [16:24] like here at the University of [16:25] Wisconsin, cardiology really wants to [16:27] know what has happened to this patient [16:29] in other departments that they've been [16:30] in before, etc. Right. Or if you have to [16:32] see a different doctor for the same [16:33] thing, it'd be great if they're calling [16:35] the same condition the same condition. [16:37] Exactly. So all of the efforts around [16:40] standardization in the country really go [16:43] back to Massachusetts General Hospital [16:44] in Boston which Massgen or MGH is the [16:48] main and largest teaching hospital of [16:50] Harvard Medical School. It would make [16:52] sense that this is where a lot of sort [16:53] of research is happening. And like you [16:55] said, in 1912, a few members of the [16:57] American Medical Association and the [16:59] American Hospital Association get [17:01] together there at Mass General in [17:03] Boston. And they start working on can we [17:05] create some standard entry practices for [17:11] physicians that we can distribute across [17:13] the country that physicians can use as [17:14] their patient diaries as they're known [17:17] back in the day. So in 1919, it takes [17:19] them seven years to do this, they do [17:21] finally introduce a standardized [17:24] treatment diary, quote unquote, for [17:26] distribution across the country. Now, [17:28] there's no incentive for doctors to [17:31] actually use this. And doctors, as we [17:34] shall see, are notoriously sort of a [17:37] independent-minded group and profession [17:39] and don't like being told what to do or [17:41] how to design the notes that they take, [17:43] which makes sense. They've been the [17:45] smartest people their whole life. [17:46] They're doing this thing that required [17:47] an incredible amount of education. It's [17:49] a very prestigious, high-paying job. [17:51] Totally. And I think there's a good [17:53] argument for this here, too. The [17:54] practice of medicine, especially then [17:56] and still now, is equal parts art and [17:59] science. Telling me how to standardize [18:02] what is going on with my patient, [18:04] probably seemed anathema at the time. [18:06] I'm the doctor. I know best. I'll figure [18:08] out what delivers the best care. Yep. [18:10] Exactly. So, this continues for a while. [18:12] And then finally in 1928, the [18:15] Association of Health Record Librarians [18:17] of North America gets set up as an [18:20] official body to standardize collection, [18:22] storage, retrieval of patient data. This [18:25] is a fun call back. I believe the [18:28] Rockefeller Foundation was behind [18:30] funding all this. Call back to our [18:31] Standard Oil episodes. I believe that [18:34] because wasn't our conclusion on [18:35] Standard Oil that the Rockefeller [18:37] Foundation is the initial funer of the [18:40] nation's medical schools? Yes. Or sort [18:43] of kicked off the movement of having [18:45] real medical schools. One of, if not [18:47] maybe the main goal of the Rockefeller [18:51] Foundations was to improve the state of [18:53] medical care in America. Makes total [18:55] sense that oh, here in the late 20s, [18:58] early 30s, this is getting funded by [19:00] them. So all this is great but the [19:03] reality is until this time in the 1960s [19:06] when computers start to arrive even the [19:10] best intentions and even let's say [19:12] doctors really want to follow all this [19:14] as long as you're dealing with paper [19:16] records there's kind of a limit of how [19:18] helpful this can all be. You could [19:20] standardize it as much as you want but [19:22] you still got to get the reams of paper [19:24] from one physical location in a hospital [19:26] to another. God forbid you're trying to [19:28] get to a different hospital or a [19:29] different health system or a different [19:31] state. It's a mess. Not to mention, [19:33] paper doesn't lend itself well to [19:34] structuring data because at the end of [19:36] the day, you can write on paper however [19:39] you'd like to. So then we finally get to [19:41] the current moment in time that we're [19:42] here in the mid1 [19:44] 1960s and two really important things [19:47] happen. And you could argue that they [19:49] are both of equal importance. One, the [19:51] arrival of the computer age. You can now [19:53] digitize this stuff. It's possible to [19:56] have portability and standardization and [19:59] the dream can really be realized here. [20:01] Yep. The other maybe more important [20:03] thing that happens is in [20:05] 1965, Congress creates social security [20:10] and the Medicare and state-based [20:13] Medicaid programs in America. Ah yes. So [20:17] we have so far given you the [20:20] technological history of medical records [20:24] and now it's time to flip over to the [20:26] sort of policy side of things which in [20:29] medicine you have to understand policy [20:32] before you can understand the business. [20:34] All business structure in this industry [20:37] is driven by what is the architecture of [20:41] policy in America. So, we rewind the [20:45] clock again to 1942. [20:48] Oh, all right. Educate me here. We're [20:50] going to catch up to Medicare here. But [20:52] in [20:53] 1942, America is at war. There is [20:56] something passed called the [20:57] Stabilization Act that imposed wage and [21:01] price controls. An interesting thing to [21:03] note, at this point in time, only 10% of [21:06] Americans have health insurance. Period. [21:09] Yeah. Mostly the paradigm in healthcare [21:12] is you pay for service. You go to the [21:13] doctor, you pay the doctor, right? You [21:15] don't yet have these runaway costs and [21:17] well, I might have a procedure that [21:19] could randomly cost a million dollars. [21:21] You don't yet have these amazing [21:23] breakthrough treatments that could be [21:25] very expensive if you were able to [21:27] achieve them. So, you mostly just paid [21:29] out of pocket. So, you've got these wage [21:32] controls, but employers still want to be [21:35] able to attract top talent. So the way [21:38] around the wage controls was to offer [21:40] health insurance as a bonus to work [21:43] around the system. [21:45] Markets always find a way. Yeah. Right. [21:48] So as soon as this happens, unions go, [21:51] "Oh, this is awesome. Our people can [21:53] make more money. Amazing. We're going to [21:54] make capital pay for our health plans [21:57] now." Yeah. Exactly. Increase wages for [21:59] workers. This is what we're all about. [22:01] They lobbyed to make it explicitly legal [22:04] since it was kind of a gray area from [22:06] the wage controls. And so the National [22:09] War Labor Board makes it official. You [22:12] can do this. Two, employers and the [22:15] health insurance companies that are [22:16] springing up then lobby to say, "Hey, [22:19] these health insurance premiums, can we [22:22] make these taxdeductible for the [22:24] business?" because the business, they [22:26] shouldn't have to pay taxes on this [22:28] money if they're going out and buying [22:30] health plans for their employees with [22:32] it. And on top of that, even though it's [22:34] a form of compensation, can we make that [22:37] not taxable as income? That would be [22:39] really great. The businesses can deduct [22:41] it and the individuals shouldn't have it [22:44] affect their income taxes, even though [22:46] clearly compensation. So, this is where [22:48] the whole origin of pre-tax stuff versus [22:51] post tax stuff in your paycheck comes [22:53] from. Exactly. yet. So, this sets it all [22:55] in motion with incentives like that. Why [22:57] would you want your healthare any other [22:59] way? What a dream. You can pay for your [23:01] premiums with pre-tax dollars. And what [23:03] a dream for employers that can now offer [23:06] this benefit that was [23:08] deductible. So, at the end of the war in [23:10] 1946, the percentage of Americans who [23:12] had medical insurance was already up to [23:14] 30%. From that 10% number just 6 years [23:17] before. Flash forward a couple decades, [23:19] 1964, it's now up to 80%. So of course [23:23] it was going to run the table and become [23:24] the default as soon as this regulatory [23:27] framework was set up. And once we did [23:28] this, the US did not have a chance of [23:30] implementing any other system. We were [23:32] just going to end up with employers [23:34] primarily being responsible for health [23:36] insurance and creating large insurance [23:38] companies to provide it. Yep. And then [23:40] of course that takes us to 1965 and [23:42] Medicare and Medicaid because the big [23:44] policy question is now what do you do [23:46] about all the people who don't have [23:47] jobs? What do you do about people who [23:48] don't have jobs, which actually aren't [23:50] that many, but what do you do about [23:51] people who are old or poor? America [23:54] needs a social safety net. It's what [23:55] we're all about. Who also, by the way, [23:57] are the highest consumers of healthcare. [23:59] Yes. So, here we are at 1965. David, [24:02] you're exactly right. Medicare and [24:04] Medicaid enter the picture as a part of [24:06] the Social Security Act. How did we get [24:09] here is sort of interesting. FDR, [24:10] Truman, and Kennedy had all tried to [24:12] pass a singlepayer system the way that [24:15] the UK did right after World War II. The [24:17] UK is in this moment of sort of great [24:19] nationalistic pride. We all have to band [24:21] together, look after one another. They [24:23] pass the national health system. The US [24:26] does not have the political will to do [24:28] it. And so we don't. You sort of have [24:30] this coverage gap to your point earlier [24:32] of all these people who are not [24:33] currently in the workforce. So the [24:36] compromise is we will create Medicare [24:38] for those over 65 and Medicaid if you [24:41] have a low income or other special [24:43] situations and we can't really fund [24:46] either exactly right. So we'll sort of [24:48] figure out how to fund Medicare out of [24:51] the federal government and Medicaid. [24:54] Let's make that the state's [24:55] responsibility. We'll help but we'll [24:57] federate that out to the states. Yep. So [25:00] all of that is our system today. private [25:03] insurers if you have a job, buying [25:06] directly from the ACA if you don't, but [25:09] again still from private insurers, [25:11] Medicare or Medicaid if you sort of fall [25:13] outside those buckets. And the important [25:14] thing to realize from all this is that [25:16] the vast majority of patients do not [25:19] feel the cost of their healthare [25:21] directly in the United States. Those [25:23] costs are so laundered through private [25:25] insurance companies and Medicare and [25:27] Medicaid that most people think about [25:29] any given health encounter as being paid [25:32] for by someone else by a part of some [25:34] system. If you're trying to unpack how [25:37] did our healthcare become 18% of GDP [25:39] versus 11% of the UK GDP or a staggering [25:43] 6% of Singapore's GDP, albeit at a much [25:46] smaller scale. Why are we 18%? A big [25:50] thing you have to understand is [25:52] psychologically every healthcare [25:53] encounter is that the system is paying [25:56] for it. I'm paying into the system. The [25:57] system is paying for it. But what does [25:59] it cost? What do I actually pay? It's a [26:02] big abstraction. Yep. Okay. So, in case [26:06] it's not glaringly obvious, why is this [26:09] so important to medical records and the [26:12] fledgling electronic medical record [26:14] industry here? Well, it's because now [26:16] with these second and third party payers [26:18] set up, you need documentation of what [26:21] happened in order to get paid. So, if [26:23] you're a hospital or you're a doctor or [26:25] you're a clinical practice before this [26:27] system, you just see your patients and [26:29] they pay you. Now, after this system, [26:33] you see your patients and then you need [26:36] to negotiate with the payer, whether [26:38] that's the insurance company or the [26:40] government about getting paid for that. [26:42] And the insurance companies and the [26:43] government, they're like, "Well, hey, I [26:45] need proof. I need documentation of what [26:47] you did. I'm not going to pay you if you [26:48] just tell me that you did this stuff and [26:51] I need it in like a really standardized [26:54] form." Yep. I need an official [26:58] standardized medical record. And so all [27:02] of a sudden you start this Fouian [27:04] bargain or slippery slope if you will [27:06] for the medical profession of okay in [27:09] order to get [27:11] paid and get paid more and more over [27:15] time as my procedures that I'm doing [27:17] become more and more complex I need [27:20] strictly codified and regulated and [27:22] standardized documentation of what I did [27:26] and I need really bulletproof workflows [27:29] and data flows between what's happening [27:32] in the patient room and then what bills [27:36] get sent over to the payers and thus the [27:38] existential need for an electronic [27:41] medical record system. But really also [27:45] and maybe more importantly an electronic [27:47] medical billing system. Yep. And the [27:50] realization that I think the whole [27:52] industry kind of had early is insurance [27:56] wants to pull together in a very large [27:58] pool. So the payers are always going to [28:02] have a tremendous amount of leverage [28:04] over individual physicians, small [28:07] hospital systems, even larger hospital [28:09] systems. So you really need to be [28:12] extremely buttoned up and extremely [28:14] standardized and extremely auditable [28:17] because you're negotiating with this [28:20] large counterparty, whatever the given [28:21] payer is, be it a government or a large [28:23] insurance company, that's going to have [28:25] leverage over you. Yep. And for the [28:27] hospitals in terms of quote unquote [28:29] customers, I mean they're not customers, [28:31] they're payers. The reason they're [28:32] called payers is they're not getting the [28:34] delivery of care. They're just paying [28:36] for it. Of that set of payers, Medicare [28:38] and Medicaid are the big gorillas [28:40] because they're covering the elderly [28:43] population who are consuming the vast [28:45] majority of care and the vast majority [28:46] of complicated care and expensive care [28:49] in the country. Makes sense. So here we [28:52] are in 1966, the year after Medicare and [28:56] Medicaid get created by the government [28:58] and there's this existential reason now [29:01] for healthcare to adopt systematized [29:03] records. And so a group once again at [29:05] Massachusetts General Hospital in Boston [29:08] starts work on the first real [29:10] computerized medical record system [29:12] called Co-Star, computer stored [29:14] ambulatory record. And it's used for [29:17] patient scheduling and registration and [29:19] clinical data and everything you would [29:21] think in an electronic health record, [29:22] but also critically billing and [29:25] interfacing here with the payers and [29:27] development for that was funded by the [29:28] National Institute of Health. The [29:31] technical requirements for creating this [29:34] though are actually pretty difficult [29:37] especially with programming languages [29:40] available at the time like forran or [29:42] whatnot. This is a very high concurrency [29:47] transaction system. A lot of users need [29:51] to use this across a health system. Even [29:53] then, even then, yeah, a lot of users [29:55] across a health system need to use this [29:58] and it needs to interface with a lot of [30:00] endpoints and especially with the [30:02] limited storage and processing power of [30:05] computers like mainframes at the time. [30:07] The group of programmers that were [30:09] working on this found that existing [30:12] programming languages couldn't really [30:14] suit their needs to build what they [30:16] needed. So they end up writing their own [30:19] new programming language called the [30:20] Massachusetts General Hospital utility [30:23] multi-programming system or MS [30:28] which is amazing that there's a medical [30:29] oriented programming language called MS. [30:32] Yes. And epic still uses mumps today or [30:36] actually it uses its modern descendant [30:38] cache but this is the standard [30:41] programming language for the industry [30:44] and database system for the industry. Uh [30:46] I don't know about the industry but for [30:48] Epic. Oh yeah, which I guess now for the [30:49] industry. Yeah. Well, that's what I was [30:51] going to say. Many of Epic's competitors [30:53] do not use this but Epic does. So, uh [30:55] it's kind of deacto the standard now. So [30:58] now there are two key features of MS and [31:00] this is why these programmers had to [31:01] create their own language. One, the [31:04] language and the database are [31:07] integrated. Right there within the [31:08] programming language is a database [31:11] structure and that makes it very [31:13] computationally efficient for handling [31:15] all these high velocity and data [31:18] inensive transactions that need to [31:19] happen or at least efficient for the [31:21] programmer because you're not switching [31:24] over to write SQL queries in the middle [31:26] of your program. Everything is sort of [31:28] in the same language. If what you're [31:31] doing is primarily building a wrapper [31:33] around a database to use modern [31:35] parlament, a glorified database that is [31:37] going to constantly being read and [31:40] written to. Kind of nice for it to all [31:42] just be one standard language. Yep. And [31:44] then the other thing that they design it [31:46] for is for multiple simultaneous users. [31:49] Imagine you've got different departments [31:51] within a hospital updating the same [31:53] patient record at the same time. You've [31:55] got then administrators in the hospital [31:57] also updating that record to know what's [31:59] been build out to a payer, what's not [32:00] been, what's been challenged, etc. [32:02] Right. You don't want collisions to [32:04] create data loss or something like that. [32:05] This is people's lives we're dealing [32:07] with. Exactly. Now, the lead programmer [32:11] of this group here at MGH at Massgen [32:14] that is creating co-star and MS was a [32:17] recent MIT graduate named Neil [32:21] Papalardo. And being the young [32:24] enterprising MIT grad that he is, a [32:27] couple years later in 1968, Neil spins [32:29] out and starts a company around this to [32:32] sell his software originally called [32:34] Medical Information Technology or MIT. [32:38] Get it? Name's taken. Yay. But soon he [32:41] changes the name perhaps at the request [32:43] of the real MIT to Metatech. And for [32:46] folks in the industry, that'll sound [32:48] very familiar because Metat is still in [32:50] business today and is the number three [32:51] player in the EMR space behind Epic and [32:54] Cerner. Yep. Pretty amazing. So, back to [32:58] Judy now here at the University of [33:00] Wisconsin. She's working with Dr. Slack [33:02] and the medical department there doing [33:04] all these various application projects. [33:06] And then Dr. Slack moves to Harvard [33:08] where he of course meets Neil and starts [33:10] working with Metatech. But he and Judy [33:13] keep in touch. Judy's part of this [33:14] community of programmers and computer [33:16] scientists building applications for [33:18] healthare systems. And like we said [33:21] earlier, what the people at Wisconsin [33:23] really want is the same thing that the [33:25] people at Harvard and MGH really want is [33:28] this integrated system where they can [33:29] get longitudinal patient records across [33:32] the continuum of care and that they can [33:34] use to bill the payers. The holy grail. [33:37] The holy grail. So through this [33:40] connection, Judy learns about MS and she [33:43] goes and learns MS, the programming [33:44] language, starts working on this problem [33:47] and legend has it that one day in the [33:50] mid 1970s, she's sitting in her living [33:52] room and has an epiphany about how she [33:56] is going to build a great system, a [33:59] single database that can do all of this. [34:02] And her quote on this is, "The sun was [34:04] shining. I was disattentive. I was just [34:06] sitting there and suddenly it all came [34:08] to me. Here's how you build it. The [34:10] integrated system. And I remember [34:12] running to the kitchen, grabbing a pad [34:14] of paper, and just writing code, code, [34:15] code, code code. And that code became [34:18] Chronicles. She called it Chronicles. [34:20] It's a chronicle of a patient's care [34:24] journey. And that code in that database [34:27] is still the core of Epic to this very [34:30] day. Is it actually the same code? No, I [34:33] mean she'll claim none of my code that I [34:34] wrote is still in production, etc. That [34:36] I'm sure that is true. But it is true [34:38] that Epic's core database, the core [34:41] single database, and this is Epic's big [34:43] differentiation. There's only one [34:45] database that every application pulls [34:48] from directly talks to. It directly [34:50] talks to whether it's the clinical side [34:52] for EMRs, whether it's the billing side [34:54] with the resolute module that Epic has, [34:56] whether it's Cosmos, whether it's [34:59] Storine or Beans for Kidney or whatever [35:01] application that Epic makes of their [35:03] hundreds, it all pulls from one single [35:06] database in Chronicles. And that's what [35:08] Judy writes here in the 1970s. Okay, so [35:11] it's just a database. Where does it go [35:13] from she writes a bunch of code when she [35:15] has this insight that there should be a [35:17] database with the patient model at the [35:21] center? Well, that's a good question [35:22] because it is just a database because [35:26] the idea of applications and certainly [35:28] graphical applications on top of it [35:30] doesn't make sense because we're still [35:31] in the mainframe world here. So what the [35:34] product is is this database and then [35:37] different departments in the Wisconsin [35:40] Medical Center can write their own [35:42] screens like queries on top of the [35:45] database that can sit on their terminals [35:47] and that they can read data out of the [35:49] database directly into their patient [35:51] rooms or departments wherever they're [35:54] sitting. So these are terminals. I mean [35:57] these are 80 character wide textonly [35:59] terminals. Yes. Green screens Unix [36:01] terminals. Okay. Okay. So, you run [36:03] Judy's database, you run Chronicles [36:05] somewhere in your hospital building on a [36:07] big mainframe, and then there's these [36:09] textonly terminals that can query it. [36:11] Yep. That's the product. Sweet. Long way [36:14] to go. So, just like the projects in [36:17] Boston at MGH were the intention was to [36:21] make the systems for use there at MGH, [36:24] Judy at first, she's just making [36:26] chronicles for the Wisconsin medical [36:28] system. She's just a programmer employee [36:31] of the medical center. But the doctors [36:34] at UW, they're going to academic [36:36] conferences all around the country and [36:38] they start telling other doctors at [36:40] other academic institutions about this [36:42] great system called Chronicles that they [36:44] have that this programmer that they [36:46] employ, Judy Faulner, has written and [36:49] demand for it starts to spread virally. [36:51] and Wisconsin gets calls and then Judy [36:53] gets calls from all these other health [36:55] systems around the country that are [36:56] like, "Oh, hey, can you write this for [36:58] me, too?" Doesn't she get asked to start [37:00] a company several times and she's like, [37:02] "No, I I just made it this one instance. [37:05] We don't need to build a whole company [37:06] around this thing." Exactly. And this [37:08] happens again and again and again and so [37:10] Legend is finally she just breaks down. [37:11] It's like, "Fine, I'll start a company. [37:14] I'll be part-time on it. We'll get some [37:15] other people who are working with me. [37:16] We'll be part-time. This will be a small [37:18] little thing." And so thus finally in [37:21] 1979 the company is born human services [37:26] computing. Amazingly generic. The future [37:30] Epic Systems. But before we tell the [37:32] story of how human services computing [37:35] became Epic. Yes. This is a great time [37:38] to thank our presenting partner JP [37:41] Morgan Payments and in particular to [37:43] dive into how essential modern payments [37:46] infrastructure has become to building [37:48] your business today. Whether you are a [37:50] large multinational business or a [37:51] startup in your early days of processing [37:53] payments. Yep. If you think about it, [37:56] there are whole companies and industries [37:57] that couldn't exist a decade ago without [37:59] the payment tools we have today. It's [38:01] absolutely essential to create [38:02] businesses with modern product [38:04] experiences like ride sharing, the [38:06] creator economy or B2B use cases like [38:08] SAS marketplaces or managing supplier [38:10] relationships. For those types of [38:12] companies, payments is their business. [38:14] Thankfully, JP Morgan has built the [38:16] stack to let them focus on their [38:18] differentiation instead of reinventing [38:19] the wheel. JP Morgan Payments has been [38:21] pioneering in this industry for decades. [38:24] They move $10 trillion a day. You can [38:27] literally never outgrow their [38:29] capabilities. Yes. So while we are [38:31] examining the healthcare landscape [38:33] today, let's look at the industry [38:35] through the lens of payments. There's a [38:37] lot of innovation promise coming with [38:39] teleaalth and AI and preventative [38:41] treatment and new clinical trial [38:43] processes. Seamless and secure payments [38:45] are critical to improving patient [38:47] experiences in unlocking innovation for [38:50] businesses and providers. And when you [38:52] zoom out, healthcare is an incredibly [38:54] complicated ecosystem of payments, [38:56] health care providers, insurance [38:57] networks, specialists, health monitoring [38:59] services, and more. It creates a complex [39:02] and friction-filled payment experience. [39:04] Who's paying who and when and under what [39:06] terms, and then you layer data privacy [39:08] requirements on top. It's wild. If [39:11] you're a company or a provider trying to [39:12] innovate in this space, you know getting [39:14] the payments piece right is paramount. [39:16] Which is why JP Morgan's array of [39:17] products, including their healthcare [39:19] payment solutions and Instamad offering, [39:21] provides a patented cloud-based [39:22] technology to securely transform [39:24] healthcare payments by driving [39:26] electronic transactions, processing [39:28] payments, and moving healthcare data [39:29] seamlessly. If you're in the industry, [39:31] you likely know that JP Morgan is [39:33] finally attuned to all these innovations [39:35] on the frontier, hosting the premier [39:37] healthcare industry event every January. [39:39] If you learned anything at this year's [39:41] event that's relevant to this episode, [39:43] come share it with us in the Slack. [39:44] Listeners can check out [39:46] jpmorggan.com/acquired to learn about [39:48] how JPMorgan's endto-end payment [39:50] solutions can accelerate your business [39:52] and discover more innovation happening [39:54] across all [39:55] industries. All right, so David Judy [39:58] leaves her job. She starts what is it? [40:01] Human computer [40:04] human services computing. Human services [40:06] computing really rolls off the tongue. [40:08] Yes. She's not planning up till this [40:10] point to be the founder of a company. [40:13] And yet for the next 47 years of her [40:14] life, she would be. How does that go? [40:17] Well, before she does anything, she [40:20] needs to raise some money. For not the [40:21] least of which to buy a computer to do [40:25] this work on, which wasn't it like 70 [40:27] grand to buy a computer? Yes. A computer [40:31] being a data general Eclipse 16bit mini [40:35] computer, which is the size of like a [40:37] washerryer system and sits in the [40:40] basement. That's right. cuz we're in [40:41] this awkward era between mainframes and [40:44] microcomputers. The microprocessor [40:45] hasn't been invented yet and so we have [40:47] these quote unquote mini computers that [40:49] are still Yeah. washing machine sized. [40:51] Yeah. We're right before the PC era [40:53] here. It's so interesting how sometimes [40:55] you have the technology wave that [40:58] eclipses all prior, but sometimes you [41:00] have these half steps along the way that [41:02] turn out not to eclipse everything. [41:04] Yeah. Many computers, they were like the [41:06] netbooks of their time, you know. Yes. [41:10] We talked about this a lot on our [41:11] Microsoft episodes, but the critical [41:14] thing about minicomputers and especially [41:17] for duty here and fledgling epic was [41:20] look a consumer wouldn't buy these [41:22] things but small businesses could and [41:25] small departments of big companies could [41:27] like an IBM mainframe system was going [41:29] to cost you a lot of money. Yeah, a data [41:33] general or deck mini computer is going [41:35] to cost you about $70,000 and Judy needs [41:39] to buy one and she also needs to fund [41:40] the company. So she goes to the bank and [41:44] she gets a bank loan for $70,000 to [41:47] finance going and buying the data [41:49] general Eclipse mini computer. And then [41:53] she rounds up a bunch of friends and [41:55] family and other people, whether [41:56] physicians or other programmers in the [41:59] University of Wisconsin system who are [42:01] working with her on these projects to [42:03] invest money in the new company. And [42:05] together they all put in about [42:07] $70,000. They value the company at [42:10] $70,000. So pre- money valuation of [42:12] $70,000. Another 70,000 of new on top. [42:17] Yep. Post money of [42:19] $140,000. taking 50% dilution from your [42:22] 70k fund raise and then you also have [42:24] this [42:25] $70,000 loan. Yep. And those were the [42:28] only primary investors ever in Epic [42:32] Systems. And there's some really fun [42:35] stories about what happened to those [42:37] shares that the other original investors [42:40] bought over time. The company has bought [42:42] a lot of them back, but not all of them. [42:44] There are still some floating out there. [42:46] There are some floating out around [42:48] there. We can't share all the stories [42:49] that we heard in the research, but one [42:51] fun thing is at one point in time, I [42:53] think this was probably the 2000s, maybe [42:57] a pretty good chunk of those shares made [43:00] their way to Sequoia Capital. Yes, [43:02] Sequoia did not put that one on their [43:04] website. Almost the only way in which [43:06] Epic intersected with Silicon Valley in [43:08] the entire company's history. But this [43:09] is a crazy point. They do $5.7 billion [43:13] in revenue today. They dominate an [43:16] industry, at least here in the United [43:18] States, and in total they raise 70,000 [43:21] of equity capital and 70,000 of bank [43:23] debt. Yes. And that's it. This is IKEA [43:26] all over again. Totally. And I think a [43:29] huge part of what enabled this building [43:33] software is not a uh necessarily a [43:36] capital light activity. It is really [43:39] hard especially in those days to build [43:42] great software. I think kind of like how [43:44] Microsoft never really raised primary [43:47] capital either. It's because Judy was [43:49] got generational talent as a programmer. [43:52] I hadn't made that linkage. That's [43:54] interesting. They had kind of a cornered [43:56] resource as a startup. You know, [43:57] Microsoft did in Bill Gates and Paul [43:59] Allen too, but Epic totally did too in [44:01] Judy. Like she could write really, [44:03] really great software. They didn't need [44:05] to go hire an army of programmers. No. [44:07] And the company stayed very very small [44:09] for a very long time. Hm. Because as [44:12] legend has it, it's three halftime [44:14] people in the basement of a building [44:16] that started the company. And I think [44:18] they got pretty far on just that. I [44:20] think so. And for years, they didn't [44:22] hire that many more people. That would [44:24] obviously change over time. Okay. I'm [44:27] 90% sure. I'm going to stump you on this [44:29] one. You mentioned the basement of the [44:30] building. So, they get their first [44:32] office space. Judy gets the company's [44:34] first office space in the basement of an [44:36] apartment building at 2020 University [44:39] Avenue in Madison, [44:41] Wisconsin. Do you know what other great [44:46] American company from Madison, Wisconsin [44:49] also started in that same office space [44:51] and I believe overlapped. I think they [44:53] were both in this same shared office [44:55] space concurrently. David, there's only [44:57] so much material out there on this [44:59] company. We were both going to find [45:00] this. Listeners, this is the American [45:03] Girl Doll Company. Oh, I really thought [45:06] I had you on this. You know, Ben and I [45:08] don't do our research calls together. We [45:10] do most of them separately, but we were [45:11] on one together and the person we were [45:13] talking to was about to say this and I [45:14] was like, "No, don't say it. I want to [45:15] stump Ben on the episode." [45:17] I did see it somewhere else, though. The [45:19] crazy thing is, I think when American [45:21] Girl moved out, Epic bought some of [45:23] their furniture. Yes, I think that's [45:25] right. So listeners, what we're talking [45:27] about is the American Girl Doll Company [45:29] founded by Pleasant Roland in Madison, [45:32] Wisconsin. My sister had I think a few [45:33] of these growing up. Oh man, Jenny had [45:36] so many of these and a lot of them have [45:38] migrated now to our house and our girls [45:40] have them. And already my three-year-old [45:43] runs around our house with the American [45:46] Girl doll magazine and she tells me on a [45:48] weekly basis, "These are the ones I want [45:50] for my birthday." She's pointing at like [45:51] a $250 doll. Yeah. incredible company. [45:55] It actually ends up getting acquired by [45:56] Mattel in the late 90s. It was a big [45:59] part of Mattel for a long time. I'm so [46:01] bummed I didn't stump you on this [46:05] one. Okay, back to the epic story. So [46:08] Judy's got the financing, she's got the [46:11] computer, but she still doesn't know [46:12] anything about building and running a [46:14] company. You know, she's a software [46:15] developer. So Warner Slack says, "Okay, [46:18] come on out to Boston. I'm gonna set up [46:20] a couple days for you to spend time with [46:22] Neil and Metatech and I'm gonna ask him [46:25] to help you out starting this company. I [46:27] gotta say this is a pivotal moment and [46:29] not just for everything she learns from [46:31] Metitech, which I know you're going to [46:33] get into all these cool things that she [46:34] brought over, but instead of I'm going [46:37] to introduce you to a business person, [46:40] he says I'm going to teach you as a [46:42] programmer everything you need to know [46:43] about running a business. And it's not [46:46] true, but it kind of rounds to true to [46:49] say Epic basically never did hire any [46:51] business people. It is essentially a big [46:54] gigantic company of programmers, [46:57] logicians, implementation people who [46:59] could be programmers who think like [47:01] programmers. That is the DNA of the [47:04] company to this day. And it could have [47:06] gone a super different direction of you [47:08] need to go figure out a sales and [47:10] marketing strategy and a business plan. [47:12] And that is just not what happened. This [47:15] is a history turns on a knife point [47:18] moment for Epic and Judy because the [47:22] likelihood that she would get introduced [47:25] to a business mentor who is also a [47:27] software developer and not a business [47:29] school graduate business guy. We're [47:30] talking about the 1970s here. The, you [47:33] know, business playbook is you bring the [47:35] grown-ups in, right? Yep. Bring in the [47:37] suits, bring in the business school [47:39] graduates, you know, fire the founders. [47:40] Well, that only happens if you raise [47:42] capital. Well, right. [47:43] That's exactly. You wonder why Judy was [47:45] so uh, you know, averse to venture [47:47] capital. Everybody else she knew who was [47:49] taking it in that era was getting fired. [47:52] Swapped in for a business guy. Totally. [47:54] But yeah, Neil, he came out of MIT. He's [47:56] a programmer. He was on the original MUS [47:58] team. So Judy goes out, spends three [48:02] days with Neil and gets a total crash [48:04] course in setting up and running a [48:06] company. And Ben, like you said, Neil [48:09] ran Metatech like a software developer. [48:11] So all the processes were extremely [48:14] standardized. He had like hardened APIs [48:16] for running the company. You know, he [48:18] had manuals and documentation for [48:20] everything. And he shared the manuals [48:22] with Judy. This is how you set up a HR [48:25] system. This is how you do benefits, [48:27] payroll, how you do college hiring, how [48:30] you promote internally. Well, yes, [48:33] Metatech recruited from the universities [48:36] there in Boston. They didn't hire [48:38] experienced programmers. They were [48:40] hiring fresh college graduates. Totally [48:43] natural that Judy's going to do the same [48:45] thing at Epic out of first the [48:47] University of Wisconsin and then other [48:49] schools. But yeah, that was the [48:50] playbook. And that DNA runs right [48:52] through to this very day at Epic. It's [48:54] mostly people from non-medical majors. [48:57] They're hiring from Midwestern schools, [49:00] from people with technical majors, just [49:02] operating under the assumption, h you [49:04] can learn this healthcare thing. It's [49:06] the same thing as Microsoft or Google or [49:08] Meta or what have you. They are going to [49:11] universities, all departments and [49:13] recruiting kids on on campus recruiting [49:15] at colleges. Yep. And she was doing this [49:17] at a time when it wasn't the norm. I [49:20] mean, I got to keep drilling in how [49:22] different the world was then in a bunch [49:23] of different ways, but one of them was [49:25] you didn't have these college career [49:27] fairs where you would just assume you [49:29] could get this amazing high agency job [49:32] right out of college. Microsoft was [49:35] really on the frontier of hiring smart [49:39] college grads and empowering them and [49:40] letting them run free. Epic was doing [49:42] the same thing and this was not industry [49:44] standard. And Microsoft wasn't even [49:47] really doing it at this point in time. [49:48] They had only just moved to Washington. [49:51] They'd only just left Albuquerque. This [49:53] is the time frame we're talking about. [49:54] Yeah. What year are we in? We're in [49:56] 1979. So, Microsoft really isn't [49:58] starting to ramp hiring yet. No. And to [50:01] that point, there is no DOS yet. There [50:04] is no real PC industry. Microsoft is [50:07] still its main product is the basic [50:09] interpreter for Judy and the fledgling [50:11] human services computing here. Yes, they [50:14] have demand from other large academic [50:18] hospitals out there and medical systems [50:20] that have mainframes set up that have [50:22] university computing infrastructure that [50:24] they can leverage and use. But your [50:27] average hospital or medical clinic or [50:29] outpatient clinic out there in America, [50:32] they don't have a mini computer. They [50:33] don't have a mainframe. The market isn't [50:36] really there for this kind of stuff yet [50:38] in a big way, except at these university [50:40] medical systems. Yeah. And part of what [50:42] she is doing is going after a different [50:44] segment than Metatech was doing is a [50:48] little bit out of deference to Neil and [50:49] his company. They're kind of going after [50:51] the small hospital market. She's going [50:53] to take the more sort of upmarket [50:55] enterprisey approach and go after the [50:57] most complex institutions, these [50:59] academic training hospitals. Later on, [51:01] the IDN's or the integrated delivery [51:04] networks which are huge hospital systems [51:06] or children's hospitals. Again, the most [51:09] complex upmarket enterprise has the most [51:13] possible needs for the most complex [51:15] software. That is where she's about to [51:17] point the company. Yep. Which she has to [51:20] for the Unix based product that she's [51:23] making because those are the only [51:24] institutions that have Unix [51:26] infrastructure, right? That can afford [51:27] these big computers. Yeah. So 1979, they [51:31] start out with four initial customers. [51:34] Four years later in 1983 is when they [51:37] renamed the company to Epic Systems. [51:39] Ben, why did they rename the company? [51:41] Well, aside from the first name just [51:43] being an awful name, you can start to [51:45] see Judy's quirkiness come through. I [51:47] mean, I think you saw it originally with [51:49] Chronicles. Her sort of abstract [51:51] creative thought, the notion of an epic, [51:54] the Greek epic is this big story, this [51:58] big longitudinal historic event. And I [52:02] think the way that she's thinking about [52:03] a patient record is that the life of the [52:06] patient is an epic. Yep. Does that kind [52:08] of jy with your understanding? Yeah, it [52:11] definitely was not a declaration of [52:13] aspirations as a company because this [52:16] was still a very very small business. [52:19] They started with four initial customers [52:20] in 1983 when they changed the name. They [52:22] only have nine customers. And in fact, [52:26] by the end of the company's first decade [52:28] of existence, so end of 1988, they only [52:32] have 24 customers and it's only doing [52:34] $1.5 million in revenue and they've got [52:36] like a handful of employees. That's the [52:39] crazy thing. Unlike Microsoft, this was [52:42] a small business from the get-go, right? [52:45] It took them a decade to get to a [52:48] million and a half dollars in revenue. [52:49] Yes, that's kind of a slow growing [52:52] startup. a great software business for a [52:55] local entrepreneur in Madison, [52:56] Wisconsin, I think is how you would [52:57] describe that. Yep. So, yeah, for the [53:00] first 10 years of the company, you know, [53:01] yes, it's growing, but it's not exactly [53:04] setting the world on fire. And part of [53:06] the reason is the computing [53:09] infrastructure wasn't there at customers [53:10] like we talked about. The other big [53:12] reason why they weren't getting crazy [53:14] customer adoption is they actually were [53:16] only doing the clinical medical record [53:19] stuff at this point in time. They [53:20] weren't doing the really important stuff [53:22] of the billing system until 1987. So for [53:26] the first, you know, what is that call [53:27] it? Eight years of the company. They [53:29] weren't addressing the actual critical [53:32] problem in the hospitals, which is help [53:34] us bill for this stuff. Right. If you [53:36] can help them make more money, they're [53:37] going to be a lot more excited to buy [53:39] your software, pay a lot more for your [53:40] software. Yep. Totally. So in 1987, they [53:43] launch a billing module called Resolute. [53:46] again on top of the Chronicles single [53:49] core database which still today is the [53:52] company's revenue cycle application for [53:55] all their hospitals. And I know we've [53:57] made this point a few times but I really [53:59] want to underline. It is still just an [54:03] application built on top of the [54:04] Chronicles database. And fast forward to [54:07] today I think that is actually the [54:10] single biggest reason why Epic wins over [54:13] their competitors. H I would argue [54:17] there's two big reasons Epic wins today. [54:19] One is reliability. Yep. But the [54:22] reliability also comes from the fact [54:25] that it's all built on one database. [54:27] You're not gluing multiple systems [54:28] together. All their competitors along [54:30] the way or almost all of them other than [54:32] Metitech became 30 other companies glued [54:35] together through M&A and take public, [54:37] take private crazy transactions. Epic's [54:39] just been epic the whole time. So you [54:41] get this system that when you buy it and [54:44] they say it's going to take X dollars [54:46] and X time to implement it does and then [54:49] you go live and it works which sounds [54:52] crazy but it does the thing that they [54:55] say it's going to do on time and on [54:57] budget or as good as anyone does in this [54:59] industry. And then the second reason in [55:01] addition to reliability, David, as [55:02] you're alluding to, is the fact that the [55:05] clinical side and the billing side are [55:08] completely stitched together in one code [55:11] base working off of one database. It is [55:14] perfect harmony. You don't have [55:16] information dropped when one system is [55:18] talking to another. If your goal is to [55:20] adopt a system that keeps track of [55:22] everything in your hospital and make [55:23] sure that you can make money from it for [55:26] your organization, this is sort of the [55:28] ideal architecture for such a product. [55:31] It's funny, I'm laughing as you were [55:32] saying earlier of it sort of does what [55:34] it's supposed to do as an EMR. I assume [55:36] you're referring to the actual clinician [55:38] doctor facing EMR side of it. If you're [55:41] looking for a system as a hospital that [55:44] ties what happens in your medical [55:46] practice to your billing and your [55:48] revenue events, there is no if. That's [55:50] all you're trying to do as a hospital. [55:52] That is what you are looking for. That [55:54] is the product. Yes. And Ben, like you [55:56] said, all the competitors out there [55:58] except arguably Metatech, in many cases, [56:02] the billing system is a separate and [56:04] often separately acquired product from [56:08] the medical record. Not only is it just [56:10] like, oh, maybe not ideal if some of the [56:13] information doesn't pass quickly or [56:15] efficiently or accurately between those [56:17] two systems, it's like incredibly not [56:19] ideal. It's like you're not getting paid [56:21] for the work you're doing. Or even [56:23] worse, maybe you're submitting [56:25] documentation that is wrong, which is [56:27] like a federal crime, right? Or even if [56:30] it's bad in the other direction, then [56:32] it's not just that your billing system [56:34] is not making it into clinical, you're [56:36] potentially causing patient harm and [56:37] costing lives. any data flowing in [56:40] either direction that's bad is really [56:42] really bad in this particular use case. [56:46] So there's a great substack called [56:47] health API guy which I'm going to [56:49] reference a few times because it's just [56:51] some of the best writing on Epic you'll [56:52] find. He put it perfectly. Epic becomes [56:55] the natural choice for enterprise [56:57] decision makers precisely because of its [56:59] integrated system architecture. Rather [57:01] than managing multiple vendors and [57:03] systems, buyers get a comprehensive [57:05] platform with a single database, unified [57:07] workflows, and built-in operability [57:09] through care everywhere, which we'll [57:11] talk about later. But Care Everywhere is [57:13] the magical button that makes it so that [57:16] your hospital records at one hospital [57:19] are easily integrated and viewed in any [57:21] other hospital that is of course also an [57:24] Epic customer. Yes. Okay. So after they [57:27] launched Resolute and now we're in the [57:30] late 80s into the early 90s, we're [57:32] finally entering the PC era. It's now [57:34] possible for hospitals and health [57:36] practices to adopt computers for a few [57:39] thousand dollars instead of a few tens [57:41] of thousands of dollars. The market [57:43] really starts to take off for Epic. And [57:47] on the back of that in 1992 they launch [57:50] Epicare which I believe is the first [57:52] graphical user interface like [57:55] Windowsbased EMR application in the [57:58] entire industry. Certainly Epic [58:00] definitely says that that is true. Yep. [58:02] So what does that mean? Go back to what [58:04] Chronicles was originally. It was [58:06] terminal access directly into a Unix [58:08] based database on a mainframe or on a [58:11] mini computer. This is not something [58:13] that your average doctor or nurse or [58:15] medical assistant is going to use. Now, [58:18] here in the 90s with Epicare, they've [58:20] created a Windows application that any [58:24] PC user can fire up on their Windows [58:26] machine and use a graphical interface [58:29] for their patient interactions in their [58:32] EMR. And oh yeah, by the way, it's tied [58:34] directly into your billing system for [58:36] your hospital or your health practice. [58:38] This is all of a sudden really [58:40] compelling. We're approaching holy grail [58:42] here if you're a hospital administrator [58:44] especially with scheduling in addition [58:47] to billing their scheduling thing is [58:48] called cadence but once you have billing [58:51] scheduling and epicare handling the [58:54] actual clinical part of it and this is [58:56] all ambulatory right we haven't gotten [58:58] to impatient yet yes so the initial [59:00] launch of epicare the guey EMR [59:04] application was only for ambulatory only [59:07] for outpatient settings not overnight [59:09] stays yes not overnight stays in the [59:11] hospital. Yeah. As a non-healthcare [59:13] person, that's always how I define what [59:15] impatient is. Yes. You hear ambulatory, [59:18] you just hear non-oververn stays. Yes. [59:21] Too confusing otherwise. But this is it. [59:23] Scheduling, billing, and Epicare on the [59:25] clinical side for ambulatory. Pretty [59:28] amazing product that is trajectory [59:30] changing for the company. It's now got [59:33] enough functionality that it's not going [59:35] to stay a small business for long. Yep. [59:37] So by 1995 they hit 18 million in [59:41] revenue up from 1.5 million in 1988. So [59:46] pretty phenomenal seven years there [59:49] where they more than 10xed. So by most [59:52] measures you'd look at this and be like [59:53] business is going well. Yes. No longer a [59:56] small Madison Wisconsin business. Right. [59:59] However we were comparing it before was [60:01] Microsoft. You made the Bill Gates [60:02] comparison. This company was started [60:04] when was it? 70 79. So only a couple [60:07] years after Microsoft, four years after [60:09] Microsoft. Yep. In 1995, Epic did $18 [60:13] million. In 1995, Microsoft had gone [60:17] public and shipped Windows 95 and did 6 [60:20] billion in revenue. [60:22] Right? Very different trajectories. You [60:24] know, the primary reason here is one is [60:27] a small vertical focused healthcare [60:30] company and one is creating the [60:32] horizontal platform of the future. But [60:35] it's worth contextualizing businesses [60:38] going well by making sure that we stop [60:40] making the Microsoft comparison from [60:41] here on out. Well, I think I want to [60:44] come back to it later in the episode and [60:46] in analysis because yes, Epic isn't [60:50] always will be constrained by being a [60:51] vertical software provider instead of a [60:54] horizontal software provider like [60:56] Microsoft or Google or Oracle or what [60:58] have you. Vertical being one industry. [61:01] Vertical being one industry. However, [61:04] the one industry that they operate in is [61:06] 18% of American GDP. So, how big can [61:09] this get? Is well, it's a question we'll [61:10] revisit. Yes. Okay. So, then finally in [61:14] 2001, they launch the inpatient version [61:17] of Epicare. So, yes, this is for [61:20] inpatient hospital stays overnight. Now, [61:23] you finally have the holy grail. You've [61:26] got Chronicles as the one single [61:27] database. You've got Epicare ambulatory [61:31] for all your outpatient clinics. You've [61:34] got Epicare inpatient for all your [61:36] inpatient activities. By the way, again, [61:38] inpatient activities are probably going [61:41] to be your majority revenue stream [61:43] because that's where the most expensive, [61:44] most complex care is happening. Kind of [61:47] similar to Medicare is your most [61:50] important payer relationship because old [61:52] people is where the most complex, most [61:54] expensive care is happening. So you've [61:56] got that and then you've got Resolute, [61:58] the billing system, all single database, [62:01] all tied together, all built on top of [62:03] it. Yeah, if you're a hospital system [62:06] administrator, this is the best thing [62:08] you could possibly imagine. And Epic is [62:11] really starting to get religion around [62:12] this point in our future is breadth. Our [62:16] customer, these hospitals do not want to [62:18] buy peacemeal solutions. They want to [62:20] buy everything from one vendor and they [62:22] want that one vendor to provide the very [62:24] best most integrated experience [62:26] possible. So we need to continue to [62:28] orient the company around that [62:30] philosophy. Yep. Now right around the [62:33] same time actually before Epicare [62:36] inpatient launches, Epic also launches [62:40] my chart. This is crazy. If you had [62:42] asked me before doing research for this [62:44] episode, when would I guess that my [62:46] chart launched? This is a [62:49] internetbased [62:51] consumerfacing medical records access [62:54] and interaction platform on the web in a [62:58] highly regulated HIPPA regulated [63:00] industry. Right? I would have guessed [63:02] like I don't know 2010 maybe mid 2000s [63:05] at the earliest. No, Epic launched this [63:08] in the year 2000 which is wild. Right. [63:11] Right around the dot bubble. Yeah. [63:14] Crazy. It's absolutely wild that they [63:17] launched this and that their customers [63:19] launched this. Well, and it's incredibly [63:21] innovative. I mean, it is truly cutting [63:23] edge. I mean, it wasn't 1994 like [63:26] Amazon, but the fact that Yeah. to your [63:28] point, the medical record thing was [63:31] being surfaced on the web. That's [63:32] pioneering. Yeah. That consumers had [63:34] direct access to. Well, I want to come [63:36] back in a minute to how important this [63:39] is for Epic and their customers. But the [63:42] origin story of my chart is kind of fun. [63:44] So it actually started as an outgrowth [63:48] of what was called epic web in 1997 [63:52] which was a project that they were [63:54] working on for remote access to the EMR [63:58] to Epicare for doctors at home. So the [64:01] idea was like, oh, you're a doctor. You [64:03] go home, you wake up in the middle of [64:05] the night, you're thinking about a case, [64:07] and you want to check the medical [64:08] records. Maybe you want to update. Some [64:10] foreshadowing here of how doctors are [64:12] going to spend their time in the future. [64:14] Yeah, exactly. Little did they know how [64:16] much doctors would hate what we're [64:17] saying here, but you want to be able to [64:19] access remotely from your home the [64:22] medical records of your patients. So [64:24] they start working on Epic Web. And a [64:26] young right out of college programmer is [64:28] working on this by the name of Sumit [64:31] Rama. Sumit today is the president of [64:33] Epic. And he goes to the then president [64:37] of Epic, Carl D'vorak, and says, "Hey, [64:40] you know, this is good that I'm working [64:42] on this, but I'm kind of bored. Can you [64:43] give me something hard to do?" Which [64:45] this says a lot about the culture of [64:46] Epic. This is a early career programmer [64:50] going directly to the president, and [64:53] that says enough on its own. Yes. And [64:54] the president is running this patient [64:56] web project, right? Is kind of a [64:58] computer architecture person himself. [65:00] And the conversation is about I want a [65:02] more interesting project. Can you give [65:04] me something more cutting edge to work [65:05] on? Yeah. Give me something more [65:07] challenging and out of that is born the [65:10] initial idea for my chart. I think they [65:13] started working on it in 1998 and then [65:15] launched it in the year 2000. So it [65:17] becomes the first integrated patient [65:18] portal. And I mean truly this is a very [65:21] fair thing to say. This is insanely [65:23] innovative. For as many reasonable barbs [65:26] get thrown at the company, some around [65:27] oh technology and oh the UI is cluji and [65:30] all this stuff. Come on. My chart 99 is [65:33] real cutting edge. People immediately [65:35] got it. Once you could access your own [65:37] medical records from home from your own [65:39] computer directly without talking to [65:41] anyone, you were never going back. The [65:44] world changed overnight. It was a little [65:45] bit like Zillow in real estate. Yes. As [65:48] soon as you can look up how much homes [65:50] in your neighborhood and your friends [65:51] and neighbors homes sold for, you're [65:53] never going to go back to not being able [65:55] to do that. We're going to have some [65:56] rich debate later about the pros and [65:58] cons of EMRs and sort of are we better [66:01] off today than we were. But I just can't [66:03] fathom being in a world where I don't [66:06] have a way to access other than going to [66:08] the physical building and asking them [66:09] for my records or placing a call and [66:10] asking for them to call me back. Oh, [66:12] records used to have to get faxed back [66:14] and forth when you moved or changed [66:16] providers. It was brutal. But the other [66:19] really compelling use case for my chart, [66:23] especially when it first launches and [66:24] even through to today, is managing [66:27] family members care. you have elderly [66:30] parents who are using the health system [66:33] as you do as you get older and you need [66:36] to help manage that but you don't live [66:37] in the same city or even if you do [66:40] having my chart access to family members [66:42] was huge and starts this whole patient [66:45] side virality now for the company and [66:49] then for their customers I imagine [66:51] initially it was very scary to roll this [66:53] out once they add self-cheduling into my [66:58] chart [66:59] This becomes the greatest thing for [67:02] hospitals. I mean the workflow savings [67:05] of calls that had to happen to [67:08] scheduling appointments are huge. The [67:10] other big thing is no shows before my [67:13] chart and self-scheduling and the weight [67:15] list that my chart manages for patients. [67:18] If a patient nosh showed that was lost [67:20] revenue for you as a hospital system. [67:23] That was a big hole in your daily [67:25] revenue operations. Yep. And now you can [67:28] automatically fill that in with another [67:29] patient on the wait list. Yep. So we're [67:32] foreshadowing this, but everyone adopts [67:34] this basically. Today there are 191 [67:36] million active users of My Chart and [67:39] this is dduplicated. This is active [67:41] unique users of My Chart today. Yep. So [67:45] on the back of this they launch My Chart [67:48] in 2000, Epicare and Patient in 2001. [67:51] The company crosses 50 million in [67:53] revenue. they feel like, okay, we're [67:56] finally like ready for the big leagues. [67:58] And just to contextualize that 50 [68:00] million of revenue, they're up to 88 [68:03] health systems now. So, they really are [68:05] starting to penetrate the market, just [68:07] going one by one by one by one to all [68:09] these different hospital systems and [68:11] selling them their software. Yep. So, [68:13] they're ready for the big leagues. And [68:14] then in 2003, they get a call not just [68:18] from the biggest player in the biggest [68:21] league, Kaiser Permanente in California. [68:25] But before we tell the Kaiser story, [68:27] which completely transformed the [68:29] company. Yep. This is a great time to [68:32] thank friend of the show, Fundrise. The [68:34] Fundrise team is awesome and they are [68:37] big acquired listeners just like all of [68:39] you. And they've been evolving since we [68:41] first worked together about 3 years ago. [68:42] So, at the time, Fundrise was mostly [68:45] known as the US's largest real estate [68:47] investment platform for retail [68:49] investors. However, they were watching [68:51] technology markets get larger and larger [68:53] than ever thanks to Moor's law. And [68:55] technology of course sort of took over [68:57] every industry like healthcare and the [68:59] team there was really feeling like it [69:01] really is a shame that all these tech [69:03] companies are staying private longer and [69:05] that means that retail investors can't [69:07] get access. So in 2022, they launched [69:10] their move to bring the democratized [69:12] model they had developed into venture [69:13] investing, which was a pretty contrarian [69:15] idea. Totally. This had been tried in [69:18] the past, but never really worked. Fast [69:20] forward to today, Fundrise has invested [69:22] in great companies like Data Bricks, [69:24] Canva, Anderil, Ramp, fellow friends of [69:27] the show, Vanta, and Anthropic, and also [69:29] Service Titan, which just went public [69:31] last December. Yeah, it is crazy what [69:33] Fundrise has done. They've taken a [69:35] retail platform that any American can [69:37] invest in and gotten preipo access to [69:39] some of the best private companies in [69:41] the world. They have enabled access to [69:43] all the value creation that's been [69:44] locked up in private companies. When the [69:46] Service Titan IPO happened, thanks to [69:48] Fundrise, tens of thousands of regular [69:50] investors got to celebrate alongside [69:52] VCs, LPs, and employees. And look, [69:55] timing really is everything here. [69:56] Fundrise is doing this at a moment where [69:58] thanks to AI, there's more value than [70:01] ever being created in the technology [70:03] sector of the economy. You can go check [70:05] out the full portfolio that Fundrise is [70:07] building at [70:08] fundrise.com/venture. And if you are a [70:10] growthstage founder looking for a great [70:12] series C or later investor, get in touch [70:15] and tell them that Ben and David sent [70:17] you. This is a paid endorsement for [70:18] Fundrise and all investments can lead to [70:20] a loss. All right. So, David, Kaiser [70:24] Permanente, the biggest of big fishes [70:27] comes knocking. Yes. So, Kaiser [70:30] Permanente is this fascinating [70:32] organization that was started and is [70:34] headquartered here in California that is [70:36] a fully integrated quote unquote managed [70:39] care consortium. And so, what that means [70:41] is that essentially they are both your [70:44] health plan, your insurance and your [70:47] hospital system allinone. It' be like if [70:49] a insurance provider married up with a [70:52] hospital system and said, "This is our [70:54] captive hospital system. You as our [70:56] members are only going to go get your [70:58] care here and we control the whole [70:59] system." Yeah. It's the closest thing [71:02] that you can sort of have to a single [71:05] payer here in the country. But [71:08] importantly, they do also need to work [71:10] with and interface with Medicare as they [71:13] have their own Medicare Advantage plan. [71:14] And then so for their older patients, [71:16] customers, once they get older, they [71:18] transition to Medicare Advantage through [71:20] Kaiser. Gotcha. Now, at the time here, [71:22] we're in 2003, Kaiser was the largest [71:26] single hospital system in the entire [71:28] country. 30 hospitals, 400 plus clinics, [71:33] 11,000 physicians, and 8.5 million [71:36] patients that are part of Kaiser here in [71:38] 2003. and they decide that they are [71:42] going to put out an RFP for a whole new [71:47] integrated entire Kaiser [71:49] systemwide EMR system. And this is when [71:53] you really should start to think, okay, [71:56] EMRs are not just medical records or [71:59] medical records tied to billing and [72:01] scheduling. This is the operating [72:03] system, yes, for this industry. Someone [72:06] described it to me as the nervous system [72:08] for a health care system. You know, [72:11] you've got 20 hospitals, a bunch of [72:13] doctors, bunch of administration people. [72:15] The whole thing is tied together by this [72:18] unbelievably complex tentacles [72:21] everywhere piece of software with [72:23] thousands of different screens and [72:25] levels of authorization and [72:27] authentication and roles and [72:29] permissions. And I mean it is incredibly [72:32] hairy and is the single nervous system [72:35] that the entire organization runs on. [72:37] Calling it an operating system is taking [72:39] it too lightly. It's like your operating [72:41] system plus your ERP system plus your [72:45] applications on top of your operating [72:47] system. It's your everything. It's quite [72:50] reasonable even though we refer to these [72:53] things as EMRs to start thinking about [72:54] EMR as a feature for a constituency of [73:00] the whole system. Yes. So in 2003, [73:04] Kaiser puts out an RFP for a new again [73:07] quote unquote ER but really new nervous [73:09] system. Yeah. For all of Kaiser and Epic [73:12] wins the deal. And Epic was a little [73:15] company just a few years before they [73:16] were a $50 million company. Yes. And [73:19] after they sign Kaiser, they go to 162 [73:23] million in annual revenue. So [73:25] transformative. Probably more than [73:27] doubled their revenue overnight. Yep. [73:29] Interestingly, let's talk prices for a [73:31] second. The headline number, as everyone [73:34] reports it, it's a $4 billion deal. They [73:38] call it a $4 billion three-year project. [73:41] And that Epic's portion is around 400 [73:44] million. Of course, not all in one year, [73:48] but the way these things work is there's [73:49] a big implementation that costs a bunch [73:51] of money up front and then there's the [73:53] ongoing license that I think eventually [73:55] would transition to a subscription, but [73:57] at this point it's licensing to use the [73:59] software. It's just funny to see these [74:01] headlines because the four billion [74:03] number, not only do they capture many [74:06] years of the deal and the [74:08] implementation, but they also roll in [74:10] there the headcount of the hospital [74:12] employees that have to do the work and [74:14] they roll in there the potential lost [74:18] productivity from all doctors across the [74:20] health system who have to sort of ramp [74:22] on the new software, which is a real [74:24] economic impact by the way, for sure. [74:26] But it's not like Epic got $4 billion [74:29] out of this. No. And I always chuckle [74:30] cuz every single one of these numbers [74:32] looks huge. Multi-billion dollar. Oh, [74:34] even if it's small health system, $300 [74:36] million project and it ends up resulting [74:39] in nowhere near that much money to Epic. [74:41] But this is how the industry has decided [74:43] to talk about the size of these deals. [74:45] Yeah, it's funny. We should start [74:47] talking about the size of acquired and [74:49] you know like fourth and fifth degree uh [74:51] tertiary impacts. We are a billion [74:54] dollar business. Just think about all [74:56] the business that our customers do when [74:59] acquired listeners. [75:01] But to your point, it is actually fair. [75:03] If Kaiser is going to engage in [75:05] switching their central nervous system [75:07] over several years, it's going to be [75:10] net4 billion of impact to them. Yes. So, [75:14] the story of how this goes down is wild. [75:17] At the time, Kaiser's two main centers [75:21] of gravity were Northern California and [75:24] Southern California. And they were [75:26] almost like separate companies under the [75:29] Kaiser umbrella. Had different systems, [75:31] had their own EMRs, different [75:33] management, of course, talked to each [75:35] other, and part of the same parent [75:37] organization. Wasn't there like a almost [75:39] like cousin organization that was [75:41] Northwest? Yes. Like this stepchild? So [75:44] they had I don't know if it was just [75:46] Northwest or they had maybe some other [75:48] smaller regional operations at the time, [75:50] but they had a Pacific Northwest small [75:52] region based in Portland, Oregon. Today [75:55] actually Kaiser has large regions [75:57] through a large part of the country. [75:58] They've grown a lot since then, but this [76:00] small little Portland region, they had [76:03] started using Epic for their ambulatory [76:06] clinics. So like not even inpatient [76:08] stuff in the hospital, but their [76:10] outpatient clinics. And at the time, the [76:13] Northern California and Southern [76:15] California big factions were battling [76:18] each other and they were each trying to [76:19] develop their own [76:21] proprietary EMR systems with software [76:24] consultants with like Accenture and [76:25] stuff. Yeah. There was this era where [76:27] hospitals thought that EMRs should be [76:30] their IP that they develop and have sort [76:33] of a competitive advantage over other [76:36] hospitals because their EMR was better. [76:37] I don't know what the thinking was, but [76:39] people wanted to own their own EMRs. I [76:42] think there may even have been some pipe [76:43] dreams of like, oh, we're going to [76:45] commercialize this and sell it to other [76:47] hospitals. It seems like not a core [76:49] competency that hospitals should be [76:51] doing. Anyway, within the Kaiser system [76:54] though, there was a fairly high degree [76:56] of rotatability of physicians, of [76:59] doctors. So if you were a doctor in [77:02] Portland with Kaiser Pacific Northwest [77:04] and you wanted to move or your family [77:06] had to move down to California, you [77:09] could transfer pretty easily to Northern [77:11] or Southern California Kaiser. So this [77:13] was happening and as physicians from the [77:15] Northwest started coming down to [77:16] California, they'd be like, "Man, what [77:18] are you guys doing? You're spending all [77:20] this money with Accenture and you know, [77:22] blah blah blah, all these consultants [77:24] and trying to roll your own. Like we've [77:26] got this thing called Epic up in [77:28] Portland that we're not even using at [77:29] the hospital and it's way better than [77:32] the stuff that you're trying to build. [77:34] Yeah. So finally after a year or two of [77:39] this battle between North and South, [77:41] they finally agree like, "All right, [77:42] ceasefire. It's come to a truce. We're [77:44] going to ditch our competing projects [77:46] and we're going to bid this out to third [77:48] party vendors." So they hold an RFP for [77:52] a new EMR for all of Kaiser, one [77:55] integrated system, and they pick IBM. [77:59] IBM is going to come in and do this big [78:01] $4 billion project for Kaiser. Nobody [78:04] gets fired for buying IBM. We're on like [78:05] the end of that era, but it's still a [78:07] little that era. I mean, this is only [78:08] 2003 here. This bid might have even [78:10] happened before 2003. So IBM comes in [78:14] and the project fails. Doesn't work. And [78:17] this is not uncommon, especially think [78:20] about ERP. The number of times you've [78:22] heard, oh, failed ERP implementation and [78:24] some CEOs explaining on an earnings [78:26] call, yeah, we lost hundreds of millions [78:29] or billions of dollars and we actually [78:31] didn't even switch systems. This sort of [78:33] thing does happen in the hospital world, [78:36] too. And Epic has bet the whole company [78:38] on having a reputation for we don't have [78:40] failed implementations, and that wins [78:43] deals. Totally. I I think this is one of [78:45] the most important reasons why their [78:48] customers love them. And we're going to [78:49] say this a bunch of times on the [78:50] episode, their customers love them. When [78:52] we say customer, we mean hospital CEO, [78:55] CIO, the chief information officer and [78:58] CFOs. When we refer to their customers, [79:00] that's who we're talking to. Obviously, [79:03] the chief medical officer and all of the [79:05] physicians and nurses and care folks are [79:07] a part of that. But a part of my [79:10] research has revealed the customer is [79:13] the hospital administration. I doubt [79:15] that's even like a controversial [79:17] statement probably at Epic themselves. I [79:19] think they consider the customer to be [79:21] the CEO, CIO, and CFO of their [79:24] customers. Yep. So the IBM project fails [79:28] and maybe it's worth another word on [79:30] that too. It's not like this doesn't [79:34] impact the physicians and the staff of [79:37] the hospitals and poor performance for [79:40] the hospital as is in the case of this [79:42] failed IBM implementation massive [79:45] detriment to the physicians. This must [79:47] have really really sucked for them like [79:49] they want stability too. And the [79:51] physicians provide the value at [79:53] hospitals and so therefore you need to [79:55] make the physicians happy to retain your [79:58] administration job. They have the [79:59] leverage in the organization because [80:01] they provide the core competency. But [80:02] still, you're not going in and pitching [80:05] the doctors when you're going in and [80:06] trying to land a customer. You could [80:08] think of a hospital almost like a media [80:10] company. The doctors and the nurses and [80:13] the clinicians. They're the on-air [80:14] talent. Yeah. The on-air talent are not [80:17] the ones, you know, at Disney or [80:19] Universal or wherever who are making the [80:21] business decisions, right? And in this [80:23] case, the medical staff is involved, but [80:24] they're not the decision maker. Yeah. [80:27] So, okay, Kaiser's now got to rebid this [80:29] project. And by this point in time, [80:31] enough Portland doctors had come down to [80:33] California and sung Epic's praises that [80:35] they're like, "All right, we should take [80:36] this little company seriously." So, they [80:38] start the RFP and it comes down to Epic [80:42] and their big main rival, Cerner. Now, [80:45] we haven't really talked about Cerner so [80:46] far in this episode, part of Oracle [80:48] today, as we will get into later. Cerner [80:50] was a much bigger company. Yep. [80:53] Interestingly, started right around the [80:54] same time, right? Yeah. Wow. I think it [80:56] was started the same year as Epic, [80:58] right? 1979. Yep. By Neil Patterson. By [81:00] Neil Patterson in Kansas City, Missouri. [81:03] Unlike Epic, which took the no venture [81:07] capital, no acquisitions, single [81:09] platform, stay private forever route, [81:12] Cerner took the lots of acquisitions, [81:16] raise capital, go public, get big route, [81:19] which could have worked, too. It did [81:21] work, too. It did work for a long time. [81:24] Yep. At this point in time, Cerner is [81:27] almost a billion dollar a year revenue [81:29] business. They're a public company, way, [81:31] way, way bigger. They're international, [81:33] etc. So, the RFP comes down to the two [81:36] of them. But it's worth saying before [81:37] they were acquired by Oracle was a [81:40] merger of 24 different companies. Put a [81:42] nice little wrapper on it and call it [81:44] Cerner, but a lot of companies along the [81:46] way. Yes. A different path than the one [81:48] Judy took, let's say. Yes. But that's [81:50] not to knock it. It is and especially [81:53] back then was a good really competitive [81:55] product. The obvious choice for Kaiser [81:58] here in 2003 would have been go with [82:00] Cerner and they actually try to go with [82:03] Cerner. So supposedly at one point they [82:05] come to both companies to Cerner and [82:07] Epic. They say look Cerner we want to go [82:11] with you for inatient in the hospital [82:13] since that's your bread and butter. This [82:15] is the most important thing. This is the [82:16] big business. We trust you. You've been [82:18] around forever. and Epic. Hey, you've [82:20] got impatient now, but you only just [82:22] launched that in 2001. Like, you're new [82:24] at this. You are good at ambulatory at [82:26] outpatient. You're doing a really good [82:28] job for us in Portland. We want to split [82:30] the baby here and do one system with [82:34] Epic for our outpatient clinics and one [82:36] system with Cerner for our inpatient [82:38] clinics. Oh, yeah. That always works. [82:40] And Judy says, "No, that is a bad [82:43] choice. That is the wrong choice to [82:45] make. And I don't care if you go with us [82:47] or them, but to do the right thing for [82:49] your patients and for your whole system [82:50] and to make billing work and have this [82:52] all function correctly and have patient [82:54] records transfer between your ambulatory [82:57] clinics and your hospitals, which you [82:58] really, really need. You should just [83:00] pick one of us. Now, this was a little [83:03] self- serving on Judy's part because [83:06] Cerner was not good at ambulatory. [83:09] It's a calculated high-risisk decision. [83:12] However, definitely super ballsy to do [83:14] this. But she knew that Epic had a good [83:18] product in both. The problem was just [83:20] that their impatient product was still [83:21] new and so didn't have trust yet in the [83:24] marketplace. Yeah. But she's still a [83:26] pipsqueak. I mean, Cerner's a billion in [83:28] revenue at Epic is like somewhere [83:29] slightly north of 50 million at this [83:31] point in time. So they go a little [83:33] further in the process and at one point [83:36] there's a technical due diligence [83:38] meeting where Kaiser asks both companies [83:41] to come in and present to them about how [83:44] their systems are going to handle all [83:46] the volume of concurrent data [83:48] transactions that Kaiser has. Remember 8 [83:50] and a half million patients 11,000 [83:52] physicians is a high volume system here. [83:55] Yes. And listeners this is a great [83:57] story. This is kind of the 11th hour of [84:00] the deal. This is, hey, we're pretty [84:02] close to a decision, but we haven't made [84:03] it yet. Can you both come to the same [84:06] building and sit in different conference [84:08] rooms and throughout the day, we're [84:11] going to bounce back and forth and keep [84:12] spending an hour with each of you, [84:14] formulate some questions from hearing [84:16] the other pitch, and then come back and [84:18] ask you those questions. And so if [84:20] you're on one team or the other, you can [84:22] kind of learn through the questions what [84:26] is being pitched in the other room and [84:28] why am I suddenly being grilled on this [84:29] new topic. So the Epic team is doing [84:32] their planning, you know, getting ready [84:33] for all the preparation for this really [84:35] big meeting. And the team had decided [84:38] that the way they were going to handle [84:40] this question was to do a theoretical [84:42] presentation about how Epic's [84:44] architecture worked, the single system [84:47] and theoretically how much load could [84:49] the system handle all at once. And the [84:53] story is that Carl Devorak, the [84:55] president, flew in to California the [84:58] night before the pitch, meets with the [85:00] team, sees this plan, and is like, [85:02] "Guys, no, we need to model out in Excel [85:08] exactly what Kaiser's transaction flow [85:12] is going to be throughout the day in [85:14] this system and how our system will [85:16] process it and prove to them that we [85:19] have far in excess bandwidth capability [85:21] to handle their system and it'll never [85:23] go down. Cuz I think they had done it at [85:25] sort of a theoretical formula level, but [85:28] Carl knew that Epic had the advantage [85:31] here that they actually could if they [85:34] played it all the way out and really [85:35] built out the spreadsheet show actually [85:38] we're going to be more performant for [85:39] you. Yep. So he and the team pull an [85:42] allnighter the night before the [85:43] presentation. I don't know if Judy was [85:44] there as well. I assume she was too. [85:46] come in and they show the model during [85:49] these meetings that Kaiser is having [85:51] back and forth between the two teams. [85:53] And as the day is going on, it becomes [85:55] really clear that Cerner has not done a [85:59] similar level of modeling and can't [86:01] actually prove to Kaiser that their [86:03] system is going to be able to handle the [86:04] transaction flow. I think this was the [86:07] moment when the tide turned that Epic [86:08] was like, "Oh yeah, we're going to win [86:10] this thing." But it still wasn't obvious [86:12] right away. I think they earned big [86:13] points there, but my understanding is [86:16] that Kaiser still went to Cerner and [86:20] said, "We're interested. Can we do an [86:23] equity deal on top of this? Can we take [86:25] part of the company in exchange for [86:28] basically giving you this big deal?" I [86:30] think Cerner said yes. Yeah. The story [86:34] as we heard it at least is that at the [86:36] last minute then right before the [86:38] decision, yeah, Kaiser came to both [86:40] companies and said, "Hey, we'd really [86:43] like warrants in your company. We're the [86:45] biggest health system in America. This [86:47] is the biggest contract you're going to [86:48] get. We want some equity in your [86:49] companies for working with us." And [86:51] apparently Cerner did offer them 10% of [86:54] the company for this deal. That's how [86:56] important this was. So then they come to [86:58] Epic and they say then they come to Epic [86:59] and they're like, "Well, what do you [87:01] have to say about that?" And Judy's [87:04] like, "No, we're not going to do it. [87:06] We're not going to do it for you. We're [87:07] not going to do it for anybody." And [87:09] it's the wrong thing to do. If we did it [87:10] for you, we'd have to do it for all our [87:12] big customers. And then that would turn [87:13] out poorly for you, too. So absolutely [87:15] not. And they still picked Epic in the [87:19] end. through that, through the [87:22] architecture bake off, and through the [87:26] no, you have to pick one of us, Epic [87:29] still won out. Yeah. and stuck to their [87:31] guns all the way through that [87:32] negotiation, which I think really should [87:35] just tell you how [87:38] important the stability and continuity [87:42] of the system across inpatient, [87:45] outpatient, and billing is because Epic [87:47] is the only one that can offer that. [87:49] Yeah. By the way, earlier when I said 24 [87:51] different companies merged together to [87:53] create Cerner, I forgot that this chart [87:56] that I'm looking at predates when Cerner [87:58] then bought Seammens. So then there's [88:01] another 12 companies that had merged [88:02] together to become Seammens that merged [88:05] also in Derner. Yep. We're going to come [88:07] back to the Seaman's acquisition in a [88:09] minute. I think that was 2014 when that [88:11] happened. But yeah, this is key. This is [88:14] why Epic wins. Yep. So they win the [88:18] deal. Basically doubles or triples [88:20] revenue overnight. The LA Times writes [88:22] about the deal when it gets announced. [88:23] Quote, "Because of its scope, the Kaiser [88:26] Epic system could become the Model T of [88:29] its industry. Not the first of its kind, [88:32] but the first to reach masses of people. [88:35] Once this happens, Epic gets elevated to [88:39] the new gold standard. If you're a [88:41] hospital system, if you're a CIO or a [88:44] CEO looking to rebid your EMR, well, [88:48] Kaiser, the biggest system in the world, [88:50] just chose Epic. and they chose it over [88:52] all of these reasons not to, there must [88:54] be something really good in there, like [88:56] of course you're now going to consider [88:57] Epic and of course Epic is going to [88:59] perform really well in these [89:01] evaluations. Yeah, it helped a little [89:04] bit at first, but after the go live a [89:06] few years in, after the whole [89:07] implementation took place and they [89:09] didn't tip over and it did go well, that [89:11] was really when the floodgates opened [89:12] sort of in that [89:14] 2006 78 time frame. Yep. So by 2007, [89:18] Epic has hit 500 million in revenue. So [89:21] another 3x what they were doing once [89:24] they added Kaiser and like 8x what they [89:27] were before Kaiser, almost 10x what they [89:29] were before Kaiser. They're really [89:32] starting to transform into a big [89:34] company. I spoke with one former [89:35] employee and I was asking what were the [89:37] inflection points in the company and [89:40] this employee said, "Oh, after we won [89:42] the Kaiser deal, we had been hiring like [89:44] 10 kids a month out of college and it [89:46] now felt like hundreds a month were just [89:48] flowing into the doors so we could [89:49] scale." Which will bring us in a minute [89:52] to Epics's epic, shall we say, Verona [89:56] Wisconsin campus. Yeah. And one thing [89:59] just for our storytelling narrative [90:01] there we kind of skipped over was [90:02] there's another thing that happened in [90:04] the early 2000s that reads a little bit [90:07] as a sort of alternate history for what [90:10] could have happened at Epic. Imagine [90:12] you're a small company and a big company [90:14] comes to you and says can we co-develop [90:16] a new product together? You can [90:18] distribute it to your customers. we'll [90:20] distribute it to our customers, you [90:22] know, under our brand name and [90:23] everything, but you get some of that [90:25] revenue and we'll sort of build this [90:26] with you and it really will charge your [90:28] business up. And if you don't know any [90:30] better and you've never done it, it kind [90:33] of sounds appealing. Maybe we should do [90:34] that. And there are many instances of it [90:38] working. So, you know, that's extra [90:40] tempting. My favorite weird example of [90:42] this in history of it not really working [90:43] is the HP iPod. Do you remember this? [90:47] Oh, yeah. That's right. I do remember [90:49] this. I think I might have had one. No, [90:51] I have one of the YouTube iPods, the red [90:53] and black one. Listeners, Google it. [90:54] It's this really odd thing. It's an HP [90:57] brand on the back of an iPod. That tells [90:59] you all you need to know about what a [91:01] weak position Apple was in at the time [91:02] that they were willing to let HP put [91:05] their brand on something entirely [91:06] created by Apple to get HP's [91:08] distribution and get some cut of that [91:10] revenue. And of course, it would have [91:11] bootstrapped their ecosystem since it [91:13] used iTunes and all that. But that's [91:15] basically what happened here. So, [91:17] Philips, the Dutch company, comes to [91:19] Epic and says, "We want to do something [91:22] focused on the radiology segment. You do [91:25] a lot of the development work. We have [91:27] the customer relationships and [91:28] distribution." Right. Cuz they're [91:29] probably selling the machines. Yes. We [91:32] Phillips will get a license to Epic's [91:34] whole IT system. Everything that you [91:36] already sell, we want to sell also. Oh, [91:38] wow. And we're going to market that as [91:41] Phillips externity enterprise to our [91:44] customers. They're gonna buy a Philip [91:46] branded version of Epic. Interesting. [91:49] And I imagine Phillips probably has a [91:52] lot of customers not in America too, [91:54] like in Europe. So this is a way to go [91:56] international, right? So Epic starts [91:58] hiring people in the Netherlands. They [91:59] build up this team. They spend multiple [92:02] years or at least a year building it [92:04] out. I think they even launch it. The [92:07] whole thing ends up folding and within a [92:10] year or two of coming out it was a [92:11] really expensive detour and the company [92:14] develops this intense scar tissue for [92:16] partnerships. Partnerships at this era [92:19] to the epic team means a stuff I can't [92:23] control outside my organization. B risk [92:26] c never in our history do we have an [92:29] example of it working. This is an [92:31] uncontrollable dependency. And they sort [92:33] of internalize this scar tissue as stay [92:35] focused on what we can control. Go [92:37] directly to the customer. Don't try to [92:40] do any fancy partnership integration [92:42] stuff with other people. And that's [92:44] oversimplifying it, but I think it's [92:46] still reasonable to say that some of the [92:48] DNA of what Epic would become and they [92:51] would take tons of arrows for this this [92:53] closed not interoperable blah blah blah. [92:55] Some of it stems from this failed [92:58] partnership with Phillips. Yep. Totally. [93:02] They're now partnering again finally, [93:05] but yeah, like 15, 20 years later. Yes. [93:08] And thanks to Health API Guy for the tip [93:10] on that. Yeah. So, at this point in the [93:13] late 2000s, Epic is eyes on the prize. [93:16] We're building everything ourselves. We [93:18] can get big customers on our own. We can [93:21] stand firm and not negotiate, not have [93:23] to give up pieces of our company. The [93:26] price is the price. And we know that [93:28] we're going to deliver. And so it's time [93:31] to invest in our future. How do they [93:34] invest in the future? Uh the Verona [93:36] campus. So for the probably minority of [93:39] you listening who know anything about [93:41] Epic as a company or been involved with [93:43] them in the past, if you have, you [93:46] almost certainly know about their [93:47] corporate campus. Yes. So there's sort [93:50] of two stories of how the campus came to [93:52] be. First is Judy's son by this point in [93:57] time is actually working at Microsoft in [94:00] Redmond as a developer. And Judy and [94:04] Carl had always been inspired by [94:06] Microsoft's way of doing things. And so [94:10] one time when Judy's visiting her son, [94:12] she's like, "Hey, can you go give me a [94:13] tour of the campus in Redmond? We're [94:15] thinking about expanding. We're out [94:17] growing our space. I want to see what [94:18] it's like there." And she's really [94:19] impressed by Redmond, as I think anybody [94:21] who goes there would be. Yep. Especially [94:23] in this era, it was sort of pre [94:26] Googlelex. It was kind of the fairy tale [94:28] tech campus. Yes. It was Google before [94:31] Google. Yep. And she's like, this is [94:33] like a college campus here. The [94:35] atmosphere that you've got sports field [94:37] and tons of buildings and all of them [94:39] are pretty short. Two, three, four [94:41] stories, a lot of outdoor space, walking [94:44] space, everyone gets an office. This is [94:47] perfect. So Judy comes back and [94:50] basically copies the campus strategy of [94:52] Microsoft almost exactly. They had been [94:55] in that renovated schoolhouse. I mean if [94:56] you think about where they came from and [94:59] sort of what they now have the [95:01] opportunity to build. They had renovated [95:03] the schoolhouse to be nice but Yeah. [95:05] That was well before they were a half a [95:07] billion dollar a year revenue company. [95:08] Yes. Exactly. So she goes out and buys a [95:11] thousand acres of farmland in Verona, [95:16] Wisconsin, which just like Redmond is a [95:19] kind of bucolic looking suburb about [95:20] half an hour outside the city in [95:22] Madison. Or I guess maybe Redmond was [95:24] bucolic before Microsoft got it built [95:27] up. But just like Microsoft kind of owns [95:30] Redmond, Epic owns Verona spiritually. [95:33] So they build this incredible campus [95:35] there. And what is very different about [95:37] it than Microsoft is it is not [95:40] utilitarian the way that the Microsoft [95:42] buildings were. A thing that was already [95:45] happening at Epic was this fairy tale [95:47] whimsical thing that you can kind of see [95:49] in their product names. But David, what [95:51] was the story we heard about the [95:53] schoolhouse fireplace? Yeah. Okay. So, [95:55] this is the second story of the Epic [95:58] campus. So, like you said, Ben, before [96:01] Verona, the headquarters was in an old [96:03] school building in Madison that they' [96:05] bought and renovated. And as they were [96:08] renovating it, the designers decided [96:10] that in one of the main conference [96:12] rooms, they were going to put a [96:14] fireplace in there to make it feel more [96:16] homey. Yes. Like a Wisconsin lodge. Yes. [96:20] And I don't even know if the designers [96:22] did that, but I think Judy and the [96:23] company were like, "Well, this kind of [96:25] feels like a Wisconsin lakes lodge. [96:28] let's lean into it. And so they [96:30] decorated it like a lodge. They brought [96:32] snowshoes and furs and like a axe that [96:34] they put on the wall and stuff and it [96:36] ends up becoming the most popular room [96:38] in the building. And anytime customers [96:40] are coming to Madison, they always want [96:42] to meet in the lodge. Yes. Conference [96:44] room. And so now they're building this [96:46] new campus institute. He's like, "Oh, [96:48] well, let's take the lodge idea and [96:50] really blow it out. Maximize it." Yeah. [96:53] So, they inject this fairy taleness [96:57] times 10. When they're building out the [97:00] new campus, they hire the same firm that [97:04] did the Disneyland California Adventure [97:08] renovation in 2008. I think it's even [97:10] more on the nose on that. They hired two [97:12] architecture firms. One is that Uhhuh. [97:15] and then they also hired the firm that [97:17] built a lot of the Redmond campus for [97:18] Microsoft. Oh, is it really? Yeah. Wow. [97:22] And it's crazy. I mean, it's Alice and [97:23] Wonderland stuff. It's Harry Potter [97:25] inspired stuff. It's Wizard of Oz [97:27] inspired stuff. Just Google pictures of [97:29] the Verona campus for Epic. And we'll [97:32] link to some in the show notes, too. [97:33] It's bananas. But it is extremely [97:36] attractive to new hires coming out of [97:39] college who want to kind of feel like [97:43] they're still in college and want to go [97:44] work for a company that seems fun and [97:46] interesting. And Epic is this sort of [97:49] two-sided culture that seemed to play [97:52] well together somehow. This goofy, [97:56] whimsical fairy tale thing and this hard [97:59] driving win at all costs [98:03] performanceoriented fierce competitor. [98:06] And it just is both. To understand the [98:08] company, you have to hold in your head [98:09] that the DNA is both of those things [98:11] concurrently. And I think it's because [98:12] that's what Judy is. Yes, 100%. That is [98:16] completely spot-on. I had in my script [98:18] here the question of all this is why on [98:21] the campus and that is exactly why we [98:24] are [98:25] hiring super smart young hungry new [98:28] college grads. How are we going to [98:32] attract them to Verona Wisconsin? Well, [98:34] we are going to create a paradise for [98:36] them. Yes. So, some interesting stats. [98:39] It's 1,700 acres. 410 of them are the [98:42] campus. The rest is the farm. now covers [98:44] 89 buildings. There are four indoor [98:46] auditoriums with 18,000 seats total. The [98:49] big one, Deep Space, is the world's [98:53] largest subsurface auditorium. There are [98:58] 11,400 seats. This is an auditorium. [99:01] This is two radio city music halls full [99:04] of people smashed together in one giant [99:08] auditorium underground on a corporate [99:10] campus. The number of seats is much [99:13] closer to a chase center. It's much [99:16] closer to a basketball arena than it is [99:18] to any other auditorium that I can think [99:20] of. It goes down 74 ft beneath the [99:23] surface. I mean, the whole logic behind [99:26] it is when they built it, they thought, [99:28] "Oh, we'll never grow to 11,400 people. [99:30] We can have our all hands in here. Nob [99:32] brainer. But this can also be the place [99:33] where we have all of our customers and [99:35] our whole ecosystem can come here." Of [99:37] course, now they don't actually fit in [99:39] there because they've outgrown it. Not [99:40] even all their employees can come to all [99:42] hands there. The logic for building it [99:44] was that they had a movie theater that [99:46] they used to do their monthly all hands [99:49] beating and they wanted to sort of have [99:50] this opportunity to do it on their new [99:52] campus, too. It's wild. I mean, when you [99:54] look at it, you just can't believe the [99:56] scale of this building. Yep. All right. [99:58] I think now is the right time while [100:01] we're talking about the campus and epic [100:03] culture to really talk about epic [100:04] culture. Yeah. Because the first thing [100:06] that you have to understand is Judy [100:07] refers to it as a software factory. When [100:10] you keep looking at it and you're like, [100:11] why is it so weird? The biggest takeaway [100:14] is in Judy's mind since they don't ever [100:18] go buy any other companies and they [100:21] don't have any competencies at the [100:23] company other than making software, what [100:25] they are is a factory that turns out [100:28] software. They take in software [100:30] developers and they turn that into [100:31] software for the medical industry. [100:34] That's funny. Apparently, she was just [100:36] ahead of her time with the AI factories. [100:38] Was it Nvidia and Dell are calling, you [100:39] know, Yeah. AI factories now. Yeah. [100:41] Yeah. Absolutely. And so it starts to [100:43] click and make more sense when you [100:45] think, well, what would a factory for [100:46] turning developers into medical code, [100:49] medical applications look like? Well, [100:51] Verona, Wisconsin. Yep. So, while we're [100:54] on culture here, one of the most amazing [100:56] things about the campus is Epic has a [101:01] list of ten commandments. It's the epic [101:04] ten commandments is like you know Moses [101:05] in the Bible here. They have them posted [101:10] in every bathroom and in every break [101:12] room across the entire campus. And so [101:15] any visitor who goes there and it's open [101:16] to the public. You can just go in and [101:18] see the ten commandments in the [101:19] bathrooms there. The ten commandments [101:21] are number one do not go public. Number [101:25] two do not acquire or be acquired. By [101:27] the way those first two things you don't [101:29] need to communicate that to employees. [101:30] Only the CEO can do either of those [101:32] things. So, it's sort of funny to put it [101:34] as commandments that shows how deeply [101:36] Judy feels it needs to run in all the [101:38] employees. And I think the other big [101:40] motivation for doing this is and having [101:42] it there in the bathrooms is every [101:43] customer who comes to visit, it's right [101:46] there to the customers too, right? We [101:48] will never go public. We will never be [101:50] acquired and we will never acquire [101:51] another company. You can trust this is [101:53] one system forever. Yep. Yep. Okay. So, [101:56] that's one and two. Number three, [101:58] software must work. Number four, reality [102:02] equals expectations. Number five, keep [102:05] commitments, even the unspoken ones. [102:07] Number six, focus on competency. Do not [102:09] tolerate mediocrity. Number seven, have [102:11] standards. Be fair to all. Number eight, [102:14] have courage. What you put up with is [102:16] what you stand for. Number nine, teach [102:18] philosophy and culture. And number 10, [102:20] be frugal. Do not take on debt for [102:23] operations. Zero of those pertain to [102:26] healthcare. [102:28] Yes. When I was looking at them, I kept [102:30] thinking, "Oh, I'm going to find [102:31] something here about every life is [102:33] important or the patient is at the [102:34] center of everything or no, this is how [102:37] to run a company. This is my opinion on [102:39] how to run a company, period." Yep. And [102:42] I think specifically to your point, this [102:45] is a pretty good way to run a software [102:46] factory. Yes. So, these hang in the [102:49] bathroom. Other interesting things about [102:51] the campus, they have wedding bells that [102:54] will play campuswide when a new client [102:56] is signed. just showing that that's the [102:58] type of commitment that this is. It's a [102:59] we've now married this client for the [103:01] rest of our lives. Yeah. It's like the [103:03] wedding march, you [103:04] know, like that's their version of like [103:06] ringing the gong. Yes. And I think we [103:09] haven't really talked that much about [103:12] what it is like to be an employee there. [103:14] So, this is a probably a good time to do [103:15] that. This is an insanely awesome [103:18] training ground if you are a smart, [103:20] ambitious person out of college. People [103:23] accuse them of being cultlike, but there [103:26] are ways in which that's a good thing. I [103:27] mean, they take you fresh out of your [103:28] career and they teach you everything. [103:30] And when I say everything, I mean how to [103:32] take notes. There is an epic way to take [103:35] notes on a yellow legal pad. There is an [103:38] epic way to write emails. And these are [103:41] like hardened practices over the years [103:43] that they just believe through [103:44] iteration, through testing, through [103:46] data. Probably going all the way back to [103:48] Neil and Metatech and yeah, the three [103:51] days in Boston. this is the best way to [103:53] do this period. And so we're just going [103:55] to teach everybody the best way to do [103:57] everything and everybody is going to be [103:58] like reasonably robotic. We can sort of [104:01] trust that once we squeeze you through [104:03] the epic system, when you come out the [104:04] other side, you are able to operate in a [104:07] way that works really, really, really [104:09] well in our machine where people can [104:11] really trust each other. So because of [104:13] that high level of trust, there's very [104:15] few middle managers. you understand the [104:18] system that everybody else works within [104:20] and you don't have to corral chaos. Most [104:23] of the people that are hiring are right [104:24] out of school, so they've been there for [104:26] a long time. They don't have budgets, [104:29] which you can kind of only do when you [104:31] have a high trust environment like this. [104:32] I mean, there's some financial controls, [104:34] of course, but Judy's got this great [104:36] story that she used to go see customers [104:38] and they'd say, "Oh, this is the right [104:40] thing to do. I just don't have the [104:41] budget for it this year, so we're going [104:42] to push it next year." She's like, [104:44] "That's stupid." or they would tell her, [104:46] "Can we squeeze this in this year [104:48] because if I don't spend this money, I'm [104:49] going to lose it in my budget." And she [104:51] would say, "Well, that's also stupid." [104:53] So, at my company, we're not going to [104:54] have budgets like that. They're super [104:56] light on titles. You might have a [104:58] business card, whether you've been there [105:00] 20 years or 6 months, that says [105:02] implementation that you hand to a [105:04] customer when you go and do an [105:05] engagement. Everyone does immersion [105:07] trips where the software developers I [105:09] mean everyone is required to spend time [105:11] in clinical settings like operating [105:13] rooms to directly observe workflows. [105:16] Yep. I think when you start you have to [105:18] do five of them and then you do more [105:20] every year. Wow. Why combinator preaches [105:23] go talk to your customers, spend time [105:25] with customers. Epic's been doing this [105:27] forever. This is the epic way that I [105:30] think the rest of the world sort of woke [105:31] up to and startups internalized as [105:33] doctrine. But every person in the [105:35] company spending time in medical [105:38] settings talking to customers hugely [105:40] valuable. On the developer side, there's [105:43] a super prescriptive software [105:44] methodology that they use to minimize [105:47] bugs. So you go through this intensive [105:49] training for 6 months when you join and [105:51] then when you start programming, the [105:52] whole system is designed around [105:54] minimizing the number of hours between [105:56] when a line of code is written and then [105:58] when it is tested. So if a bug is found [106:02] then you drop everything as the original [106:04] developer and you fix it so that you [106:07] still have the whole context fresh in [106:09] your head. You don't sort of go months [106:10] and then the system gets tested and bugs [106:13] aren't allowed to compound into bigger [106:14] problems this way. They get caught right [106:16] away. And I believe the rule right is [106:18] that every developer must fix their own [106:21] bugs. That's my understanding too. It's [106:23] this method of software engineering that [106:26] places way more importance on a zero bug [106:29] environment because lives are on the [106:31] line than other ones. So you're not [106:34] necessarily going to ship software the [106:35] fastest way. You may not even ship the [106:37] most innovative, clever, amazing, [106:38] cutting edge. You're just going to make [106:40] sure that you're shipping bug-free [106:41] software. Yeah. Well, I think the [106:43] reasons are twofold. One, it absolutely [106:45] is right that lives are on the line. If [106:47] a order for the amount of dose in a [106:50] prescription gets messed up because of a [106:51] bug, a lot of people are going to die. [106:53] Yep. And also the complexity required [106:57] for the revenue cycle and billing for [106:59] your customers is of paramount [107:01] importance. You cannot have bugs there [107:04] either or at best the hospital's going [107:07] to lose a lot of revenue opportunity. At [107:09] worst, they're going to get sued for [107:11] federal crimes, for medical fraud. Yep, [107:14] great point. So that means you need a [107:17] highly robust system. Yes, the work done [107:20] as an implementation person at the [107:22] company is insane. It's like military [107:25] level logistics. You're handling [107:26] multiple customers, all of which are [107:28] among the most complex systems on earth. [107:32] Peter Ducker famously referred to [107:34] hospitals as the most complex form of [107:36] human organization that we have ever [107:38] attempted to manage. You have to [107:40] understand all these dependencies at the [107:42] customer and the status of a dozen [107:45] interrelated things on a daily basis. [107:47] You really are in this high [107:51] adrenaline, high stakes leadership role [107:55] as a really young person. You can work [107:57] 10, 12 hours a day, but a lot of them [107:59] love it because you're winning. You're [108:01] doing really big things right out of [108:03] school. You're doing it with other [108:04] really bright people. So they really try [108:06] to get high IQ, high EQ, often like very [108:10] sweet Midwestern kids to take these [108:13] customer-f facing roles. Yeah. And I [108:15] think also because of the flat [108:16] organization, like you're also doing it [108:18] alongside [108:19] other senior people and learning from [108:22] them directly. Like the story about Carl [108:24] and Sumit, you know, Sumit is a young [108:26] programmer working on a project with [108:29] Carl the president who's also a [108:31] programmer leading the team. like that [108:32] happens. If you're an ambitious, career [108:35] focused person, there is nothing more [108:37] fun than winning in a high stakes [108:40] environment with other high performers [108:42] as a team. That plus the whimsy [108:45] encapsulates pretty well the epic [108:47] culture. And the result of that is it [108:50] becomes the number one thing in your [108:52] life. If you talk to a lot of these [108:53] people who spent time there or still [108:55] work there, you're all in and you're in [108:58] the middle of nowhere. This is the other [109:00] job the campus does. You're not really [109:01] getting exposed to other things you [109:03] could leave and go do. They make it very [109:05] easy for your whole life to become epic. [109:07] Yeah. You could drive to the big city of [109:09] Madison. Yes. I know we keep making the [109:12] comparison and it's a direct one with [109:13] the campus here too, but the company [109:16] that this reminds me the most of is [109:18] those early days of Microsoft. This is [109:21] exactly what being at Microsoft in the [109:23] 80s and 90s was like when we talked to [109:26] people doing that research. And I heard [109:27] from some people too when I was [109:28] researching for this two or three times. [109:30] Palanteer came up of these sort of [109:32] brighteyed, bushy tailed, smart young [109:34] people where you're deployed into these [109:37] really intense environments, but you [109:39] know your stuff. You've been through the [109:40] training. You've been through the [109:41] process. You're armed with good tools [109:42] and you're going to go make it happen. [109:44] Yep. So, the other side of this is it is [109:47] up or out. They aggressively trim modern [109:50] whatever percent of performers and they [109:52] work you really really hard. And so with [109:55] the vast amounts of new hires, they're [109:57] trying to figure out if you're going to [109:58] cut it. And it's more cost-effective to [110:00] replace you than keep you as dead [110:02] weight. And so you have lots of [110:03] attrition in the first few years, but [110:05] you know that once someone's been there [110:06] for a while, they're good. You can count [110:08] on them. Yep. Their method of hiring is [110:11] crazy. And this is more common now to [110:13] give out programming tests as a part of [110:15] the hiring process. Oh, this story is so [110:18] good. Can I tell it? Yes, please. So, we [110:21] mentioned that Judy's son was a [110:23] programmer at Microsoft and that's how [110:26] she went to visit him, part of the [110:27] inspiration for the campus. His [110:29] involvement in the company actually [110:30] predates him becoming a programmer at [110:33] Microsoft or even an adult. In the late [110:36] 80s, Judy was hiring software developers [110:39] and found that interviewing them just [110:42] wasn't that predictive of whether they [110:44] were going to be great software [110:45] developers or not. Meanwhile, her son, I [110:47] think, was in like seventh grade or [110:48] something like that and was doing these [110:51] programming competitions around the [110:53] state and she was talking to him was [110:56] like, "Well, you do these programming [110:57] competitions and they're these tests and [110:59] then you're winning some of them and [111:00] that's a pretty good judge of whether [111:02] you're a good programmer or not. They're [111:04] reasonably predictive. Do you think you [111:06] could write one of these types of tests [111:10] for me and I can use it at Epic and test [111:12] software developers as we're hiring them [111:14] and see if they're any good? And for 18 [111:16] years, listeners, that was how they [111:19] tested to determine if someone should [111:20] work at Epic. And many many times, I [111:23] don't know if the number is close to [111:24] 100% or but some large percent of the [111:26] time, they just don't interview you. [111:28] They believe that their tests are [111:30] predictive enough that that's it. You [111:32] can get a job offer. you come visit [111:34] campus and all that, but like you get a [111:35] job offer after scoring high on this [111:38] test. And it's not just that test now. [111:40] There's other tests. There's a Rembrandt [111:41] test. They've got a bunch of tests. It's [111:43] all systemized and cataloged. And [111:46] everybody takes one on the way in for [111:48] every role in the company. Yeah. The [111:50] culinary team, we heard, takes not full [111:52] software developer tests, but logic [111:54] tests on the way in. Yeah. Amazing. But [111:57] yeah, Judy's teenage son wrote the first [111:59] test that is now no longer being used. I [112:01] think the answers have gotten out on the [112:02] internet, but for a long time was used [112:05] and it's a funny story, but this was in [112:07] the 80s. This was way before Google was [112:09] doing this kind of stuff. This was [112:12] really out there hiring practices at the [112:15] time. Yep. So, we've talked a lot about [112:17] the internal culture. The most important [112:19] thing that we haven't really dove into [112:21] yet is how it touches the outside world. [112:24] The Epic culture is completely customer [112:28] obsessed. And I mean that the way we [112:30] talked about customer before. There's a [112:32] great quote from Jeff Gotney, the chief [112:34] information officer for Rush University [112:36] System for Health. He says, "You get [112:38] what you pay for 100% of the time [112:41] despite Epic being quote not cheap." You [112:44] kind of see that echoed over and over [112:45] and over again in these customer [112:46] conversations is reliable. It worked. [112:49] They didn't overpromise on something. It [112:52] is fully integrated. At this point, no [112:54] one gets fired for buying Epic the same [112:56] way it used to be true about IBM. Yep. [112:58] Customers are like always number one. [113:01] Everything they vote. I mean the way [113:03] that they pick the next things to build [113:05] is when all the customers come to campus [113:07] for their annual conference, they ask [113:10] for ideas and then they vote and take [113:12] customer input as a way for figuring out [113:15] what are we going to do next. Totally. [113:17] There are basically only three roles at [113:20] the company. There's software developer, [113:23] project manager who are the [113:24] implementations managers doing the new [113:26] active implementations for new customers [113:29] and then there are technical specialists [113:30] who do ongoing customer support. There's [113:34] not a sales or marketing department. [113:35] There are eight or so salespeople quote [113:38] unquote in the company who only react to [113:41] inbound requests and they all came from [113:43] either project management or technical [113:45] specialists. That's the whole company. [113:47] The technical specialists are the [113:48] biggest group in the company. Every [113:50] single customer of the 607 whatever [113:54] hospital system customers that Epic has, [113:57] every single customer has their own [113:59] technical specialist teams for every [114:03] single product that they use. So like if [114:06] you're a hospital system, you have your [114:08] own technical specialist team for your [114:10] Epicare EMR, for My Chart, for Resolute, [114:13] for Cosmos, for you name it. Anything [114:15] you use, you have your own dedicated [114:16] team for that. And then on top of that, [114:19] every customer has their own dedicated [114:22] BFF, best friend forever, like BFF who [114:25] is a single person within Epic and their [114:28] sole job, their only job is to make sure [114:30] that you as a customer are successful [114:33] with their products. So this means they [114:35] do things like they grade you as a [114:39] customer every year benchmarked relative [114:42] to what other customers are doing and [114:45] how you're doing with the epic tools. So [114:48] they will send separate report cards [114:50] every year to your CEO, CIO and CFO and [114:55] they will grade you one through five on [114:57] a bunch of dimensions and then they will [114:59] show you benchmark data against other [115:02] customers at your peer set of like [115:04] relative similar size hospital systems [115:06] how you're doing. Wow, that's crazy. [115:09] Yes, I've never heard of any other [115:11] company that does stuff like this. Well, [115:13] they have a lot of leverage in the [115:14] customer relationship. I think at this [115:16] point in history, you know, [115:18] 2025, when the customer wants to do [115:20] something a certain way and Epic wants [115:22] to do something a certain way, [115:24] ultimately Epic is customer focused. So, [115:26] they will do whatever the customer [115:27] wants, but they're going to lay out very [115:31] compelling, convincing arguments why [115:33] their way is the correct way. And what [115:36] this leads to is things like a standard [115:40] package. A lot of times when people are [115:42] setting up Cerner, every implementation [115:44] looks completely different. Epic is [115:46] highly opinionated. Please use as much [115:49] standard stuff as you can so that we can [115:51] easily push out updates or we can easily [115:55] add in new modules for you or [115:57] interoperability all works exactly the [115:59] way that we're thinking it should. They [116:00] have a strong negotiating position with [116:02] customers when they're saying, "Uh, I [116:05] think you should do it this way." [116:06] Because at some point they may choose to [116:08] just say, "You know what? I don't think [116:09] you're ready to be a customer yet. I [116:11] think we're gonna focus elsewhere this [116:13] year because we're going to pick up only [116:15] 10 to 20 maybe 30 new customers and [116:18] we're happy to wait. We're happy to wait [116:19] until you're ready to work with us. And [116:21] they actually have the leverage to pull [116:22] that off. Now, to your point about the [116:24] standard implementations, I believe this [116:26] is the only way that you can get [116:29] discounts on pricing with Epic is by [116:33] doing either fully or mostly standard. I [116:37] think there probably are some tiers [116:39] based on how standard your [116:41] implementation is and the more you [116:42] deviate from it, the more you have to [116:44] pay. I could see that. I know if you [116:46] stay up toate on things like database [116:48] maintenance and versioning and all that, [116:50] then they give you discounts. Yep. So [116:52] all of this sounds sort of epic acrewing [116:55] power for Epic, which it is. There is [116:58] also this almost altruistic part of the [117:01] company and their customer [117:03] relationships. And I think this comes [117:04] from just Judy and who she is as a [117:06] person. Altruistic capitalism does feel [117:09] like ultra competitive, [117:11] ultra value maximizing, altruistic [117:14] capitalism is kind of like how I would [117:15] describe the company. Yeah, I think [117:17] that's how you sum up Epic. And this [117:19] altruistic piece is despite certainly [117:22] being [117:23] expensive, they basically never raise [117:26] prices once you're a customer. They do, [117:29] but their average yearly increase is [117:32] about 2%. Across the board, so below [117:35] inflation. Compare that to lots of other [117:37] software companies out there, and [117:39] they're not raising prices 2% every [117:41] year. They also do things like they have [117:44] never in its entire history changed the [117:46] price of my chart. We got told this [117:48] multiple times. I think that's a red [117:49] herring. We don't know what the bottom [117:51] line price looks like. We got told one [117:53] peace meal. It's not like we have a full [117:55] contract in hand of what it looks like [117:57] to be a customer and add up all the [117:58] subcomponents. Good point. Two points to [118:01] make. Yes, they're customer obsessed. [118:02] Yes, they're listening. But yes, of [118:04] course, it's to do what's in Epic's [118:06] long-term interest. I mean, the core [118:07] functionality of if you're a customer, [118:10] the reason you pick Epic is to turn an [118:12] interaction with a patient into as many [118:14] dollars as possible for the health [118:15] system without risking downside with the [118:18] lowest risk possible. Yes. So, that is [118:21] why the customer is picking them. And [118:23] what is Epic trying to do? Epic is [118:25] trying to win deals and stay in forever [118:28] and achieve world domination. So they do [118:32] these very interesting things that feel [118:34] very customerfocused and are but are [118:37] also very valuable for Epic. Like [118:40] they'll recommend alternative [118:42] third-party new pieces of software to [118:45] their customers. Some new thing comes [118:46] out. Oh, it's 2020. Tellahalth suddenly [118:49] is really important. Here's a HIPPA [118:52] compliant way to do Zoom. By the way, [118:54] we're going to start working on our own [118:55] teleahalth thing. It'll be out soon. You [118:58] should feel free to use Zoom right now. [119:00] By the way, when our new module comes [119:02] out, it's just going to be free to you. [119:03] You're already a customer. It's just [119:04] going to be free to you. So, you're [119:06] certainly not going to go around [119:07] shopping for some new thing when you [119:10] know that it's just going to come for [119:11] free to you. And you're not going to [119:12] take it right away cuz it's going to be [119:13] too bare bones. But at some point, it'll [119:16] get good enough where you can say, "Oh, [119:18] yeah. I'll just adopt the Epic version [119:19] now. part of my enterprisewide [119:21] agreement. It's this kind of amazing [119:23] bundling strategy that certainly reminds [119:25] me of the Microsoft episode. Yep, you're [119:28] totally right. I think that's the right [119:30] way to look at it. They're customer [119:31] obsessed because in the long term that [119:34] is the right thing to do for Epic also. [119:37] And this is not a new realization. There [119:39] is a Jeff Bezos quote and I just thought [119:42] of I don't have it in my notes, but it's [119:43] something about in the long run, there [119:45] is no difference between what the [119:47] customer wants and what Amazon wants. [119:50] And I distinctly remember him saying [119:51] this in like 99 or something, some [119:54] really old video. Yep. Epic is [119:56] absolutely customer obsessed. Again, CEO [119:59] CIO of Hospital Systems because if you [120:02] deliver for them, you deliver for Epic [120:05] in the long run. Yep. Well, I mean [120:07] that's kind of like that's like the goal [120:08] of running a company theoretical [120:10] underpinning of capitalism, right? Like [120:14] Right. So, all that to say, it is a very [120:17] fascinating corporate culture and [120:21] organization. Yes. All right. So, [120:23] listeners, if you know anything about [120:25] this industry, you know that we haven't [120:26] gotten to the important part yet. [120:30] We've gotten to the Kaiser deal. That [120:31] was a big deal. Gotten to moving to [120:33] Verona. That's an important part to [120:34] understand the culture and it's like [120:35] this cutesy thing that most journalists [120:38] who write about the company kind of [120:39] latch on to oh cool campus like I should [120:42] go take some pictures and write a cool [120:43] story about campus. There's a whole big [120:46] crazy thing that happens as a part of [120:49] the great recession and some legislation [120:51] that gets passed transformative for the [120:54] industry and causes a whole bunch of [120:57] good and a whole bunch of bad to happen. [120:59] But before we do that, this is a great [121:01] time to thank good friend of the show, [121:04] Service Now. Service Now has been [121:05] working on something really exciting, [121:08] CRM. Now, I know we've talked about [121:11] Service Now as the AI operating system [121:13] for the enterprise, but what you might [121:15] not know is they've also quietly been [121:18] building a billion dollar plus CRM [121:20] business for years. Yep. And part of the [121:23] reason it's been hiding in plain sight [121:25] is because Service Now has always been [121:28] one single platform for all your [121:29] company's data and operations. So it's a [121:32] totally natural consequence that Service [121:34] Now is also one of the best ways for the [121:37] whole enterprise to access and interact [121:39] with customer data. It's fascinating [121:40] because Service Now approaches CRM from [121:43] a completely different angle. [121:44] Traditional CRM systems were built [121:46] decades ago primarily to record data and [121:49] report on activities not to take action [121:51] or connect customerf facing teams across [121:54] your business. Service Now believes the [121:56] new age of CRM is not just about selling [121:59] but also about delivering service across [122:01] the whole totality of your organization. [122:04] Oh, you mean like delivering service [122:06] now? [122:08] Hey, but yes, exactly. Think about it. [122:11] If you have separate systems for sales, [122:13] service, and operations, you end up with [122:15] what Service Now calls human middleware. [122:18] Employees manually updating different [122:20] systems with the same information. But [122:22] Service Now is all just one platform and [122:25] one single source of truth. Ah, sounds [122:27] like a familiar strategy. Ah, kind of [122:29] like what we've been talking about this [122:30] whole episode. Yep. Service or sales [122:33] interactions automatically get updated [122:34] everywhere on Service Now because it's [122:36] just one platform. So since service now [122:39] is already your enterprise operating [122:40] system, shouldn't it also be your CRM? [122:43] And timing is everything with the rise [122:45] of AI agents. So let's say a customer [122:46] contacts support about a delayed [122:48] shipment. In traditional CRM, you might [122:50] log the complaint, but the root cause [122:52] remains hidden in your operation system. [122:54] And with Service Now's modern CRM [122:56] approach, AI agents can seamlessly [122:58] access your data across departments, [123:00] from inventory to logistics to consumer [123:02] history, and not just track the issue, [123:04] but actually resolve it. It's not just [123:06] about managing customers. It's about [123:08] connecting your entire business to serve [123:10] them better. Yep. So, if you want to [123:12] learn more about Service Now's modern [123:14] approach to CRM, go to service [123:17] now.com/acquired and just tell them that [123:18] Ben and David sent you. Okay. So, we're [123:21] coming through the 2000s here. We just [123:24] got through the Verona campus. You know [123:26] everything about Epic's culture now, or [123:28] at least everything that we could [123:29] discern from the outside anyway. And [123:32] we're in this pretty interesting era in [123:34] 2006. We're in the Bush administration [123:37] here in America. And Bush says in a [123:40] State of the Union, "For all Americans, [123:43] we must confront the rising cost of [123:46] care, strengthen the doctor patient [123:48] relationship, and help people afford the [123:51] insurance coverage they need. We will [123:53] make wider use of electronic records and [123:56] other health information technology to [123:58] help control costs and reduce dangerous [124:01] medical errors. Wide applause. This is [124:04] one of the craziest things to me in [124:06] doing the research. the narrative out [124:08] there and certainly in as much as I knew [124:11] or paid attention to any of this was [124:14] that Obamacare and the Obama [124:17] administration were the ones who really [124:19] pushed EMR adoption and meaningful use [124:21] and all this stuff that we're really [124:22] going to get into. Totally bipartisan, [124:25] totally started in the Bush [124:26] administration. Yeah. And just to give [124:28] you a little look into how much the [124:30] window changes of which side represents [124:32] which party, the next paragraph, George [124:35] W. Bush also says, "We will do more to [124:37] make this coverage portable so workers [124:39] can switch jobs without having to worry [124:41] about their health insurance." I mean, [124:44] it's crazy. If you're a Republican [124:46] during the Obama era talking about [124:48] having coverage that's portable across [124:50] employers, that's a scary thing to be [124:52] talking about just four years later. [124:55] Yep. So funny. So you're starting to get [124:58] these political winds of, hey, everyone [125:02] wants a good system for electronic [125:05] medical records. We think in the [125:07] abstract there will be a lot of good [125:09] that comes from it. Well, I think [125:10] there's an even deeper motivation than [125:12] that. In 2006 already, I think everybody [125:16] knows in America the current health care [125:19] system sucks. Yes, there are great [125:21] things about it, but overall this is [125:25] cost disease run rampant. Huge portion [125:27] of GDP, massively inefficient. We want [125:30] to try and fix it. How to fix it, we [125:32] don't all agree on. How to fix it, we [125:33] don't know. But this promise of [125:35] digitizing it and incentivizing EMR [125:40] adoption is held forth as a promise that [125:43] can deliver us from this problem. Right? [125:45] It feels like a step in the right [125:46] direction. And at the time only 13% of [125:50] healthcare facilities in America had an [125:52] ehr system at all. This is a massively [125:56] not digitized industry here in the [125:58] mid200s which I think there's some push [126:01] back on that 13%. People would say well [126:03] the definition of ehr just changed. So [126:06] actually there were a lot more than 13% [126:07] that had it. But that's the best data [126:08] that we have is that fair enough. Okay. [126:11] Now before we get to the actual [126:13] legislation and what happened there's [126:14] one other concept to sort of have in [126:16] your mind which is interoperability [126:18] because this word comes up every time [126:20] Epic comes up or any EHR comes up and [126:22] it's worth knowing sort of the buckets. [126:25] So there's the first and easiest [126:27] interoperability which is Epic to Epic [126:29] at a different hospital. I want to [126:31] transfer my records from one hospital to [126:32] another. They also have Epic. It's the [126:35] same technology that should transfer [126:37] easily. Two is epic to another EMR [126:41] that's at a different hospital. Epic to [126:43] Cerner, Epic to Metate, Epic to All [126:45] Scripts, Epic to some Home World system. [126:48] You can imagine the technical reasons [126:51] why that would be harder. They have [126:53] different architectures. All the other [126:54] ones arguably except Metatech don't have [126:57] a single database, you know, etc. And [126:59] standards have not been quite as [127:01] standard in this industry as they are in [127:03] SAS software easier B2B software mode. [127:06] Now, before we get to the third [127:08] category, let's just think about some of [127:09] the incentives that are probably at play [127:11] here. Epic transferring to themselves, [127:14] fine, easy. A hospital transferring to a [127:16] different hospital, you can imagine [127:18] these places are competitors. They're [127:20] businesses. They may not want to support [127:22] that unless it's really in the patient [127:24] interest, but they're at least going to [127:25] be sensitive to it. Epic to another EMR [127:28] at another hospital. Well, not only the [127:30] hospitals maybe don't want that. Maybe [127:32] Epic doesn't love that either, right? [127:35] interests are sort of aligned against [127:36] this happening, right? Epic doesn't want [127:38] that and the hospital doesn't want that. [127:40] And you know, they talk about a lot of [127:41] good reasons why, hey, actually, we do [127:43] want that, but let's just call a spade a [127:44] spade and say there's no business reason [127:46] why Epic would love for that to happen [127:49] other than of course it's in the [127:50] patient's interest, and everybody really [127:52] should do what's in the patient's [127:53] interest because we all get care from [127:55] multiple places. Then there's this third [127:57] category, Epic to third-party [128:00] application that wants to use data from [128:04] Epic or sit on top of Epic or [128:06] interoperate somehow with Epic. As a [128:10] company, they have been very very [128:12] careful about this one over the years [128:14] and for good reason. I mean, you don't [128:16] want patient data to leak. Epic has [128:18] publicly referenced the fact that they [128:19] have never had a Cambridge Analytica [128:21] situation because they're very very [128:23] careful about sharing data with other [128:26] application developers. But the result [128:28] of this is that it's been much harder to [128:29] integrate with Epic as an application [128:31] developer than you would be used to in [128:33] any other software category. Yep. And [128:36] again, very justifiable reasons why that [128:39] should be the case. We're dealing with [128:40] HIPPA data here. Yeah. But also [128:43] massively in Epic's interest for that to [128:45] be the case. It's very convenient as a [128:47] strategy. Yes. So, as for that first [128:49] scenario, Epic to Epic, it's actually [128:51] awesome. They have this thing called [128:52] Care Everywhere. Today, there are 20 [128:54] million patient records exchanged daily. [128:56] I used it as a part of prepping for this [128:58] episode to join all my My Chart accounts [129:00] across the three health systems I have [129:02] in Seattle. It just seems to work pretty [129:04] well. There's a great story of how it [129:06] actually happened and how it came to be. [129:08] Judy made the call personally that Care [129:11] Everywhere wasn't a thing or shouldn't [129:13] be a thing where their customers could [129:15] pick and choose. If you were enabling [129:17] it, it worked across all Epic customers, [129:20] even if they were your local competitor. [129:22] And so, here's the story. One of the [129:25] first customers to accept that new [129:27] software that came with Care Everywhere [129:30] unknowingly agreed to the [129:31] interoperability feature. And he later [129:34] admitted that he would have declined it [129:35] if he realized that it was included in [129:37] the contract that he was signing. Yeah. [129:38] This is the CEO of that health system [129:40] later admitted that. Yeah. Of that [129:41] hospital. And Judy described it as pure [129:44] luck that Epic was actually able to move [129:46] ahead. And so after that win, they then [129:49] made it mandatory for all of their [129:51] customers and they retrofitted old [129:53] versions of the software to support care [129:55] everywhere. So every Epic customer in [129:57] the US has it and can share with every [130:00] other Epic customer in the US. So this [130:01] is kind of their talking point of hey [130:04] actually we have lots of [130:05] interoperability. We launched this thing [130:07] called care everywhere. It shares [130:09] millions and millions of records all the [130:11] time. Yeah. They talk about hey we are [130:13] the biggest sharer of medical record [130:16] data of anybody. Right. So coming into [130:20] 2008, that's sort of the historical [130:23] baggage that you should know about in [130:25] Epic land of they believe they're doing [130:28] a lot of interoperability in the way [130:30] that they like to do it. Almost no one [130:32] is great at interoperability. I saw some [130:36] transcripts of talking with hospital [130:37] administrators who were like [130:39] interoperability in this industry is [130:40] just laughable, period. No one's [130:42] incentivized to share with each other [130:44] unless they absolutely have to. And it's [130:45] not easy and these are incredibly [130:47] complicated systems and there's lots of [130:48] risk. One CIO put it to me in an [130:51] interesting way when I was chatting [130:52] about this with him because it's patient [130:55] data and health data. There's this sort [130:57] of we feel like interoperability should [131:00] be a thing. You should be able to share [131:01] your data everywhere and like sure of [131:03] course you should. I don't want to [131:04] discount that whatsoever. However, he [131:06] was like imagine this were a different [131:08] industry. Imagine this were the airline [131:11] industry. Does United share their [131:14] customer data with Delta even though [131:16] it's customer friendly to do so? You [131:17] know, even though it's the customer's [131:19] data, no, of course not. They never [131:21] would. So, do I feel great about sharing [131:24] my patients data with my competitor down [131:27] the street? No, of course not. Right? [131:29] This is where you really start to feel [131:31] the friction. I'm sure you're listening [131:34] to this and you're emotionally getting [131:35] charged up at this point because you're [131:38] feeling your inner capitalist think, [131:41] well, it's a business. They should do [131:43] the things that make sense for them as a [131:45] business. And you're feeling your inner [131:48] human being who is part of a society [131:52] that has your own health needs thinking, [131:54] well, maybe but this thing shouldn't be [131:56] subject to the bad parts of capitalism. [131:58] maybe this thing should function [132:00] differently and business interests, it [132:02] feels yucky to me that business [132:04] interests are governing the way that [132:07] things get done. You should feel that [132:09] tension because this is one of those [132:10] bothies. Well, that kind of describes [132:12] the whole healthcare industry, right? [132:14] Make no mistake, this is definitely a [132:15] for-profit industry. Yes. And it's [132:18] highly regulated and there's all these [132:21] issues. Yep. Okay. So, we're into the [132:24] meat. 2008 the great financial crisis [132:27] happens. At some point we should just do [132:29] a whole episode on the great financial [132:31] crisis and the collapse and the [132:33] government bailouts and how we managed [132:35] to figure out how to not collapse as a [132:37] nation and a global we pulled out of it [132:40] and it was amazing. Took some time but [132:41] we pulled out of it. But one element of [132:43] pulling out of it just like during co is [132:47] the government said all right we need [132:49] stimulus baby stuff. We need stimulus. [132:51] We need programs that we can throw money [132:54] at the economy. Interest rates go to [132:56] zero. Money out all over the place. Turn [132:59] on the printers. Figure out how to get [133:01] people doing productive stuff and reward [133:03] them for it. And there's the monetary [133:04] policy stuff of interest rates to zero. [133:06] But there was also the fiscal policy [133:08] stuff of we need to helicopter money [133:10] into the economy. So the question is [133:11] when you're going to helicopter money [133:13] into the economy, you could mail [133:14] everyone checks. And that happened [133:16] during co. Yep. Not a ideal way to do [133:20] it. A more productive thing to do that [133:22] calls all the way back to the Civilian [133:24] Conservation Corps in the FDR sort of [133:26] New Deal era is we want to look for [133:29] shovel ready projects. What are things [133:31] that we just kind of all agree on are [133:34] good ideas for the country that should [133:36] happen and let's just pass some [133:38] legislation that rewards the crap out of [133:41] people with money, with free money for [133:44] doing the things that we think are a [133:46] good idea anyway. The hope is you get [133:49] this double whammy. You get the money [133:51] distributed, which is just a goal. [133:53] You're just trying to create jobs, [133:54] trying to create economic motion, and [133:58] you get something accomplished that is [134:00] widely agreed upon to be a good idea. Oh [134:02] my gosh, a shovel ready project right [134:05] here. Electronic medical records. Let's [134:08] do it. Let's just figure out how to [134:10] finally make that a thing. So the goals, [134:13] one fiscal stimulus to stimulate the [134:17] economy after the great recession, two [134:19] promote the adoption of electronic [134:22] health records, three promote the use, [134:26] which is different, the actual use, the [134:28] meaningful use of electronic healthcare [134:30] records. Not only do we want people to [134:32] become users in your system that never [134:34] log in, we want them logging in all the [134:37] time using the stuff and promote the [134:40] adoption of interoperable standards so [134:43] that providers could share patient data [134:46] nationwide. These are the goals of the [134:50] high-tech act, the 2009 health [134:53] information technology for economic and [134:55] clinical health act. I love how they [134:58] name these things. I know. which gets [135:00] folded into the almost trillion dollar [135:03] American Recovery and Reinvestment Act. [135:06] Yes. So the mechanics in 2009 when the [135:09] bill is passed, there is $27 billion [135:12] available in direct incentive payments [135:15] to hospitals that implemented electronic [135:17] medical records. And if you include [135:19] broader incentives, other health IT [135:21] projects, data exchanges, training, the [135:23] total actually comes to $36 [135:25] billion. This is amazing. Imagine if [135:28] you're a software startup and the [135:30] government just starts mailing checks to [135:33] your customers, not only for signing up [135:36] and becoming a customer and paying you [135:37] money, but then paying them even more [135:39] for actually using your software. It's [135:41] like the best thing that could happen if [135:42] you're a software vendor. This is the [135:44] government going to your customers and [135:47] saying, "Hey, I'm going to give you [135:49] money, but it's like a dedicated use [135:51] credit card. The only thing you can use [135:54] this money for and you must use this [135:56] money for is to buy this software [135:59] startup's products and then you must [136:01] also use them. And if you don't buy it [136:03] and you don't use it, not only are you [136:05] not going to get the money, I'm going to [136:06] penalize and you're going to have to pay [136:08] me money. Yes. Carrot and stick. So [136:11] specifically, it amounted to $44,000 to [136:15] $64,000 per physician in incentive [136:18] payments over a few years for adopting [136:19] it. Now, that actually went to the [136:21] hospital, but it was basically you get [136:22] paid for how many physicians are doing [136:25] this. Yeah. So, imagine a huge system [136:27] like Kaiser with 11,000 physicians. It's [136:30] a lot of money. Yes. High-tech contained [136:33] the phrase meaningful use, which is this [136:36] crazy phrase where if you say meaningful [136:38] use or you even say mu to anyone, it is [136:41] a triggering term in this industry. [136:44] First meaningful use was a carrot and [136:46] then as a stick. So after the stimulus [136:48] ran out, David, as you were saying, [136:49] after about 5 years, health systems [136:52] would face meaningful financial [136:53] penalties for not meaningfully using [136:57] health IT software like EHRs. There's a [137:00] uh KFFF article entitled Death by a [137:03] Thousand Clicks that has a quote on [137:04] this. The EHR vendor community, then a [137:07] scrappy $2 billion industry, griped at [137:10] the litany of requirements, but stood to [137:12] gain so much from the government's 36 [137:14] billion injection that it jumped in [137:17] line. As Rusty France, CEO of EHR vendor [137:20] NextGen Healthcare, put it, and NextG is [137:22] one of the top 10 epic competitors. The [137:25] industry was like, I've got this check [137:27] dangling in front of me, and I have to [137:29] check these boxes to get there. And so, [137:31] yeah, I'm going to do that. For everyone [137:33] that wants to just think this was the [137:35] most amazing thing ever for Epic, yes, [137:37] it was obviously net very good. They had [137:39] to do a bunch of hoop jumping through to [137:43] make sure that it was exactly the thing [137:46] that the bill was going to reward. And [137:48] everybody in this industry had to make [137:50] sure that they spent a bunch of [137:52] development time and a bunch of rep [137:53] prioritization in the company to make [137:55] sure that it was exactly the thing that [137:57] high-tech was saying that it needed to [137:58] be. Yep. It's more nuanced than this is [138:01] really good for Epic. Of course, it was [138:02] really good for Epic, but really it's [138:04] that this was really good for the entire [138:06] EMR industry. The competitive dynamics [138:09] within the industry are almost like a [138:11] separate kind of question. Well, okay. [138:13] So, who's it going to benefit the most? [138:16] Probably the most reliable one. If I [138:18] wasn't going to do this before and now I [138:20] just have a check dangling in front of [138:21] me as a hospital to implement an EHR, I [138:23] want the one with the integration that's [138:25] 100% going to work. I'm taking no risk. [138:27] And also that epic one, I think someone [138:30] told me before that was an expensive [138:32] one. Well, I guess it doesn't matter [138:33] that it's expensive anymore because it's [138:35] free. So, I'm just going to buy the good [138:37] one. Yeah, right. It's effectively free. [138:39] So, Epic by being the high price, highv [138:42] value vendor and the one that you could [138:44] count on working with the lowest risk, [138:47] they were going to win in this [138:48] situation. Yeah, that's a great point. [138:50] And it's like you're getting a stimulus [138:52] to buy a handbag and you could get the [138:54] Target one or you could get the Birkin [138:56] bag and they're all free. And it's not [138:59] exactly like that, but it rounds to [139:00] that. It also is interesting because it [139:03] basically rewards the most reliable [139:05] software, not necessarily the most [139:07] innovative software. You end up not [139:09] taking any risk in situations like this [139:11] because you're not going to get rewarded [139:13] unless you actually stand the system up [139:15] and then start getting meaningful use. [139:17] So you end up with the one that's just [139:19] going to work. So, what did high-tech [139:20] actually do? We'll go through sort of [139:22] each of the goals. Did it accelerate [139:24] usage? Absolutely. Market penetration of [139:27] electronic healthcare records went from [139:29] 9% of hospitals in 2009 to 95% by 2014. [139:34] That's insane. David, what other markets [139:36] have you ever heard of that saw usage go [139:39] from almost nobody to almost everybody [139:41] in 5 years? [139:44] Uh, certain categories of software [139:46] during the pandemic. Yeah, that's a good [139:48] point. I would imagine maybe Zoom during [139:50] the pandemic is the closest thing. I [139:52] mean to be an Epic employee at this [139:54] moment where suddenly your category is [139:57] just free for all your potential [139:59] customers. It's crazy. And digitization [140:01] was great for patients. Even though like [140:04] a lot of doctors will complain about it, [140:06] it is totally amazing to be able to [140:07] message your doctor about something and [140:09] pull up your own records electronically. [140:11] As we talked about with my chart, [140:12] schedule your own appointments. You have [140:14] all of the above. Manage your family's [140:15] care. Yes, absolutely. Okay, so then the [140:18] next goal, cost reduction. There are [140:20] stats both ways on if EHRs generally [140:23] reduce costs. We found one stat that [140:25] claimed that costs in hospitals went [140:27] down by 10% with the introduction of [140:29] EHRs. There's also arguments that the [140:31] adoption of EHRs may actually have [140:33] accelerated the ordering of unnecessary [140:36] care as well as increased billing codes [140:40] for either more codes or higher dollar [140:42] codes for the same procedures. So that's [140:44] sort of the counterargument to did it [140:46] decrease costs in the system. The [140:49] critics of EHRs say that basically more [140:53] stuff is getting ordered from exam rooms [140:55] for the exact same procedures than was [140:57] happening before. And frankly that's the [141:01] goal of of hospital buying an EHR. I [141:04] mean part of it is revenue maximization [141:06] right? Again, these are commercial [141:09] entities. You know, even though many of [141:10] them may be nonprofit organizations or [141:13] part of a university system or what have [141:15] you, at the end of the day, this is [141:18] revenue being generated by a hospital [141:20] with a management team and with its [141:24] primary staff being doctors and nurses [141:26] who like to make money. The reason you [141:29] go to medical school to be a doctor or [141:31] nurse is to get a good job and make a [141:34] good living and the prestige of it. and [141:36] you want to help people, but it's a [141:37] highpaying job. Yep. So, there's a great [141:40] read called An Epic Dystopia, and I will [141:42] say this article presents kind of a [141:44] one-sided view of things. This is the [141:45] American Prospect article. Yep. But [141:47] there's a quote, one doctor, for [141:49] example, stated her supervisor regularly [141:51] contacted her and said something like, [141:53] "That appointment was a two. Don't you [141:55] think it might be a three?" And [141:57] obviously, in this case, a three could [142:00] bill more to insurance than a two. or [142:02] someone else gave me a quote. Hey, if [142:04] you did X, almost certainly you did Y in [142:06] that appointment. If you use those [142:08] codes, you could charge more. And if you [142:10] have a system, a computer system versus [142:13] not having one, it's more likely that [142:15] you are billing for more stuff. That [142:18] would be the counterargument to this [142:20] mass implementation of EHR's saving [142:23] costs for the system. Yep. Data [142:25] interoperability, not really. The [142:28] legislation was prescriptive about [142:30] meaningful use. It really wasn't [142:32] prescriptive about data standards that [142:34] people talked about. That's a goal of [142:35] the legislation, but it's not like there [142:38] were incentives for data [142:40] interoperability their way there were [142:42] for meaningful use. And so people follow [142:45] incentives. The data interoperability [142:47] did not really happen which was very [142:49] convenient for Epic who had a whole [142:51] system that they made themselves and [142:52] they didn't really need to integrate [142:53] with anyone to make the software suite [142:55] useful. So again, as the leading player [142:57] in the industry and the one that can [142:59] kind of do everything themselves, they [143:01] didn't need interoperability and it [143:02] wasn't explicitly rewarded. So customers [143:05] didn't prioritize it either. And there [143:07] is a fact pattern that opens Epic up [143:11] certainly to criticism here, which is [143:12] that Judy was on Obama's health IT [143:15] council through all of this, which was [143:17] advising the Obama administration on the [143:20] high-tech act. And other competitors had [143:23] people on other committees, too. It's [143:24] not like she was the only one with a [143:25] voice in government. Right. I was going [143:26] to say that is a convenient fact that [143:28] gets left out by a lot of people making [143:30] that argument or that all the [143:32] competitors, Cerner, all scripts, etc. [143:34] They were all on those councils too. [143:36] Yep. But let's just look at the data [143:39] interoperability thing. It didn't [143:41] happen, but it's happening more now 15 [143:45] years later. And so at least we have the [143:47] foundation of digital records so that as [143:50] pushes and shves happen to make it more [143:53] interoperable. Without operability, you [143:55] can't have interoperability. And I think [143:56] for me that's kind of my takeaway of the [144:00] high-tech act and meaningful use here, [144:03] at least when it comes to the industry [144:05] of EHRs. Yeah. Anytime you've got [144:08] government coming in and distorting a [144:10] market, you're going to get weird stuff [144:11] happening. And especially anytime you [144:14] have government coming in and regulating [144:16] how products are to be developed and [144:18] used. Yes. In a market, you're going to [144:20] get really messy stuff happening. Okay. [144:23] So, there were two really bad side [144:25] effects here. One, by literally defining [144:28] what is meaningful use. Congress not [144:30] only accidentally designed the software [144:32] and specified the features, but they [144:34] also accidentally told doctors how they [144:36] needed to do their jobs. And so doctors [144:39] are like running around now clicking [144:41] nine things that they never needed to [144:42] click before. Some of which are poor [144:45] implementation by their hospital system. [144:47] Some of which is because the software is [144:49] really complex. But a lot of it is like [144:50] just making sure that this is what the [144:52] legislation requires. Yes. Complies with [144:55] meaningful use so that the hospital then [144:58] doesn't come and get hit with a [144:59] financial penalty. Yeah. This is sort of [145:01] the most unfortunate aspect of [145:03] meaningful use. I think I think it's the [145:05] second most. I think the biggest one is [145:06] they increase the regulatory burden of [145:09] practicing medicine period. So now since [145:11] hospitals can face big financial [145:13] penalties, there's all these mandatory [145:16] fields and workflows everywhere forever [145:18] even after the stimulus runs out. And so [145:21] it just generally increased the [145:22] operational overhead of the whole [145:24] industry. The thing to optimize for now [145:26] is hitting the MU definitions legally [145:28] rather than the spirit which is help [145:30] doctors and patients get more out of the [145:32] system. And you end up with things like [145:34] health systems merging with other health [145:37] systems because the cost of doing [145:38] business is higher, which is a [145:39] pre-existing trend, but it's certainly [145:41] an accelerant. There's more overhead, [145:43] more burdensome regulation, just more [145:46] crap in the system, more waste. So all [145:49] that unquestionably bad at the same [145:52] time, what's the alternative? We going [145:53] to be in the paperbased system? Like [145:55] that's not good. At least we have [145:56] adoption here. At least we have adoption [145:59] here is what a lot of people have sort [146:01] of chocked this up to. It's definitely [146:03] better to be digital. Is this the best [146:05] form of digitization that could have [146:06] happened? No. Is digitization overall [146:10] good? Yes. And I think if you zoom way [146:14] way way out, you give the most credit to [146:17] the government and the legislators who [146:19] created all this. Was the goal actually [146:22] a jobs program and an economic stimulus [146:25] for the country? If that was the goal, [146:28] A+. A lot of jobs got created because of [146:31] this. 18% of our GDP's worth of jobs are [146:35] in this field. So it's now a huge part [146:38] of our society is people who work in [146:41] medicine for better or for worse and [146:43] related fields to medicine like IT [146:46] administrators, you know, epic [146:47] administrators at hospitals. These are a [146:50] lot of jobs and these are a lot of good [146:51] jobs. And play it one step forward. [146:54] These are all domestic American jobs not [146:57] dependent on imports, not dependent on [147:00] manufacturing that have created viable [147:03] great career paths for a lot of people [147:05] in this country from a policy [147:07] standpoint. Like maybe not bad. N I hate [147:11] it. David, I'm not saying I love the [147:13] idea of jobs programs. I'm just saying [147:15] if you zoom way out of what was the goal [147:17] of government in doing this, did they [147:18] succeed? If that was the goal, they [147:19] succeeded. Yeah, it was good to create [147:22] fiscal stimulus for that period of time [147:24] where we really needed it. Should the [147:26] government be in the business on a [147:27] durable basis of using taxpayer funds to [147:32] prop up industries and create and [147:34] maintain a bunch of jobs in those [147:35] industries? No, absolutely not. And so [147:38] it's a very low bar what you're saying. [147:40] Did it stimulate the economy? I mean, [147:42] mailing checks would have stimulated the [147:44] economy. I look at this as like, yes, it [147:45] stimulated the economy and it had the [147:47] free stapled coupon of we got some [147:50] amount of electronic medical records. [147:52] Even though they're not interoperable [147:53] and even it's not the best system we [147:55] possibly could hope for as a country, [147:57] it's a better one than we would have had [147:59] otherwise. And it kind of came for free [148:01] with the stimulus checks. Yep. So, back [148:05] to the epic story here. Yeah. So, what [148:07] did meaningful use do for them? Well, [148:10] interestingly, if you just take a step [148:12] back and say what is the byproduct of [148:14] the government paying off a whole [148:16] industry to adopt something new, what [148:19] you're basically doing is pulling [148:20] forward the future and saying this was [148:22] all going to happen eventually. We want [148:24] to make it all happen right now. And to [148:26] be clear, I think especially after the [148:28] Kaiser win, Epic was going to become the [148:31] dominant player anyway, absent of [148:33] meaningful use. They were starting to [148:35] run the table on competition. So, that's [148:37] the question. they were starting to. But [148:38] what you definitely do when you throw [148:41] $36 billion at the problem is you say, [148:44] "We're going to close the door on anyone [148:47] that might become a dominant player [148:49] after this." The current competitive set [148:52] is now what we're dealing with. And we [148:55] think the current competitive set is [148:57] good enough that this is shovel ready [148:59] enough that we want the best ones to get [149:02] implemented everywhere. These are stuck [149:04] in there for multiple decades. So, yeah, [149:06] the switching cost here is enormous. [149:08] Yes. And that's a fair trade-off. I [149:10] don't even think it was that intentional [149:12] of a trade-off that they made by passing [149:14] this legislation, but it is definitely [149:16] true that a new EHR that may have been [149:19] more theoretically innovative in some [149:21] way that could have come after high-tech [149:24] wasn't going to happen, at least for a [149:26] long time. Yep. Because you pulled [149:29] forward so many RFP processes to a [149:32] single point in time. Yes. and the [149:35] majority of them chose Epic, which they [149:37] probably would have anyway, but a lot of [149:39] those processes would have happened in [149:41] the future, at which point in time other [149:43] competitors may have emerged, right? [149:45] There's no green field bidding for those [149:48] theoretical future innovative [149:49] competitors that everyone's bidding [149:52] against your current system. I buy that, [149:54] but that's kind of nitpicky, right? It's [149:56] a little bit of like a for whom are we [149:58] holding the door open. So, if you're [150:01] looking for a smoking gun on regulatory [150:03] capture, I think the right way to [150:05] characterize this is as regulatory [150:07] tailwind rather than regulatory capture. [150:10] I don't think the folks at Epic loved it [150:13] either. My impression is they felt like [150:16] they were going to take this market and [150:17] now they had to do it in sort of this [150:20] Frankenstein style way. Well, it [150:21] definitely made the product worse. Yeah, [150:24] it made the product worse. It increased [150:26] burdens for everyone. you know, it made [150:28] it more sure that they were going to win [150:30] the market because if something's going [150:32] to happen five years from now or it [150:33] could happen now, you'd rather it happen [150:35] now and it be certain. But they [150:37] certainly don't love it that doctors [150:38] have to check 57 boxes on every patient [150:42] visit, right? So all this compliance [150:44] burden and the software being frankly [150:46] hard to use once it's legislated. Yeah. [150:49] Product development by legislation, [150:51] which to be clear isn't just Epic, it's [150:53] all of them. And also just the [150:56] complexity of software of this magnitude [150:59] with this many different screens and [151:00] participants and all that. A big [151:02] criticism is that it takes time away [151:05] from patients. There was a 2016 study [151:07] that showed that entering data into EHRs [151:10] consumes about 2 hours of doctor time [151:12] for every 1 hour spent providing [151:14] hands-on patient care. There's all this [151:16] other research around doctors now work [151:18] 11 12 hours a day. They're like [151:20] constantly responding to messages, which [151:22] again is a hospital configuration thing. [151:24] There's a way to triage messages. [151:26] Another downside to the high-tech act [151:28] and meaningful use is that before [151:31] digitization and moving all this [151:32] activity from hospitals into the [151:35] EHR, there were a lot of regulations and [151:39] guidelines that maybe weren't always [151:42] followed and that was sort of for the [151:44] best in the hospital. So things like [151:46] it's supposed to be doctors themselves [151:48] who put in orders for pharmacy for meds [151:51] or for scans or stuff like that. But in [151:54] the old world before everything moved [151:56] into the EHR, doctors were supposed to [151:59] do it, but they would have their medical [152:00] assistants do it. If you're seeing 20 [152:02] patients a day and a whole bunch of them [152:04] need scans or meds or whatever, you can [152:07] just be like, "Hey, medical assistant, [152:09] do all this." And they can do it during [152:10] the day. Yeah. I'll sign them all later. [152:12] Exactly. But now once everything's [152:14] digitized and in the EHR, that means [152:16] that the rules have to be followed [152:18] because they're tracked, right? And so [152:19] that means that all of a sudden doctors [152:22] don't have as much help and slack in the [152:24] system as they used to. Yeah, it's [152:26] interesting. So this is the sort of [152:28] thing that ends up contributing to [152:29] doctor burnout. Yep. But you do have to [152:32] compare it to how it used to be. There's [152:34] a great old study from 1970 that found [152:36] that communications activities such as [152:39] managing physical records accounted for [152:41] 35 to 39% of total hospital operating [152:44] costs. That's not the doctor's time, but [152:47] especially in the paper world, it was [152:49] always a huge amount of cost and time [152:51] from the hospital, right? You're [152:52] shifting around the burden of this a [152:54] little bit and creating more burden [152:55] through all the meaningful use [152:57] regulations, but this is not a new [152:58] problem. You know who else doesn't love [153:00] it? So, in 2017, Obama himself told Vox [153:03] that he felt that the high-tech [153:05] legislation did not live up to what he [153:06] wanted. And here's his quote. The fact [153:08] that there are still just mountains of [153:10] paperwork and the doctors still have to [153:12] input stuff and the nurses are spending [153:14] all their time on this administrative [153:15] work. We put a big slug of money into [153:17] trying to encourage everyone to digitize [153:19] to catch up with the rest of the world [153:21] and that's been harder than we expected. [153:23] The quote, I actually don't think it was [153:25] at all about catching up with the rest [153:26] of the world on digitization. I think we [153:27] were a leader there. But if the hope was [153:29] that we could somehow downgrade our [153:31] costs to be in line with the rest of the [153:33] world, that would have been great. And [153:34] that did not happen. Yep. One of the [153:36] CIOs I talked to had a great quote on [153:39] this. He said that meaningful use and [153:42] the high-tech act wildly succeeded at [153:45] digitization of the industry. It did [153:49] absolutely nothing on digital [153:51] transformation of the industry. Hm. We [153:54] digitized, but we didn't transform in [153:58] the way that I think people were [153:59] optimistically hoping for. That's [154:01] interesting. And the good news is that [154:04] can still come now that all the records [154:06] are digital. Yes. Now that we've [154:09] digitized. Yes. You actually can do [154:11] interesting things with the data that [154:12] you couldn't have done before. And we'll [154:14] talk about when we get to the end of the [154:15] story in a couple minutes, the promise [154:17] at least right now of AI and ambient AI [154:21] in making a lot of this ownorous process [154:24] just disappear for humans. Yes. So on [154:28] the back of all of this, Epic continues [154:31] to win epically. In 2011, they hit a [154:35] billion dollars in revenue. They keep [154:37] winning all the big systems that come up [154:39] for RFP. John's Hopkins, Cedar Sinai, [154:43] UCSF, list goes on and on and on. For [154:47] Epic's competitors, they had always been [154:50] merging and consolidation had always [154:52] been happening. But now with meaningful [154:54] use being in place, consolidation really [154:57] picks up, which again, all of which is [154:59] just going to acrue to Epic's advantage [155:01] because consolidation in this space [155:03] means the products get more complex. [155:04] They don't work as well altogether as a [155:06] suite and Epic's advantage just becomes [155:08] all the more pronounced. So 2008, All [155:11] Scripts merges with Micus. 2010 that new [155:15] All Scripts entity then merges again [155:17] with Eclipses. They keep on acquiring [155:20] more smaller players. Today that company [155:22] is called Veradime. 2011, Metitech [155:25] acquires LSS Data Systems. And then in [155:28] August 2014, one of the big ones that [155:30] we've already alluded to happens. Cerner [155:32] buys Seaman's Health for $1.3 billion, [155:36] which that starts the downward spiral [155:39] for Cerner. That's my understanding. [155:42] Yes, I heard in the research that even [155:45] today in 2025, customers will still [155:48] refer to themselves as either Seaman's [155:50] customers or Cerner customers. It's [155:52] still not fully integrated. Meanwhile, I [155:55] think post Oracle acquisition, Oracle's [155:57] trying to rewrite it all anyway, right? [155:59] And that's uh not going well, which [156:01] we'll come to later. Yep. So, this leads [156:05] us right into the Department of Defense [156:09] contract. The motherload of all [156:12] opportunities if you are an EMR VA [156:15] contract. Oh boy. The motherload of all [156:18] opportunities. Or if you are an American [156:20] taxpayer, the motherload of everything [156:23] that is wrong with governmental waste y [156:25] in America. [156:27] But before we get into that story, we [156:29] have some happier and better news to [156:31] tell you. Yes. And that is to thank [156:33] longtime friends of the show, Crusoe. [156:36] Yes. Thank you, Crusoe, for giving us a [156:38] moment of levity here and a moment of [156:40] excitement and optimism looking to the [156:42] future. Crusoe is not just an AI cloud [156:44] platform. They are actually reimagining [156:46] how AI infrastructure gets physically [156:48] built from the ground up. Yep. In the [156:51] past, we've talked about Cruso's climate [156:52] aligned approach powering their GPU data [156:55] centers with stranded energy. But what's [156:57] truly unique about Crusoe is that [156:58] they're a fully vertically integrated AI [157:01] infrastructure business. So Cruso's [157:03] founders, Chase and Cully, realized [157:05] early on that in order to become the [157:06] best AI cloud platform, you actually [157:09] need to build the entire stack from [157:10] sourcing the energy to building the data [157:12] centers and ultimately providing the end [157:14] AI cloud service layer. So the best [157:16] example of this is the giant 1.2 2 [157:19] gawatt AI data center that they are [157:21] building in Abalene, Texas. They broke [157:23] ground in just June of 2024 on an empty [157:26] field and now within a year they are [157:29] about to turn on the first few [157:30] buildings. Bringing on new capacity of [157:32] this size in less than a year is crazy [157:35] fast for the industry. Yep. And to put [157:37] that 1.2 2 gawatt in perspective. If you [157:40] take the entire data center capacity [157:42] located in Northern Virginia today, so [157:45] where like 70% of the world's internet [157:47] traffic flows through, all of that is [157:50] only 4.5 gawatt. So Cruso's single AI [157:53] factory in Abene will be more than a [157:55] quarter the size of what powers a huge [157:58] portion of the internet today. And [157:59] that's just one of their locations. [158:01] Yeah. Crusoe handles all of the [158:03] incredible complexity of building [158:05] physical AI capacity so that their [158:07] customers can just focus on innovation, [158:09] not infrastructure. And they just [158:11] announced two new services to do that. [158:13] Their managed inference service frees [158:15] developers from managing servers and [158:18] their autocluster service eliminates the [158:20] operational burden of model training. [158:23] Yep. So when you're thinking about where [158:24] to run your AI workloads, make sure you [158:27] consider Crusoe because they operate and [158:29] innovate at every single layer of the [158:31] physical stack. Cruso delivers unmatched [158:33] price to performance, speed to market, [158:35] and a platform that customers can rely [158:36] on for years to come. It's an incredible [158:38] company, and Ben and I are excited to be [158:40] investors in their recent funding round [158:42] alongside Nvidia, Founders Fund, and a [158:44] whole slate of other great investors, [158:45] too. That we are. To learn more, head on [158:47] over to [158:48] cruso.ai/acquired. That's [158:51] crusoe.ai/ [158:52] ai/acquired or click the link in the [158:54] show notes and when you get in touch [158:56] just tell them that Ben and David sent [158:58] you. Okay, David, the DoD debacle. The [159:02] 2015 DoD debacle. So, the Department of [159:06] Defense decides that they're going to [159:08] bid out a global EHR contract for all of [159:12] their hospital and health care across [159:15] all branches of the military. Now, this [159:17] is at the beginning just active duty US [159:20] military. So, separate from the VA [159:23] system which comes later, the Veterans [159:24] Administration, all the veterans of the [159:26] US military, this deal in 2015 is up [159:32] until this moment in time the biggest [159:34] healthcare IT contract of all time. It [159:37] ends up being a $4.3 billion deal. So, [159:40] even bigger than the Kaiser deal, it [159:43] remains the biggest until two years [159:45] later when the VA contract does go out. [159:48] That is a $10 [159:50] billion [159:52] deal. Now, Epic, of course, bids on both [159:55] of these contracts, and as you would [159:57] expect, just like with Kaiser, it comes [159:59] down to Epic and Cerner, but not [160:01] directly because just to add to the [160:04] insanity of this system, it's basically [160:06] impossible unless you bid on government [160:08] contracts all the time to bid on a [160:09] government contract directly. So, you [160:11] need to pony up with one of the [160:13] companies that subcontract off of [160:16] someone who makes their bread and butter [160:18] all the time by contracting with the US [160:20] government. Yes. So, they each choose [160:22] their partners and there's all sorts of [160:25] drama around this. Basically, the knives [160:27] come out on all sides. As we shall see [160:30] with Cerner, if you win one or both of [160:32] these contracts, I mean, this will keep [160:34] your company going for years and years [160:35] and years. So, drama aside, Cerner ends [160:40] up winning both of these contracts, the [160:42] DoD and the VA. And sadly, in what will [160:47] perhaps not shock you at all, both of [160:50] these projects, despite being huge at [160:52] the bidding phase, go massively over [160:55] time and budget. It's really, really [160:57] bad. So the DoD system, the active duty [161:00] system only just fully went live late [161:04] last year in 2024. This contract got bit [161:07] out in 2015. So that is a 9year [161:10] implementation process. The VA system is [161:15] way worse. It is nowhere near fully live [161:19] today. And Cerner won both of these. [161:21] Cerner won both of these as a [161:23] subcontractor. Won both of these. The VA [161:25] system was bid out in 2017. The latest [161:28] announcement from the VA about this [161:30] project is that it will be live at all [161:33] VA sites by, this is a direct quote, as [161:36] early as [161:38] 2031. What? So even the most optimistic [161:42] scenario that they're saying, which [161:44] seems very unlikely, is this is a [161:47] 14year project. The latest I had seen is [161:50] that it's still not live. They actually [161:51] gave an as early as 2031. Now it is live [161:54] at a few sites. It's not like nothing [161:57] has happened over the last decade, but [162:00] oh my god. I mean, if as an American [162:03] citizen or watcher of America, you were [162:06] disgusted by the whole meaningful use [162:07] thing. If you're an American taxpayer, [162:10] this is just beyond disgusting. So why [162:13] what took so long or what made it so [162:15] expensive? I don't really know. I think [162:18] it's a combination of a lot of things. [162:19] Or I guess what are the incentives? [162:21] Well, it's really unfortunate, but [162:24] there's an old saying that you make more [162:27] money on failed government contracts [162:28] than successful ones. And I think that's [162:32] probably the incentives that are [162:34] applying here. Oh, cuz as long as the [162:37] contract doesn't end and you're still [162:39] implementing, you can still keep finding [162:41] more costs. You can still keep getting [162:44] paid. Now, I think it would be unfair to [162:47] blame Cerner for this. I mean, maybe [162:50] some of the blame lives on them, but [162:52] they are a subcontractor to big [162:55] government primes that are the GC here. [162:57] I mean, we talked about this on the [162:59] Lockheed episode. This is just how the [163:00] military-industrial complex works now. [163:03] So, you've got the compounding layers of [163:05] bureaucracy of the government, the [163:08] military, EMRs generally, as we've [163:11] talked about. Like, this is just like a [163:13] rat's nest of awfulness of bureaucracy. [163:17] And you know, you've got these [163:18] multi-deade now projects with many [163:21] billions of dollars of taxpayer dollars [163:24] going into this. Still not live. Just [163:26] freaking brutal. Do you think Epic is [163:29] glad they didn't win? Well, so here's [163:32] the thing. Epic, of course, was part of [163:34] this process, bid on the RFPs, and I'm [163:36] sure would have loved to have won them. [163:37] Cuz even the $10 billion initial price [163:40] tag before the overruns is twice Epic's [163:43] revenue today. is transformative for [163:46] your business in the best case? Yes. [163:49] Kind of like Kaiser was transformative [163:50] for Epic back in the day. Yeah, that's [163:53] it. Talking to Epic customers and CIOS [163:55] and the research, they are like down on [163:57] their hands and knees thankful that Epic [164:00] did not win this deal cuz Cerner just [164:03] got dragged so into the process, so into [164:06] the muck here. And there were a lot of [164:08] other compounding factors happening here [164:10] like Cerner like we've talked about [164:12] already was now postsemen's what you say [164:15] 36 companies that had been acquired to [164:18] yeah get pulled together to make Cerner. [164:20] Neil Patterson the founder and CIO and [164:24] longtime leader right around this time [164:26] he gets cancer and then he passes away [164:28] in 2017. So you're a founder and your [164:30] leader is passing away in the midst of [164:32] this very complex process. After that, [164:35] Cerner cycles through a whole bunch of [164:36] different leaders over the next few [164:38] years. And meanwhile, Epic kind of [164:40] unburdened by this DoD and VA shitow for [164:45] lack of a better word. And they just [164:47] keep winning deal after deal in the [164:49] large system providers. And in their own [164:52] way, I mean, this is a testament to how [164:54] Judy thinks a company should be run. [164:56] We're going to stay extremely focused. [164:58] We have a very particular playbook. We [165:00] listen to customers. We meet our [165:02] promises. We're hardcore about our [165:04] software engineering and our product [165:06] architecture. When our customers find [165:08] new use cases, we develop that software [165:10] too. She runs a company in an [165:12] unconventional way and that [165:14] unconventional way is proving to be the [165:17] right way to win this market. Yep. So [165:19] after the DoD deal, Epic wins Partners [165:22] Healthcare in Boston, which is the sort [165:25] of new entity of Harvard and Mass [165:27] General MGH and like the original big [165:30] research institution. They win all of [165:32] the Mayo Clinic, all of their hospitals. [165:34] That's one of the biggest contracts [165:36] again outside the DoD craziness. They [165:38] win Cambridge in the UK. They win Inner [165:40] Mountain Health. Recently, they win [165:42] Common Spirit Health. The list goes on [165:44] and on and on. 2018, Epic hits 2.7 [165:48] billion in revenue. They announced that [165:50] they have all 20 of the top academic [165:52] hospitals in the US News and World [165:54] Report rankings, which is critical for [165:57] them because winning the top academic [165:59] hospitals means that all the new doctors [166:02] and nurses that are coming through [166:03] getting minted by these institutions and [166:05] trained are all trained on Epic. I think [166:08] they sell them a separate educational [166:10] license, too. I think they treat the [166:12] university hospital and the academic [166:15] classroom as separate. I mean, I'm sure [166:18] it helps to be in the hospital, but I [166:19] think you also buy it as a school. I'm [166:22] sure that's right. But also a huge part [166:23] of medical education is practicing in [166:25] the hospital. And I think now 90% of med [166:28] school students train on Epic. Yes, that [166:31] is the current statistic. 2019, Epic [166:34] hits 3.2 billion in revenue. And then [166:37] finally, December 2021, as we've been [166:39] alluding to, Cerner gets bought by [166:41] Oracle for $28 billion. So Cerner, as we [166:45] talked about, was and still is a big [166:47] company. I mean, they've acquired all [166:49] these companies over the years. There's [166:50] a lot of revenue there and they're big [166:53] international and they have the DoD and [166:56] VA contracts ongoing, which was a lot of [166:59] money coming in. So they're doing maybe [167:01] call it 5 a half billion a year in [167:03] revenue, maybe a little more, but that's [167:06] been flat to declining since 2018. [167:09] Oracle again acquired them in 2021. [167:11] Oracle now no longer reports Cerner's [167:15] financials as a separate segment within [167:17] the company. And they talk about Cerner [167:20] and Oracle Health being a quote unquote [167:22] headwind to overall growth and [167:24] profitability for the company. There [167:25] have been a lot of layoffs within Oracle [167:28] Health. You know, maybe it's too early [167:30] to tell, but I think this is probably [167:33] one of Oracle's worst acquisitions of [167:35] all time. M you'd be hardressed to find [167:38] an analyst who would say that this was a [167:40] great acquisition by Oracle, shall we [167:42] say? Yes, to put it lightly. Meanwhile, [167:44] this is all just great for Epic and [167:46] totally reinforces Judy's whole thesis [167:49] of company building in this space and [167:51] her whole story to customers of one [167:54] single integrated platform that is [167:57] hardcore about software development that [167:59] is 100% customer focused that will never [168:02] acquire another company, never go [168:04] public, never get acquired. You can see [168:06] why it's just this warm hug embrace to [168:09] their customers out there. Yeah, it [168:11] becomes more and more different than [168:13] everyone else in the market the longer [168:15] time goes on. And the lesson here, which [168:17] is really interesting, is they started [168:20] slow. They started really slow and they [168:23] did things in a way where they almost [168:26] built up momentum for the future in a [168:29] way that would pay off 20, 30 years down [168:32] the road rather than inorganically [168:35] trying to take shortcuts and pull the [168:38] future too far forward into today. the [168:41] first 20 years kind of looks like an [168:43] unimpressive company because they were [168:45] sort of just building strength, building [168:47] muscle, growing the way that you need to [168:49] grow in order to be as bulletproof as [168:53] they are today. That's such a good [168:55] point. And I was trying to square in my [168:57] mind earlier when we were telling the [168:59] meaningful use story how on the one hand [169:02] it really was obviously good for Epic [169:04] and an accelerant and it further [169:06] cemented the market dynamics where they [169:08] were already emerging as the leader. On [169:10] the other hand, when you talk to people [169:12] within Epic, they're very mixed on [169:14] meaningful use. Yeah. And I could never [169:16] figure out if that was like lip service [169:18] of, oh, it didn't help us that much. But [169:20] when you look at the graphs of customer [169:22] counts and of revenue, it was going [169:24] great. Yes, it did pull forward some [169:26] growth, but the trend lines were going [169:28] to get here anyway. It's not like it was [169:30] some big massive step function in two [169:33] years. That is not how the graphs look. [169:35] As we told the story, they had the [169:37] better product, they had the better [169:38] customer offering, they were on the way [169:40] to winning anyway. I think to your point [169:43] here about they've always been just [169:45] philosophically allergic to artificial [169:49] growth. I wonder if that's part of why [169:51] to the extent they are serious about [169:52] really being ambivalent about meaningful [169:54] use that's the reason why yeah they want [169:56] to win slowly. They absolutely want to [169:58] win and they're going to win but they [170:00] want to do it slowly because it will [170:02] help them win harder in the long run if [170:04] they win slowly. Yes. Yes. [170:08] So meanwhile for them, I mean, I think [170:11] things have never looked brighter [170:13] basically. I mean, there's some risks [170:14] that we'll talk about, but they're now [170:17] starting to sell international and build [170:20] that as a real market. It's not huge, [170:22] but it's, you know, call it 10 to 15% of [170:25] the business now. 2023, they went live [170:28] at London's Guys and St. Thomas' NHS [170:31] Trust, which was, I think, their biggest [170:35] ever single implementation and happening [170:38] in the UK. The UK is becoming a real [170:40] market for them. Interesting. You know, [170:43] there's another big tailwind that's [170:45] happening for them. And like you've [170:47] probably been hearing from me over the [170:48] course of this whole episode, there are [170:50] tailwinds that don't make you feel good, [170:52] but they're still tailwinds for Epic's [170:53] business. the acceleration of the trend [170:56] where hospital systems are merging and [170:58] rolling up all the smaller local [171:00] practices. Oh, totally. It's just an [171:02] excuse to rebid. Yes, there's a bunch of [171:04] reasons this is happening, but one of [171:06] them is the Affordable Care Act that [171:08] also increased the compliance burden of [171:10] practicing medicine, which was already [171:12] really heavy from all the regulations [171:14] over the years, including high-tech [171:16] adding big penalties for HIPPA for [171:18] violating HIPPA. That's another thing we [171:20] didn't talk about in high-tech. That was [171:21] one of the things that sort of came [171:22] stapled onto it was even more strict [171:25] oversight on HIPPA. So you now require [171:28] scale as a local medical practitioner to [171:32] be able to afford the compliance burden. [171:34] And what that means if you play it out [171:36] is all health care or most health care [171:39] gets provided by these mega giant [171:42] hospital systems rather than community [171:45] clinics. which really benefits Epic cuz [171:47] their strategy from 30 years ago of [171:49] going over the biggest and most complex [171:52] health systems now looks totally genius [171:54] cuz those are the only customers really [171:56] left standing. Yep. Totally. I actually [171:59] think this is one of the most messed up [172:00] things in the US healthare system. Aside [172:03] from the really big hard to negotiate [172:07] with private health insurers who have [172:10] gotten a lot of power over the [172:11] ecosystem, you now have also really big [172:13] local pseudo monopoly hospital systems. [172:16] And so this consolidation is not good. [172:19] This consolidation will not lead to [172:21] prices coming down. Let's just put it [172:22] that way. Yep. But you see why it's been [172:25] happening. You and I entered doing this [172:26] episode expecting to be like, "Ah, the [172:29] hospitals are really the bad guys. [172:30] They're charging so much money." I don't [172:32] think the hospitals make very much [172:33] money. They don't make much money at [172:35] all. They're barely surviving, which is [172:37] why they're merging, right? And any [172:39] money that the very successful ones are [172:41] making, they tend to plow back into [172:42] growth because, like we've been talking [172:44] about, hospitals kind of need scale in [172:47] this day and age. And so, it's new [172:49] buildings, it's acquiring other [172:51] practices locally. Everyone is just [172:54] chasing scale because they constantly [172:57] are having to do business with other [172:59] scale players. So they need to size up [173:01] to do that. Yep. So that brings us to [173:05] the other big I think potential growth [173:08] vector for Epic going forward that [173:10] they're absolutely pursuing is they have [173:13] enough scale on the provider side now [173:16] that they can start going to the other [173:19] big players in the system, the payers [173:21] and then also the biotech and pharma [173:24] side and say we actually have ways that [173:27] we can work with you guys now and offer [173:30] compelling products to you. So they call [173:32] this the system of connectedness or the [173:34] quoteunquote grid of care that you might [173:36] sometimes hear from Epic. But this is [173:38] building products for and working with [173:40] payers, the insurers and pharma [173:43] companies and other stuff like home [173:44] health companies or post-accute rehab [173:46] etc. But really the big opportunities [173:48] are payers and pharma. One example of [173:51] this is prior authorizations. Yep. Prior [173:55] authorizations are a huge problem in the [173:58] American health care system. This is [174:00] when a patient needs or a doctor thinks [174:04] that a patient needs either a medicine [174:06] or a procedure that the payer, the [174:09] insurance company says, "Well, you can't [174:11] just go do that. It's expensive enough [174:13] that you need to ask us first on a [174:15] case-by case basis if we'll cover it for [174:17] this person." Yeah, exactly. We need you [174:19] to prior authorize this. Epic now has [174:22] enough data and enough scale on the [174:24] hospital side that they can go to the [174:26] payers and say, "Hey, prior off. our [174:28] customers don't like it, you don't like [174:30] it, we have all the data, we can just [174:33] automate this for you. I would argue [174:35] this is an area where you are seeing one [174:38] of Epic's favorite business strategies, [174:41] which is we sell the main thing to a [174:44] hospital system. We see what other [174:46] vendors they're using. We figure out if [174:48] they should also buy that from us if we [174:50] have some competitive advantage other [174:52] than just single vendor to be able to do [174:54] it. Sometimes just being a single vendor [174:57] with the same offering is enough. But [174:58] then they try to figure out if they can [175:00] also do that and include it in their [175:02] enterprise agreement. They want to sell [175:04] everything to the hospital and be the [175:07] single IT vendor. You know, the [175:09] hospitals do still buy other IT [175:13] services. So the epic playbook is go [175:16] from selling a hospital one thing to [175:18] figure out how to then sell the hospital [175:20] everything. And then once you've sort of [175:22] exhausted that, then figure out who else [175:23] you can sell stuff to and that's why [175:26] they're looking into payers and that's [175:27] why they're looking into pharma [175:28] companies. It's how can we leverage the [175:31] asset that we have to then now that [175:34] we've sort of saturated this market sell [175:36] to other customers too. Fair point. All [175:38] right. So that basically brings us to [175:41] today. If you're in this ecosystem, [175:43] you're going to say there's 50 things [175:44] that Epic does that you didn't talk [175:46] about. That is true. There's no chance [175:48] that we could talk about all of that on [175:49] this episode, but there is one really [175:52] cool interesting thing that they are [175:54] doing right now that we do want to talk [175:55] about as we catch up to today and then [175:57] we'll give you the stats on the business [175:58] today and we'll talk about the future [176:00] and then we'll do our analysis. They [176:02] launched something a few years ago [176:04] called Cosmos. Yeah, this is pretty [176:06] cool. This is awesome. So they've [176:10] realized, okay, there is data on [176:12] patients that is owned by patients in [176:14] hospitals, not by us. That lives on a [176:17] bunch of different servers. Some are on [176:18] prem at hospitals, some are in our [176:20] cloud, some are on AWS, some are on [176:22] Azure. But either way, there's all this [176:24] interesting structured patient data [176:26] stored in Epicare instances. Can we [176:30] anonymize that and get it all in an epic [176:34] hosted instance of something very [176:38] queryable and very useful for a bunch of [176:41] activities? They did it. It's called [176:44] Cosmos. And Cosmos has [176:48] 295 million patients worth of data in [176:51] it. All anonymized. Yep. And I think [176:54] that across those 295 million patients, [176:58] there's data from 15 billion individual [177:02] patient encounters. Yes. So like [177:04] structured data from a doctor [177:08] encounter 15 billion times with [177:10] patients. It's crazy. So we were asking [177:13] the question earlier, do EMRs or EHRs [177:15] really benefit patients? And there's [177:17] some arguments that aside from how [177:19] awesome it is to just access your [177:20] records at home, there's a bunch of [177:22] stats around yes, it benefits patients, [177:24] it's patient safety, it's patient ease. [177:26] There's studies that say 45% of patients [177:29] reported improved quality of care with [177:31] EHRs while only six noticed decline. [177:34] Epic has a stat in 2023, its system [177:36] prevented 66 million potential adverse [177:39] drug interactions and 250,000 potential [177:42] surgical errors. All of this seems very [177:44] plausible. If you're tracking all this [177:45] stuff digitally, you can do intelligent [177:47] things that improve care. But on top of [177:51] that, more interestingly, what if all of [177:54] the data was actually in one place and [177:57] you could do stuff with it? What is the [177:59] value of having something digital versus [178:01] paper? This is sort of like the utopian [178:03] dream that people have always had and [178:05] that the Bush administration and the [178:07] Obama administration had. Epic is [178:10] finally doing it. Yes. And one very [178:13] clear interesting illustrative example [178:16] is imagine there's something wrong in [178:18] the water supply but all the records are [178:22] on paper. So unless you're going and [178:25] pulling individual files and then [178:27] looking at some specific test result [178:29] across a whole bunch of people which you [178:31] probably need a research grant to do to [178:33] actually go and sit down and pull all [178:34] these individual files you're not going [178:36] to find it. But if you have a whole [178:38] database full of information and you can [178:40] quickly and easily look horizontally [178:42] across a whole bunch of records, this is [178:44] Flint, Michigan. This is how they [178:46] figured out there was a water problem in [178:48] Flint, Michigan. And you would not have [178:50] found it in a paperbased world. Cosmos [178:53] is that on steroids. What can you figure [178:55] out about the whole world if you have [178:58] 295 million patients worth of data in an [179:01] anonymized accessible database and it's [179:04] accessible to any institution for free [179:07] that contributes data in. So of all the [179:09] Epic customers, anyone who opts in gets [179:12] to access and query it for research or [179:15] cool stuff like I have a really unique [179:17] patient who has some crazy condition [179:19] I've never heard of. Has this ever [179:21] happened to anyone else in the world? [179:22] And what was the result? Yep, this is [179:25] what's really cool. They've productized [179:28] this and made it free that any of their [179:31] customers can just turn it on. So, Ben, [179:34] exactly like you're talking about. I'm a [179:35] doctor. I have a patient. There's [179:37] something going on. Cosmos can surface [179:40] to me. Hey, here are other examples of [179:43] similar cases anonymized. They call them [179:45] lookalikes, I think. Yeah, lookalike [179:47] patients throughout the history of [179:50] everybody that's ever been on Epic. And [179:53] here are the doctors who treated those [179:55] patients. So, this is already happening [179:58] quite a bit, I think, with rare diseases [180:01] and crazy cases. I think so, too. It'll [180:04] be useful for clinical trials. It'll be [180:05] useful for research. I do think it's [180:08] free right now, which is pretty [180:09] interesting. We'll have to see how they [180:12] price and package it in the coming [180:13] years, but yeah, you're right. This is [180:15] like the utopian dream of wouldn't it be [180:18] cool if we had all this data in one [180:19] place. Yep. And then that really [180:22] precages something we're going to talk a [180:23] lot about in analysis which is wow [180:26] imagine what you can do with AI with [180:28] this data. Yep. But we will get to that [180:30] in a minute. All right. So the business [180:32] today, let's just make sure we're all on [180:33] the same page on where things stand. [180:36] They have 607 customers with 3,200 [180:40] hospitals in between them. As you know, [180:42] they've never lost a single one, which [180:44] is insane. They add 10 to 25 new health [180:47] systems as customers each year. Their [180:50] customers represent 590,000 physicians [180:53] and [180:54] 495,000 staffed beds. And that's across [180:58] 325 million patients worldwide. Yep. 280 [181:02] million of which are in the US. Yep. So [181:04] basically almost all of America gets [181:07] seen in some part of their healthcare by [181:10] a epic system. Yes, that is correct. In [181:14] revenue last year in 2024 they did 5.7 [181:17] billion which grew 13% year-over-year [181:21] over the last 5 years and 16% last year [181:25] from 4.9 billion to 5.7 billion. David, [181:29] I wanted to bring something up with you. [181:31] They actually do less revenue than I [181:33] would have figured. Yes. And for people [181:35] that are like 5.7 billion is a lot. What [181:37] are you talking about? If you divide it [181:38] by the number of customers, it comes out [181:40] around $10 million a year a customer. If [181:42] you live in the enterprise software [181:44] world, that's not crazy. This is the [181:47] central nervous system of these giant [181:50] health systems. This is what everything [181:52] else relies on. They run their entire [181:54] business on this. It costs a lot of [181:56] money because there's all these people [181:57] administering it. I talked to someone at [181:59] a midsized health system that said [182:02] there's a hundred full-time employees [182:04] administering Epic at her hospital [182:06] system, but Epic actually only captures [182:09] about $10 million a year of the value [182:12] created. So, there's this sort of [182:14] interesting dichotomy of yes, it's a lot [182:17] of spend, but the minority of it [182:20] actually goes to Epic. Yeah, it is this [182:23] element of I know I said altruistic [182:25] earlier when talking about the company. [182:27] You're right. It's not altruistic, but I [182:29] think it's really, really important to [182:31] understand. It's very long-term [182:35] orientation. They could charge their [182:37] customers a lot more money. They [182:40] definitely could. I think I mean, they [182:42] didn't used to be able to. I think they [182:44] used to saturate willingness to pay. Now [182:45] that they're this sort of dominant, they [182:47] probably could. There's so many costs [182:50] that get paid out to the consultants, [182:52] the Accenture, Deote, IBM, Nordic, the [182:55] post-implementation people, the [182:56] optimization people, the headcount [182:58] internally, the lost revenue from doing [183:01] the switch and taking that downtime. I [183:04] mean, this costs health systems a lot. [183:07] Epic actually just doesn't capture that [183:09] much of it. I'm not as convinced that [183:11] they have room to take price as much as [183:14] I'm surprised they don't capture more of [183:17] the overall value of a central nervous [183:20] system for a hospital system. They could [183:23] probably capture some more. And as we [183:24] get further and further in the future [183:26] and they get more and more dominant, [183:27] they certainly will gain more pricing [183:29] power. I think they could. I mean, [183:30] again, to your point of on average $10 [183:33] million a year cost for a piece of [183:36] software that is so absolutely vital to [183:39] your business, I think they could [183:40] capture more. And I think part of the [183:43] reason they don't is they never want to [183:47] give a customer any kind of incentive to [183:51] rebid or consider another solution. It's [183:53] kind of like the Jeff Bezos quote about [183:55] AWS in the early days that he wanted it [183:56] to be like an irrational and [183:58] irresponsible decision not to use AWS. [184:01] Oh. To pick someone else. Yeah. Yeah. I [184:04] think they want it to be an irrational [184:05] and irresponsible decision for a health [184:08] care system not to use Epic. [184:12] Yeah. I think that's right. And just one [184:14] other number to throw out to analyze [184:16] this on a different vector. When I said [184:18] 5.7 billion, topline isn't that big of a [184:20] number in healthare. United Healthcare, [184:24] which owns an insurance company and a [184:25] hospital system, does $400 [184:29] billion a year in revenue. That's wildly [184:34] apples to oranges because they're [184:36] actually administering care in the case [184:37] of their hospital system. So, it's a [184:39] different thing. But that company does [184:42] $35 billion in IBIDA. United Healthc [184:45] Care's profit dollars are six times [184:47] Epic's entire revenues. So Epic has [184:50] become this incredibly important and [184:53] powerful lynch pin in the system despite [184:56] not actually having that big of a profit [184:59] pool in the industry. Yep. Or even maybe [185:02] better put of having that big of a [185:03] revenue pool in the industry. I came [185:05] into this episode expecting to be like, [185:08] "My god, this company just mints money [185:10] and is such a for how much people talk [185:13] about it as a part of the healthare [185:14] system. I just expected it to have more [185:16] cash moving through it." All right, [185:19] other quick stats. 14,000 employees [185:22] today. The market share is only 42% of [185:25] hospitals, but it's the largest and most [185:27] successful hospitals. So, they actually [185:29] have a bigger share of all care [185:30] performed. 58% of ambulatory physicians [185:33] now use Epic. 79% of the US ends up [185:38] using Epic in one way or another. In [185:40] terms of profit, we heard a few [185:42] different estimates. There are Ebida [185:43] margins somewhere between 30 and 35%. So [185:46] just to be conservative, we'll take that [185:48] at the low end and say about $1.7 [185:50] billion in Ebida if those estimates are [185:52] correct. So then that kind of brings us [185:55] to what is Epic worth? Which is sort of [185:57] a silly question because if you own [185:59] shares, it would be very difficult to [186:00] ever sell shares to anyone other than [186:02] the company itself. So what's the point [186:04] of valuing something that you can't own? [186:06] And what's the point of owning something [186:07] that'll never provide a financial [186:08] return? Yeah, there's never going to be [186:10] a liquidity event for your shares, [186:12] right? But it's worth doing. So we [186:16] started this episode talking about [186:18] Judy's one of the most successful, if [186:19] not the most successful female founder [186:21] of all time. Forbes has a list of this [186:23] of the most wealthy self-made women. I [186:26] think their estimate is silly low. Yeah, [186:29] wildly wrong. In 20121, Forbes estimates [186:32] her net worth to be 7.6 billion, [186:35] implying that the company is worth what [186:38] 15 billion. That is a ludicrous [186:40] valuation for Epic. If you do the very [186:43] conservative thing of looking at [186:45] Cerner's EBIDA multiple of 30 times when [186:48] Oracle bought them and you assume [186:50] conservatively again a 30% IBIDA margin [186:52] at Epic that would give you a number of [186:56] 51 billion valuing Epic at 51 billion. [187:00] But come on, Epic's revenue streams are [187:02] way more durable than Cerner's. And Epic [187:05] is growing while Cerner was going [187:07] through a hard time. You could get to a [187:09] similar number of that 50-ish billion if [187:11] you just slap a 9x revenue multiple, [187:13] which is a reasonable public market [187:15] software comp, but this company has [187:18] insane durability. Yeah, it's missing [187:21] the point that I think this is maybe the [187:23] most durable software company in [187:25] history. Yes. My guess is if this is [187:29] public, which it never will be, [187:31] investors would value it somewhere in [187:33] the hundred billion neighborhood, giving [187:35] Judy's shares a value of about $50 [187:37] billion and making her the wealthiest [187:40] self-made woman in the world and one of [187:42] the most successful entrepreneurs, [187:44] period. And that would put Epic in one [187:48] of the most valuable 150 or so companies [187:51] in the world, right along with Shopify, [187:53] ARM, Loheed Martin, and Starbucks. Yep. [187:57] Another thought exercise to kind of come [188:00] at this question is if in some alternate [188:05] universe, Microsoft or Google or Amazon [188:09] or I don't know, maybe even Berkshire [188:10] Hathaway, maybe Apple, if they could [188:13] acquire this company, how much would [188:16] they be willing to pay for it? I think a [188:18] hundred billion is on the very, very low [188:21] end of that spectrum. That's 4% of Apple [188:23] or Microsoft. Of course, they'd be [188:25] willing to pay 4% of themselves for [188:27] this. There's the market, there's the [188:29] durability of the revenue and the [188:31] profits. Oh, I see. You're arguing that [188:34] on a financial basis alone. Yes. And [188:36] also the strategic value on top of that [188:39] and the strategic value both in terms of [188:44] becoming the most important partner to [188:46] this huge industry to the biggest [188:48] players in this huge industry in [188:50] healthcare, but then also just the data [188:52] and the AI world that we're in today. I [188:54] mean, my god, Cosmos, you know, 15 [188:56] billion patient interactions. You think [188:59] Google or Microsoft or whomever wouldn't [189:01] pay a lot of money to own that? Like, of [189:03] course they would. Part of me is like, [189:05] "Oo, thank god Judy's putting it in a [189:07] trust." All that to say, this is never [189:09] going to happen. So, final chapter of [189:11] the story here is Judy's 81 years old. [189:15] What is going to happen at some point in [189:17] the future? Even if she's the next [189:19] Warren Buffett here, we're still talking [189:21] about in the next couple decades. You [189:23] know, there will have to be a [189:24] transition. Like you said, Ben, she owns [189:26] about half the company economically and [189:28] 100% of the voting shares. She has been [189:31] transferring her non- voting shares into [189:34] her foundation called Broots and Wings. [189:36] So, she signed the giving pledge and is [189:38] transferring her wealth into the [189:40] foundation. They then sell those shares [189:43] back to the company at a price that [189:46] changes every year. and that funds the [189:48] foundation. Judy has announced that upon [189:51] her death, all of her voting shares, so [189:54] the 100% voting control that she has of [189:56] the company, will transfer into a quote [189:58] unquote purpose trust that will be [190:01] administered by three [190:04] constituencies. One, her family, her [190:07] husband if he's still alive, Gordon, and [190:09] their three children. Two, a set of five [190:13] longtime senior managers from Epic. Mhm. [190:17] And three, I believe three customer [190:20] CEOs, CEOs of big customers. And that [190:24] that group will manage the trust. And in [190:27] the trust [190:29] bylaws are several things. One, the [190:33] company can never be sold or taken [190:35] public. Ironclad, that can never happen. [190:38] Two, the next CEO of Epic must meet two [190:42] criteria. Oh. Be a software developer, [190:44] right? Yes. one must be a longtime Epic [190:47] employee and two must be a software [190:50] developer. Ah, I love the idea of the [190:54] CEO being someone who came from the core [190:57] skill of the company. It's very Costco. [191:00] You know, you think about all the Costco [191:01] CEOs have been people who were core to [191:04] Costco's operations over the years and [191:06] had worked there for like 30 plus years. [191:08] Nike's heyday was when a guy who started [191:10] in sneaker design ended up running the [191:12] company, Mark Parker. There's sort of a [191:15] beauty to when that happens. There is a [191:17] playbook to this. Yes. One last thing to [191:20] know about Epic today looking forward. [191:22] We talked a lot about interoperability [191:24] and their notorious lack of partnership [191:28] or ability to access their data or [191:30] integrate with them in the past. A lot [191:32] has changed, but there would be a whole [191:34] another 4hour podcast on specifically [191:37] how there are many programs each with [191:39] different methods of integration, [191:40] different names, different standards. [191:42] They've all evolved over the years. They [191:44] come with things like revenue shares [191:46] back to Epic. And if it is the rare true [191:48] partnership, they can also come with [191:50] warrants to own a chunk of your company. [191:51] But the bottom line is you actually can [191:53] do a lot more building on top of and [191:55] integrating with Epic than you could [191:57] have 10 or 20 years ago. All right, [191:59] that's our epic story. Should we move [192:01] into analysis? Let's do it. So the first [192:05] part of analysis is a section called [192:07] power which we have shamelessly ripped [192:09] from Hamilton Helmer's book seven [192:12] powers. The question is what is it about [192:15] any given business that allows them to [192:17] achieve persistent differential returns [192:19] or to be more profitable than their [192:22] closest competitor on a sustainable [192:24] basis. And there's seven of these. [192:26] There's counterpositioning, scale [192:27] economies, switching costs, network [192:29] economies, process power, branding, and [192:31] cornered resource. And as usual, we have [192:34] to separate the takeoff phase from the [192:37] phase that they're in today. David, I'm [192:39] curious. Do you have one for the takeoff [192:42] phase? Do you feel like you can analyze [192:44] why in the takeoff phase they gained [192:46] power? Well, it's funny. They didn't [192:49] take off for quite a while. Right. And I [192:52] might argue they didn't actually have [192:54] any power during the takeoff phase. [192:56] Yeah, I don't think so. I mean [192:58] ultimately their biggest point of [193:00] differentiation being the single [193:04] platform didn't matter that much in that [193:06] phase because everybody else was [193:09] starting up new and they hadn't built [193:10] that much of it out and there wasn't [193:12] that much computing adoption among [193:14] hospitals generally so like didn't [193:16] really matter. I think what was [193:18] happening is they had this kind of nent [193:20] market in the proto EHR world where they [193:24] were able to sell something small enough [193:26] at enough of a profit to be [193:27] self-sustaining while they beded their [193:29] time and built out the bigger suite [193:32] which they were able to basically use [193:35] time as a resource that most people [193:38] don't flex. Most entrepreneurs want to [193:40] do something in a short period of time, [193:42] usually because their capital structure [193:43] demands it. And so you launch a platform [193:46] and try to get other people to build the [193:47] applications because you're in a really [193:49] really dynamic fastchanging market. Or [193:51] you launch an application on a platform [193:53] that somebody else built similarly [193:55] because you're in a dynamic changing [193:56] market. And Epic's market was developing [193:59] slowly over a long period of time. And [194:01] they were willing to run a small [194:02] business for a long period of time and [194:05] then sort of grow into building all the [194:08] functionality for the eventual huge [194:10] market that wanted all the functionality [194:12] tied together. Yep. Okay. So then what [194:14] powers do they have today as the scale [194:16] incumbent? Yeah. A lot. A lot is the [194:20] answer. A lot. A lot. I mean the most [194:23] obvious one is switching cost. I mean, [194:25] this is the biggest switching cost piece [194:27] of software ever. Ooh, that's [194:29] interesting. Is that true? Do you think [194:31] it's bigger than people's ERP instance? [194:33] For sure. This is like an ERP on [194:35] steroids. This is the most important [194:37] nervous system for your entire business. [194:40] Oh, yeah. And by the way, if things go [194:41] wrong or it goes down, people die. [194:43] That's a very good point. So, people [194:46] spend, you know, including all the [194:48] opportunity costs and everything, [194:49] hundreds of millions or billions to put [194:51] this in. Are there any ERP or CRM [194:54] implementations that cost a billion,5 [194:57] billion, 10 billion dollars? [195:00] Maybe, but I don't think any that are so [195:05] core to everything about your customer's [195:08] business. You're right. Even if it was [195:09] really expensive to put it in, you're [195:11] still more likely to swap out one of [195:12] those systems than you are your EMR at [195:15] this point. Right. It's so much more [195:17] than a medical record. Yes. Yes. So [195:20] that's definitely true. Scale economies [195:22] is definitely true to the thing we were [195:24] talking about a minute ago. You can't [195:26] make the investments in product breadth [195:30] without having the scale that you have [195:31] today. Yep. You also can't do their [195:34] clever bundling in a bunch of free stuff [195:38] that you might want to use in the future [195:40] unless you have the scale economies. [195:42] Also, I mean, they just develop a lot of [195:45] software that they either throw in for [195:48] free or offer to you in the future. And [195:51] the only way you can develop that much [195:53] software is have a huge number of [195:55] customers to amortize it across. Yep. If [195:58] you're trying to start a new EMR today, [196:00] you have to write a lot of software to [196:02] be competitive. Yep. And that's scale [196:04] economies. Yep. Network economies. Yep. [196:07] You bet. You betcha. They hit this [196:10] interesting tipping point where they [196:12] could win just on we're the best most [196:17] reliable platform for a long time and [196:20] then at some point it hit some tipping [196:22] point where there was a second value [196:23] proposition which is most other [196:25] hospitals are also on Epic so your [196:28] patients and your doctors will all [196:30] appreciate it if you adopt Epic to make [196:32] it easier to interop with the records at [196:34] all those other hospitals. It's like a [196:35] icing on the cake. Oh, you're talking [196:37] about interop. I mean there's interrupt [196:39] but there's also do you think you will [196:41] need to hire more doctors in the future [196:43] either to grow or as your existing [196:46] doctors retire right pick the standard. [196:49] Yes. Well, what do they know how to use [196:52] and what are they trained on? Epic. [196:54] Great. Yeah, that's true. It's a second [196:56] network economy in addition to the care [196:59] everywhere concept of being able to [197:02] share my records across institutions. [197:04] There's one institution in Seattle that [197:05] I go to for my son's food allergies. [197:08] They're on Cerner. I have three [197:10] institutions that I've gone to over the [197:12] years for various other stuff that are [197:14] all on Epic. I don't know why that one [197:17] hospital is still on Cerner. Yeah, it's [197:19] really annoying to you, isn't it? I'm [197:20] sure everyone else in Seattle is like, [197:23] why don't you just do the thing that all [197:24] the other big hospitals do so that it's [197:26] all the same platform? That's network [197:28] economies. Yeah, it's kind of amazing [197:29] that they have built a network economy [197:31] business. Like who would have thought [197:33] medical software? There is value to [197:36] every participant in the system from the [197:38] other hospitals to me as a patient to [197:39] doctors. If that one hold out hospital [197:42] that for some reason is on Cerner just [197:43] switches to Epic. It's the definition of [197:45] network economy. I think they also have [197:48] branding. They're the most trusted by [197:50] all the most prestigious institutions. [197:52] So at this point, if they had no other [197:55] powers and they just came in at [197:56] identical bids on the table, 100% of the [197:58] time people say Epic cuz it's going to [198:00] work. Like I know your brand. Don't get [198:02] fired for buying IBM. Yep. Exactly. Yep. [198:05] I didn't think about that. I was like, [198:06] "Oh, no. They don't have branding. It's [198:07] not really a consumer brand." You know, [198:09] my charter. But no, no, you're totally [198:10] right. They totally have branding. That [198:12] is in part why they built that massive [198:14] auditorium to bring the whole universe [198:16] of Epic together every few months. Yep. [198:20] And then I think you could argue that [198:21] they do have process power, I think, is [198:23] the last one I would say in their [198:26] development process and the language [198:29] that they use. and it being kind of [198:31] unique to Epic. This is more of a [198:34] defensive one. I think it's about the [198:36] training of young employees, the way [198:37] that they sort of make them the Epic [198:39] way. Well, interesting. There's the Epic [198:42] way. I buy that generally, but I think [198:47] specifically you can point to their [198:49] software developers who they are hiring [198:51] out of college mostly building on cache [198:54] and mumps and they don't don't develop [198:57] transferable skills. You know, it's not [198:59] like you can easily then go work at [199:01] Cerner, right, and port that over. I [199:04] mean, sure, you could like if you're a [199:05] great software developer, you can learn [199:06] other languages and other frameworks, [199:08] but if you spent your whole career all [199:10] developing in this sort of proprietary [199:12] epic system, and I should say [199:14] proprietary is the wrong word. They [199:15] don't make mumps or cach a separate [199:17] company called InterSystems makes them [199:18] and they license it from them. But most [199:20] of the rest of the industry does not [199:22] operate on that. It is a bit of a [199:24] process power. Yep. I think that's fair. [199:26] All right, that does it for power. [199:28] Great. Do you have any playbook themes? [199:31] Yes, I do. The one that jumped right out [199:35] at me was the heavy spend on research [199:38] and development. And this one is from [199:39] friend of the show Arvin Navaratnam at [199:41] Worldly Partners in the always excellent [199:44] writeup that he does associated with [199:45] these episodes that we'll link to in the [199:47] show notes. Epic of course builds, as [199:50] we've talked about a million times, [199:51] their whole systems from scratch and [199:53] they don't buy companies. It's all [199:55] in-house development. The result of that [199:57] is 35% of their operating expenses are [200:01] spent on R&D. If you compare this to [200:03] Athena Health, a competitor, that's 10% [200:06] or Oracle all up is 23%. Or if you look [200:10] outside this industry, it's actually [200:12] equivalent to Apple at 36% of their [200:16] operating expenses of R&D. Amazon's 28%. [200:19] So 35% is quite meaningful. It's quite [200:22] high. Yeah. The only one that beats them [200:24] is Google at 45%. And think about all [200:27] the money that Google is just dumping [200:28] into R&D for all these sort of [200:30] future-looking projects and you know [200:32] they're a core technology company. So on [200:35] the one hand, duh, because Epic doesn't [200:37] really buy other companies or spend any [200:39] money on sales and marketing and they [200:41] keep their GNA low. So what are their [200:43] other expenses? Of course, it's going to [200:44] be high on R&D. Yes. Unlike most [200:46] companies, they spend zero on sales and [200:48] marketing, right? if they classify those [200:50] eight people that they employ as [200:52] quotequote salespeople as you know [200:53] headcount maybe they spend I don't know [200:55] not much yes small but I think to get to [200:59] this point that we've been sort of [201:00] hitting in a bunch of different ways as [201:03] long as you can win deals in the short [201:05] term early days without sales and [201:07] marketing spend and you can just spend [201:09] all of your costs or as much as possible [201:11] on R&D that R&D does compound over time [201:15] to give you a big competitive advantage [201:18] later in life that other forms of spend [201:21] do not. Yep. Especially in software. [201:24] Yes. My next one is about growth. So, I [201:28] was reading an article that said Faulner [201:29] once described her approach as climbing [201:31] a mountain, not trying to see the entire [201:33] mountain at once, but by focusing on the [201:35] next hill in front of her, and it [201:37] reminded me of something I've read [201:39] before. Does this remind you of [201:40] anything? [201:42] Uh, I've drawn a blank. Go for it. Okay. [201:45] So, in Paul Graham's legendary 2012 [201:48] essay, Startups Equal Growth, he's [201:51] talking about how the whole thing you [201:52] need to do as a startup is focus on next [201:55] week's growth. He then says, "In theory, [201:57] this sort of hill climbing could get a [201:59] startup into trouble. They could end up [202:01] at a local maximum. But in practice, [202:03] that never happens. Having to hit a [202:04] growth number every week forces a [202:06] founder to act, and acting versus not [202:08] acting is the high bit of succeeding. [202:11] Nine times out of ten sitting around [202:13] strategizing is just a form of [202:14] procrastination. Whereas founders [202:16] intuition about which hill to climb are [202:18] usually better than they realize. Plus [202:20] the maxima in the space of startup ideas [202:23] are not spiky and isolated. Most good [202:25] ideas are adjacent to even better ones. [202:28] I just thought I don't know it just [202:30] jumped off the page at me that Judy who [202:33] has almost certainly never read any of [202:34] Paul Graham's writings had this same [202:37] insight. I don't try to see the entire [202:39] mountain at once. I just try to focus on [202:41] the hill right in front of me. And that [202:43] hill is adjacent to a bigger and more [202:45] interesting hill. Oh, I'm so glad you [202:48] brought this up. I was thinking about [202:50] Judy as this incredible software [202:52] developer who became a founder of a [202:55] company and this incredible business [202:57] person and didn't follow any of the, you [202:59] know, MBA rules except that she did. She [203:01] just discovered them all for herself in [203:03] Wisconsin. Well, right. And I was [203:06] thinking about it. I was like, man, her [203:08] story in this company kind of reminds me [203:09] of Paul Graham and Y Combinator. And [203:11] there's so many resonances here, but [203:14] they're also kind of like a different [203:15] world. If Y Combinator weren't about [203:18] funding companies and being a venture [203:20] capital firm and instead was espousing [203:24] to founders instead of raising money, [203:26] build your companies independently [203:28] without capital, we would have a lot [203:30] more epic systems running around out [203:31] there. Wow. But, you know, she also [203:33] built this in a different era where you [203:35] could do this without capital. Yeah, I [203:37] think that's right. But the distance [203:39] between Judy Faulner and Paul Graham is [203:42] actually not nearly as far as you would [203:43] think. Yep. It's applying computer [203:46] science principles and common sense ways [203:50] of running a company that are like [203:52] obvious to you but contrarian to the way [203:54] that most people seem to run companies [203:57] and just being a programmer turned [203:59] business person to its logical extreme. [204:01] Yep. Okay. One that we did not talk [204:04] about is their litigation strategy. It's [204:07] worth just touching this so people are [204:09] aware. They are aggressive on defending [204:12] IP and as they would put it on defense [204:15] against patent trolls. They're also [204:18] extremely aggressive on employment [204:20] practices law. They have non-competes to [204:23] keep secrets secret and they enforce [204:25] them. They are also why employers can [204:28] force their employees into arbitration. [204:29] They won a landmark case around forced [204:32] arbitration. Yeah. Went to the Supreme [204:34] Court, I believe. Yep. They're embroiled [204:36] in lawsuits with other health IT [204:38] companies trying to use Epic software in [204:40] a way that Epic believes violates terms [204:42] of service. There's not one of these. [204:44] There's a bunch of these, including an [204:46] antitrust lawsuit. Yes. A big one with a [204:49] startup called Particle Health that is [204:51] at a really pivotal point right now. [204:53] Yes, David. They are being accused of [204:55] violating the Sherman Antitrust Act. [204:57] Yep. And that suit is ongoing. Yep. We [205:01] just put this out there to say this [205:03] company is seemingly not shy about using [205:06] the law to protect things that they [205:08] believe are theirs and to deepen their [205:10] moat. Interestingly though, they have [205:13] never been in a lawsuit or even in [205:15] mediation with a customer. There's a [205:17] sacredness to that customer relationship [205:19] that they hold really dear. It comes [205:20] back to when we were talking about the [205:21] company culture. Everything at the [205:23] company is about customers. Yep. All [205:26] right. The next one is one I call [205:28] enterprise sales on steroids. So [205:32] normally in enterprise sales you have a [205:35] bunch of things. Long sales cycles, [205:37] stuff where the buyer is not the user. [205:39] So you end up with cluji UI and it's not [205:42] just because people don't care, but the [205:44] breadth of enterprise software is really [205:46] really wide. So it takes forever to [205:48] write everything. And so you're not [205:50] going to keep updating it every time [205:51] there's new UI trends. And on top of [205:54] that, a million screens look really [205:56] similar for different users, for [205:58] different use cases. So they all just [206:00] kind of end up looking like drop downs [206:01] and buttons. Hard to blame someone [206:04] specifically for that. There's tons of [206:06] customization, long implementation [206:08] times. This is enterprise sales and [206:10] enterprise software generally, not in [206:12] healthcare. And then in healthcare, here [206:14] comes the steroids. you have the added [206:16] layer of regulation from HIPPA and other [206:19] medical system compliance stuff and data [206:22] sensitivity and lives are on the line. [206:25] And so in most industries when you're a [206:29] startup looking down the end of the [206:30] barrel and you're like all right I am [206:32] not doing this top down enterprise sales [206:34] thing I'm going to do productled growth [206:35] going to get some bottoms up adoption [206:37] we'll play the Slack or the Figma or you [206:39] name it but HIPPA makes productled [206:42] growth much harder. So from the great [206:44] Substack health API guy, under HIPPO [206:47] regulations, a business associate [206:48] agreement typically needs to be signed [206:50] by someone with the proper authority to [206:52] legally bind the organization, not [206:54] individual providers or employees. So [206:56] David, this is something I kept [206:57] thinking, why doesn't one doctor or one [207:00] practice adopt something and then get [207:02] bottoms up adoption and that's how we'll [207:04] get the next EHR in the future? No, [207:07] these are all CIO sales because they all [207:10] need to be, right? You can't have the [207:12] typical Oh, a small team within a big [207:15] company is going to buy this on a credit [207:16] card. Nope, not going to happen. And [207:18] there's some exceptions. I know there's [207:19] some PLG that happens in healthcare, but [207:21] it's way, way harder to figure out how [207:23] to do that than normal B2B SAS. Yep. And [207:27] my last one, Judy is a wacky, wacky [207:32] founder. And so is every truly great [207:36] entrepreneur that we study. She is truly [207:38] of one. I mean, she has stated that it [207:41] is more likely that she'll die than [207:43] retire, just like Warren Buffett always [207:45] jokes. She is singularly focused on [207:47] making Epic great in her life, just like [207:51] Ingvar Comprod and IKEA. She's totally [207:54] obsessed with rapidly incorporating [207:55] feedback from customers, just like Jeff [207:57] Bezos. And she believes correctly that [208:00] the nerds will prevail in the long run, [208:02] at least correctly for her market, her [208:04] point of view, her industry. She is [208:06] completely unique in the way that all [208:08] the great founders we study are. Yep. [208:11] You foreshadowed in the intro that Judy [208:13] is almost certainly the most successful [208:15] by any measure female founder in [208:17] history. I think that is actually true. [208:19] I looked up to try and see who else [208:22] might even come close. You could maybe [208:25] make a case for Estee Lauder. However, [208:28] despite Estee Lauder reaching about a [208:30] hundred billion dollar market cap during [208:32] the pandemic, it's only worth about 20 [208:34] today. I don't know what happened there. [208:35] Taylor Swift and Oprah, which we've [208:37] talked about, and we did our Taylor [208:39] Swift update at the Chase Center show [208:40] last year. We think the enterprise of [208:43] Taylor Inc. is only worth about 11 [208:44] billion. I mean, only quote unquote. I [208:46] love that you and I have a have like an [208:49] ongoing Taylor Inc. calculator. Yeah. [208:51] Yeah. Yeah. Oprah's less than that. [208:54] Forbes currently thinks that Diane Hris [208:56] is the wealthiest female founder in the [208:59] world, or at least in America. She's the [209:01] founder of ABC Supply, which is one of [209:03] the largest roofing suppliers in the [209:05] country. You know what's crazy? This [209:06] means that the two most successful [209:08] female founders in America are both [209:10] based in Wisconsin. I know. Isn't that [209:12] amazing? I was going to say that. I love [209:14] that you found that, too. It is a big [209:16] and great company for sure. However, ABC [209:19] Supply does about $20 billion a year in [209:23] revenue. I assume not at anywhere near [209:26] Epic's margins. Yep. So, I don't think [209:28] it is worth the call it I think minimum [209:31] hundred billion that I think Epic is [209:32] worth. You're now up to a minimum [209:34] hundred billion. [209:36] Well, I'm going by my rubric of if [209:38] somebody could acquire this company, [209:40] what would they be willing to pay? Yes. [209:42] That's kind of funny. If someone offered [209:44] you shares in Epic right now at 100 [209:46] billion, would you pay it? Oh, good [209:49] question. If I could get profit [209:52] distributions from Epic, yes, I would [209:54] love to hold Epic shares for the rest of [209:56] my life. I would for sure pay that. [209:57] Yeah, I totally agree. If I were banking [209:59] on a liquidity event, then no. [210:03] Yeah, but me personally, yeah, I would [210:05] love to hold shares. Yeah, when people [210:07] are speculating on the value of [210:09] something, the interesting way to turn [210:10] it around is always, okay, are you buyer [210:11] at that price? Like, open your wallet. [210:13] Makes it feel much more real. I would [210:15] for sure buy shares of Epic at $100 [210:16] billion. There's no way I will ever be [210:19] allowed to, but I would love to. That [210:21] sounds like a standing offer, and the [210:22] show is not investment advice. [210:26] in this category. Man, it is crazy how [210:30] powerful founder continuity can be. I [210:32] was thinking about this. How many other [210:34] companies have had a single founder [210:36] leader for 47 years? I mean, Jensen's 31 [210:40] years into Nvidia. Zuckerberg is 21 [210:44] years into Facebook. Obviously, there's [210:47] a lot running smaller companies. There's [210:50] Bergkshire Hathaway, right? That's [210:52] actually the exception. Buffett's been [210:54] running Bergkshire for over 50 years, [210:56] but in Judy's case, having 47 years to [210:59] imbue the founders's personality into [211:02] the organization as it's built all the [211:04] way to this scale is really rare and [211:07] really powerful. Yep. And as you've [211:10] alluded to, I mean, she really is like [211:12] Jensen. She's still running the company. [211:15] She's still highly engaged. Yep. [211:17] Absolutely. All right. Baron bullcases. [211:20] Yep. Let's do it. All right. A few [211:22] different things contributing to the [211:24] bear case if I were to make one. US [211:26] customers had very big dollars they [211:28] could pay. The market is big because one [211:31] the US economy and population are large [211:33] and two unfortunately healthcare is a [211:35] large percent of that. So if the future [211:38] is coming from international that could [211:39] be less fruitful much lower willingness [211:41] to pay in most countries for healthcare [211:44] administration stuff. So that's one sort [211:47] of barecase element. The other would be [211:50] if this particle suit ends up getting [211:53] actually picked up and there's a [211:55] material event that happens for Epic [211:58] around antitrust and a violation of the [212:00] Sherman Act. That's a huge huge problem [212:04] and that is a company changing thing if [212:07] that comes to bear. So we'll have to [212:09] watch the news on that. The third is [212:11] there is legislation that happened in [212:13] the last few years about information [212:15] blocking. There's an act that passed [212:17] called the cures act that essentially [212:20] says that an EHR or health system cannot [212:23] block access to information. So even [212:26] though Epic makes the very reasonable [212:27] argument around data privacy and [212:29] security and in the long run patient [212:31] safety by locking down data, the Cures [212:34] Act makes things like screen scraping or [212:37] Chrome extensions or RPA legal as a [212:40] means to extract data from Epic and [212:42] there's nothing they can do about it. [212:43] That is required allowed. Interesting. [212:46] So, at this point, I don't know how that [212:49] could really displace them. What also [212:51] sort of begs the question of, yeah, what [212:52] would you do? Okay, you scrape Epic [212:55] data, you're not going to build Epic, [212:57] right? Epic has always been hyper [213:00] sensitive about locking that stuff down [213:02] for good reasons for their customers, [213:04] but also for their own durability [213:06] reasons. And I think this is going to be [213:09] a tiny crack in the armor. The fear is [213:11] that someone deploys to like 70% of your [213:14] customer base. They're all using some [213:15] Chrome extension. They're all feeding [213:17] the data into some nice UI. That nice UI [213:21] is the thing everyone prefers to use and [213:23] then that UI vendor is like, "Oh, [213:24] instead of using Epic on the backside, [213:26] you could use my own home world thing on [213:28] the back side." But the probability of [213:30] that happen is near zero. Yeah, not [213:32] going to happen. So, I'm not that [213:34] worried about that. The other thing is [213:37] maybe if interoperability comes to pass [213:40] in a bigger way, then you could kind of [213:44] see the best of breed applications [213:46] becoming a more dominant paradigm versus [213:48] the all integrated paradigm. Again, this [213:52] feels pretty handwavy to describe as the [213:54] bare case. Yep. Well, and should that [213:57] happen, they still have the Microsoft [213:59] playbook, which they absolutely run. [214:01] Okay, let's even say there are better [214:03] best of breed point applications out [214:05] there. Great, Mrs. or Mr. hospital CIO. [214:10] You could go pay money for those or you [214:12] could just keep buying your Epic [214:14] Enterprise license where you get access [214:16] to all of these point applications that [214:19] you need for one price. Yep. And then [214:22] the last one is paradigm shift. And this [214:25] is always the bare case for any dominant [214:27] company. Is AI going to be such a [214:29] dominant paradigm shift that it changes [214:34] the needs of health systems? Or is [214:37] valuebased care going to be such a [214:39] paradigm shift that it changes what a [214:42] health system is? Maybe you don't go to [214:44] the doctor most of the time. I have a [214:46] hard time even imagining what it could [214:48] be that you wouldn't need a system like [214:51] Epic. Maybe new care delivery models [214:54] will make the old type of EMRs obsolete. [214:57] This requires some imagination and again [214:59] it's very handwavy. Epic also has their [215:01] eye on all of this. So it's hard to [215:03] imagine any of those things being the [215:05] Yeah. Right. My question to you is going [215:06] to be okay. How much do you want to talk [215:08] about AI here or how much do you want to [215:09] talk about it in the bullcase? Yeah, [215:12] it's probably more a bullcase. All [215:13] right. Well, let's do bullcase. Great. [215:16] All right. The bullcase is they've [215:19] successfully expanded from EMR plus [215:21] billing to all these other specialties [215:23] and modules. They basically served all [215:27] those same customers even more products. [215:29] Those customers A will need more [215:31] products in the future. B there [215:33] continues to be more health systems they [215:35] can sell to. And now C they are [215:38] expanding the customer base from just [215:41] health systems to other types of [215:44] companies, pharma payers, researchers [215:48] from this new dominant position that [215:49] they're in. So they're sort of expanding [215:51] to a whole new set of potential [215:53] customers who could pay them because of [215:55] their dominant position. I think that's [215:57] extremely credible. Yep. They could have [215:59] an even larger business. I mean, prior [216:01] off seems like the first and most [216:03] obvious business opportunity for them [216:05] here. In some ways, hospitals are kind [216:07] of crappy customers cuz they don't [216:09] generate much profit. Insurance [216:11] companies are probably much better [216:12] customers if you can figure out [216:14] something really compelling for them. [216:15] Yep. Unfortunately, again, I feel like [216:17] every time I come up with some profit [216:19] pool in healthcare, I'm like, "Boy, I [216:21] really wish that wasn't there." I have [216:23] to sort of think philosophically on why [216:25] I keep feeling that way. But I'm [216:26] definitely I end up critical of any [216:29] thing that becomes too profitable in the [216:31] healthare system. Well, behind that [216:33] though is that you as a consumer, we as [216:36] consumers feel like some of these [216:38] entities are not in any way providing [216:40] value to us. Right. I feel like I'm [216:42] getting ripped off. Yeah. I mean, I saw [216:44] a study that said 30% of spending in the [216:48] healthcare, you know, in the 18% of GDP [216:51] goes to waste. Truly waste. And I saw [216:53] another one that quantified it and said [216:55] $800 billion of waste is in the system. [216:59] The entire GDP of Switzerland's worth of [217:02] waste exists in our health care system. [217:06] Yep, sounds about right. Another [217:08] interesting thing that someone brought [217:09] up is they asked me how much I pay in [217:11] premiums a year for health insurance for [217:13] the family. And I was like, I don't [217:14] know, I think somewhere in the [217:15] neighborhood of 25 [217:17] $30,000. And their comment back was [217:20] like, in what world are you ever going [217:23] to use 25 or 30,000? Don't like imagine [217:26] what's on the bill cuz that's all made [217:27] up. All the numbers are made up. The top [217:29] line, the negotiated rate, what I [217:31] covered for, they're all made up. Just [217:33] imagine like dollars out of your pocket [217:35] for healthare. And don't even just pick [217:37] this year. Pick a 10-year period because [217:40] stuff might come up. You'll have [217:41] surgeries in some years but not others. [217:43] What are you actually willing to pay and [217:46] go out of your pocket dollars give a [217:48] doctor for everything you need over the [217:49] next 10 years? And compare that against [217:51] everything you're paying into the system [217:54] if you were to be paying your own [217:56] employer side, employee side, and all [217:59] the out-of- pocket stuff. We're not [218:01] getting a good deal. We're just not. And [218:03] we can all feel it. Yep. All right. [218:05] Continuing the bullcase. So, there's a [218:08] whole category called ambient listening. [218:10] And one of the biggest complaints that [218:12] people have with EHRs is doctors spend [218:14] too much time in front of the computers. [218:16] They're typing in notes. They're clicky [218:18] clicky on nine different drop- down [218:20] boxes and alerts. And sometimes that [218:23] gets fixed by having someone in the room [218:25] typing while the doctor is doctoring [218:28] with you. But either way, there's the [218:30] scribe, but the doctor's like also at a [218:31] computer cuz they just need to read the [218:34] medical records. So, the doctor ends up [218:35] in front of a screen. They're wasting a [218:36] bunch of time typing or they're hiring [218:38] someone to type. Oh my god, AI would be [218:41] amazing for this if only there was an [218:44] ambient listening AI scribe that could [218:47] be writing down and categorizing and [218:49] structuring all the data that comes out [218:51] of an interaction with a physician. [218:54] Well, good news. It's happening. It is [218:57] happening and it works. So, David, [218:59] there's a few companies that are [219:01] partners of Epic, right? Yep. That's [219:03] right. So one is Microsoft and Nuance. [219:06] Microsoft bought a company called Nuance [219:08] a few years ago. They used to do like [219:09] Dragon Dictation. Yeah. So they've [219:12] always been in kind of the voice space [219:14] for healthcare. So they now have an [219:16] ambient AI product and then there are [219:19] two big startups. one called a bridge [219:21] based out of Pittsburgh and then another [219:23] one called Suki that have really good [219:26] ambient AI products that plug into Epic [219:30] and yeah are an AI scribe and people who [219:34] use them, physicians who use them love [219:36] them. You can just focus on the patient [219:38] and it scribes it all. Yep. This would [219:40] be a great solve to some of the doctor [219:43] burnout and fatigue and too much screen [219:45] time that a lot of physicians are [219:47] seeing. Yep. That's the bullcase for [219:49] this. These products are all partners of [219:52] Epic and in some cases Epic may have [219:54] relationships or warrants with some of [219:56] these companies or just revenue. You [219:58] could also imagine Epic might want to [220:00] build their own product of this someday. [220:02] It's a big opportunity. Epic is kind of [220:04] the choke point of the industry at this [220:06] point to decide what software [220:08] innovations reach hospitals. I don't [220:11] think it's overstating it to say if [220:13] you're developing new breakthrough [220:16] software to be used by physicians in [220:18] hospital systems, Epic's kind of the one [220:21] that gets to decide are you going to [220:23] reach that customer or not. Yep. [220:25] Exactly. So, I think the basic bull case [220:28] on ambient voice, ambient AI is this is [220:32] a great new product revenue opportunity. [220:36] The mega bull case to borrow from our [220:38] IPL episode here is one CIO was talking [220:41] to me. He's like, "Look, I don't have [220:43] high confidence that this is going to [220:45] happen, but as ambient AI becomes better [220:48] and better, is there a version of the [220:50] future where the EHR quote unquote [220:52] itself just kind of fades into the [220:54] background and this all just becomes an [220:58] AI operating system?" Like, what is it [221:00] primarily doing? the EHR, it's capturing [221:04] data about what happened. It's recording [221:06] it both for clinical uses and then [221:09] importantly for billing purposes and [221:11] then what needs to happen with that for [221:13] billing purposes. Well, that data needs [221:15] to get shipped over to the payer whether [221:17] that's the government or an insurance [221:18] company and then needs to get judged and [221:21] adjudicated and then payouts need to [221:23] happen. Like is there any real reason [221:25] why this can't all be done by AI in the [221:28] future? Why do we actually really need [221:31] userf facing software here or a lot of [221:34] it? So yeah, there may be some future [221:36] where and this is you know a utopian [221:38] type future but a lot of the [221:41] administrative bloat and costs in the [221:43] system actually gets taken care of by [221:46] ambient AI sort of just sucking it all [221:49] up here. I'll believe it when I see it. [221:51] But it sounds nice. Exactly. But a real [221:54] CIO of a real big hospital system made [221:56] this case to me. It's just so hard to [221:58] believe bloat ever goes away. Getting [222:01] administrative costs down is so hard. [222:03] What are you going to do? Put a whole [222:05] bunch of people who have good [222:06] administrative jobs at hospitals out. [222:09] That is going to be hard for our whole [222:10] society. For sure. I mean, this is the [222:12] question about AI, right? But you're [222:14] right. I think it's a bullcase if you're [222:15] a shareholder of Epic, though, right? If [222:17] Epic becomes that system. Yep. All that [222:20] to say, like every industry, AI has big [222:26] potential here. Yep. I think that's [222:27] fair. And that's on top of everything [222:28] we've already talked about. And no one [222:30] is switching off of them. Revenue from [222:31] their install base will just continue to [222:33] grow. They've become the standard that [222:35] everyone will switch to. Once they get a [222:37] customer, they don't leave. Honestly, at [222:40] this point, regulation is probably also [222:42] a bullc case. This is sort of what [222:44] happens once you become an incumbent. [222:45] And regulation tends to shut out new [222:48] entrance and entrench the older [222:51] companies cuz the older companies are [222:52] the ones that have the resources to [222:54] comply. So they've got their ear to the [222:56] ground. They've got a ton of resources. [222:58] So when new data interoperability stuff [223:00] like there's one called TEFA that is in [223:02] the works. Epic can be the first to [223:05] implement it well. And so that can [223:08] further entrench them as they comply [223:11] correctly with new regulations and new [223:14] compliance. Yep. As we talked about, [223:17] there's probably pricing headroom at [223:19] this point, which may or may not be [223:21] great for their customers, but it's [223:22] great for Epic, and they probably can [223:24] start extracting more value if they [223:26] wanted to pull forward some of the [223:28] future into today. I would be really, [223:30] really surprised if they do that. I [223:32] would too. I don't think they will. Then [223:34] the last one is becoming a platform. [223:37] They really haven't done this to date. [223:38] They've built it all themselves, but I'm [223:40] curious if they start platformizing a [223:43] little bit more. and they've always had [223:45] little things here and there, App [223:46] Orchard, but I'd be curious if they at [223:49] some point make a real play to be a [223:51] robust platform layer upon which other [223:53] applications build on top of. To your [223:55] point about vertical software, maybe [223:57] that's just not the way this ends up [223:59] playing out. But they have the [224:00] opportunity if they want it and think it [224:02] would be more valuable than their [224:03] current path. Yep. All right. Should we [224:07] do quintessence? The biggest question I [224:09] have is why did it uniquely work in this [224:13] field to build the entire platform and [224:15] all the applications yourself? Any other [224:18] time you get that pitched, pass, I'm not [224:20] investing in that. That's a stupid [224:21] strategy. Why in this industry did it [224:23] work? So, I sort of phrased this as a [224:27] different question to myself, but I [224:28] think it's the same thing. I asked why [224:30] did Epic win, but I think it's the same [224:32] question here. And to me, it just [224:35] smacked me in the head. This is vertical [224:38] market software. This is the correct [224:41] playbook to win in vertical market [224:44] software. And it is very different than [224:47] building horizontal market software. H [224:51] that's interesting because in vertical [224:54] market software you are only serving one [224:57] category of user. the deeper and deeper [225:00] and deeper you can go into their [225:03] business operations and solving their [225:06] problems the better. Whereas if you're [225:09] building Slack or Salesforce let's pick [225:13] a horizontal complex piece of software [225:16] Microsoft Office. Yeah, Microsoft Office [225:18] great canonical example or Windows [225:20] actually right? You need to be really, [225:24] really, really careful about going too [225:26] far into one customer segment. And this [225:30] goes all the way into product [225:31] development. You hear all the time from [225:33] really great product people and [225:34] engineers at the big horizontal [225:36] technology companies that the shest way [225:40] to design bad products is to listen to [225:42] your customers, right? You internalize [225:44] the feedback and then you decide what [225:45] the best feature would be based on their [225:47] own experience or based on collecting [225:49] all the different experiences. And that [225:51] I think is generally right for building [225:53] horizontal software and products. [225:56] However, when you're building vertical [225:57] market software and products, you [225:59] absolutely want to listen to your [226:00] customers. And that is what Epic is [226:02] really good at. That's interesting. And [226:04] you want to build exactly what they want [226:05] and you want to understand who your [226:07] customer is and your customer is the CIO [226:10] and the CIO and the CFO of the hospital [226:12] system. Right. I now agree with you. My [226:16] answer it probably just layers on top of [226:18] why it's healthc care specific not just [226:21] vertical specific. It is all the we were [226:24] talking about carrots and sticks earlier [226:25] all the sticks in this industry. If your [226:28] software is bad they die. So the right [226:32] thing is one big integrated approach [226:34] rather than any risk at all of [226:37] discontinuity between different [226:40] applications. So that's one. Two is all [226:43] the compliance stuff. Again, if you have [226:47] any data leak, whoa, it's [226:48] catastrophically bad for HIPPA. If you [226:52] don't function correctly, oh, you get [226:54] your Medicare, Medicaid subsidies [226:57] crushed for violating, you know, [226:59] meaningful use stuff. There's just all [227:01] these sticks with any mistakes. And so, [227:04] in a situation where you can't make any [227:05] mistakes, the single vendor playbook is [227:08] the right playbook. Even if it's going [227:09] to take 47 years to build the dominant [227:12] company. Yep. You have to write a lot of [227:15] software and it all has to work together [227:17] really well. Yep. And one other thing I [227:20] was thinking about about this idea of [227:22] Epic as a vertical market software [227:25] company. I don't think we've ever really [227:28] covered any other vertical market [227:30] software companies on the show before. [227:32] No, I don't think we have either. We [227:34] sort of have. And I was thinking about [227:35] why is that? I'm like, oh well that [227:36] makes sense. Generally vertical market [227:38] software companies are not going to get [227:39] that big. You're limited by your [227:41] vertical market. This is a rare case [227:45] where because the market they operate in [227:48] is so big, the American healthcare [227:50] market and it's something we all use and [227:53] it's something we all use and it's [227:54] something that is so important. You [227:56] actually can get a really valuable [227:58] company built serving just one and [228:02] market. And then I was thinking about it [228:03] and I was like, Epic has got to be the [228:06] most valuable vertical market software [228:08] company in the entire world. What else [228:10] could possibly be up there? And so I [228:12] started racking my brain. I was like, [228:13] Bloomberg was the only one I could come [228:15] up with that could plausibly maybe be as [228:18] valuable. We got to do a Bloomberg [228:20] episode someday, but that'll be our [228:21] second vertical market software company [228:23] that we cover. But then I was googling [228:25] and I was thinking about it and I was [228:27] like, well, people talk about Viva [228:28] Systems. Yeah, Viva, which is also [228:31] healthcare. They're probably the largest [228:33] public company, but I was like, "Yep, [228:34] they're definitely not worth as much as [228:36] Epic." And they are also operating in [228:37] healthcare. They're not worth as much as [228:39] Epic. No, it's about a $20 billion [228:41] market cap company. Oh, wow. Yeah, I [228:43] don't think so. You know, the only other [228:45] company I could come up with in the [228:47] vertical market software industry that's [228:49] also in this kind of big league is [228:51] Constellation Software. Of course, you [228:53] know, fan favorite friend of ours, but [228:55] they bundle together a bunch of [228:56] different verticals. Exactly. there a [228:59] roll up. And I think some of their [229:02] biggest companies are vertical market [229:05] healthcare and healthcare IT companies [229:07] within Constellation. Yep. Mark Leonard, [229:10] I'm sure, is a big epic fan. Or maybe [229:12] not fan because he probably operates [229:14] some competitors. Yeah, I'm sure he has [229:16] respect for Judy. I am sure. I mean, [229:20] she's got a gold medal in the capitalism [229:22] Olympics, so you got to respect it. All [229:25] right, that winds it down. Should we do [229:27] some carveouts? Let's do carveouts. [229:28] Let's do something fun. All right. I've [229:31] got two. I've got a regular carveout, a [229:33] today carveout, and then I have a [229:35] preemptive carveout. Great. My today [229:38] carve [229:39] out a friend of mine turned me on to a [229:42] very, very popular OG YouTube video that [229:46] I had never seen. Ken Block, the rally [229:49] car driver, did a YouTube video back in [229:53] 2012 of San Francisco. They shut down [229:57] the streets in San Francisco. Whoa. And [229:59] he did this [230:00] incredible rally car 10-minute long [230:03] video through San Francisco. It's like [230:05] watching the old movie Bullet with Steve [230:07] McQueen and the iconic Yeah. Yeah. But [230:08] it's like that with a 600 horsepower [230:11] rally car driven by one of the best in [230:13] the world. I can't believe I'd never [230:15] seen this before. And it's like my s I'm [230:16] like, "Oh, I know those blocks. I live [230:18] there." You know, so cool. Amazing [230:21] video. and he sadly died recently in an [230:24] accident, which is why I was kind of [230:26] back in the news. But I did some more [230:27] research into him and I knew nothing [230:29] about him before. He was one of the [230:31] co-founders of DC shoes, you know, like [230:33] the skateboarding shoes. He was one of [230:35] the two co-founders. Then after they [230:37] sold it to Quicksilver, he got into [230:40] professional rally car driving, became [230:42] one of the best professional rally car [230:44] drivers, filmed this whole series of [230:46] amazing videos in cities. Super awesome. [230:49] I can't believe I'd never seen it [230:50] before. So, we'll link to it in the show [230:51] notes. Huh. I don't know anything you [230:53] just said, so I got to go look it up cuz [230:56] none of that made any sense to me. [230:57] Right. It's a short enough very [230:59] impressive video. Very cool to watch. [231:01] Sweet. That's my main carve out. And [231:03] then my preemptive carveout is I am so [231:06] excited for the Switch, too. Ah, if it [231:09] ships, if they start taking pre-orders [231:11] again. Uh, yeah, maybe by the time this [231:13] comes out. I'm not worried about it [231:14] hitting the ship date. I think they got [231:16] to sort out the US price and everything, [231:18] but a it looks incredible. Basically [231:21] everything on our Nintendo episodes that [231:24] we've hoped Nintendo would do with the [231:27] successor system to the Switch, they're [231:29] doing full backwards compatibility, [231:31] maintaining the online accounts. [231:34] System looks incredible. Shipping with a [231:36] new Mario Kart. It's going to be just [231:38] awesome in and of itself. I can't wait [231:39] to have one for me. Also, my older [231:42] daughter, like I have been waiting since [231:44] the day she was born to play video games [231:46] with her, and by the time it comes out, [231:49] like I'm thinking she might finally be [231:51] ready for Mario Kart, and I'm just so [231:53] excited. Sweet. Oh, that'll be so fun. [231:57] All right, mine. Go for it. I just [232:00] rewatched the movie Knives Out. Oh, [232:03] yeah. Good one. Daniel Craig. Extremely [232:05] fun. So good. It's just pure popcorn. [232:07] It's visually stunning. Everyone loves a [232:10] Who Done It. Daniel Craig's awesome. The [232:12] whole cast is great. So, if you haven't [232:14] seen that, you'll really enjoy Movie [232:16] Night. My second one is I must be the [232:19] last person to discover this album. I've [232:22] been listening to the album Brat by [232:24] Charlie XCX. It is awesome. Well, you're [232:27] definitely not the last person because [232:29] don't worry, Ben. I will always be [232:30] behind you in discovering music. I just [232:32] got to say like I'm the oldest malist. [232:36] And also the fact that this came out [232:38] 2024. Everything about this is like a [232:40] ridiculous carveout. But I was watching [232:42] Charlie XCX at the Grammys in her [232:45] performances there and ever since I've [232:46] just had the album on repeat and it's [232:48] amazing. It'll transport you to another [232:51] and much more fun world than the one you [232:53] are currently sitting in in front of [232:54] your computer. [232:55] I love it. And then my last one is Odza [232:58] just released an album called Music to [233:01] Refine to and it is remixes from [233:04] Severance. It's just good. Leave it on [233:06] lowfi in the background while you're [233:08] doing work. I love Odza. I love [233:10] Severance. It's my world colliding. I'm [233:13] grateful that all of us have you too. [233:15] Keep us up to date on music here. My [233:17] music is Frozen in Amber at like 2006. [233:20] Uh if you want to feel like you're at [233:22] like a much cooler party than you've [233:23] been invited to in 20 years, listen to [233:25] Brat. [233:25] [Laughter] [233:27] Great. H. All right. One last thing, [233:30] listeners. Remember, New York City, July [233:34] 15th. We can't wait to see you there. [233:38] acquired.fm/nyc to make sure that you [233:41] get the latest and you're first to know [233:43] as soon as we say what the heck is going [233:44] on. On that note, thank you to our [233:46] partners this season, JP Morgan [233:48] Payments, Service Now, Crusoe, and [233:50] Fundrise. You can click the link in the [233:52] show notes to learn more. Thank you to [233:54] all the people who spent time with us on [233:56] this one. We talked to dozens of people [233:58] and I think we can probably thank four [234:00] or five of them specifically on air as [234:02] always. Arvin Navaratnam at Worldly [234:04] Partners for his great writeup linked in [234:06] the show notes. Brendan Keeler who [234:08] writes the Substack health API guy. He's [234:10] awesome. Brian Lawrence is the founder [234:13] of Oakcliffe Capital and was a great [234:16] thought partner in this episode. Lan [234:18] Wang, a healthcare technology investor [234:21] and managing partner of JSL Health [234:24] Capital. Scott Gaines, a longtime IT [234:27] executive from Highland and Cover My [234:29] Meds. Patrick Wingo, the head of [234:31] research at Eleon, someone who has [234:34] studied the health IT space very [234:36] diligently and formerly at Palunteer. [234:38] John Bertrren, the CEO of digital [234:40] diagnostics, an adviser at 8VC and a [234:42] former Epic employee. He really helped [234:44] me understand the culture and what a [234:46] unique place it is to sort of grow and [234:48] develop there. And to Dr. Abed Mir, a [234:51] Stanford physician and investor who used [234:54] Epic firsthand before becoming an [234:56] investor and was inspired to become an [234:58] Epic shareholder himself. David, I know [235:01] you've got a few folks as well. Yeah, a [235:03] whole number of health system CIOS who I [235:06] spoke with. Huge thank you to all of [235:08] you. One that I wanted to call out [235:09] specifically was Mike Feffer, the CIO of [235:11] Stanford Medicine. two fun ones. Two of [235:15] Jenny and my friends from our Princeton [235:18] undergrad days who are now doctors at [235:21] big epic sites. So Molly Cantor who's a [235:24] doctor here at UCSF in San Francisco and [235:27] Karine Davala who is a doctor at MGH at [235:30] Harvard MGH in Boston. Nice. Got to [235:33] speak with both of them about the [235:34] physician experience of using Epic. And [235:38] then finally, we have a number of people [235:40] to thank at Epic itself. Since there [235:44] really isn't any canonical long form [235:47] piece about the company out there, we [235:49] thought we usually don't do this, but [235:51] let's just email Judy and see if we can [235:53] chat. And yes, she responded and we did. [235:55] Thank you, Judy, for spending time with [235:57] us and helping us understand all the [235:59] missing details we had on Epic's company [236:02] history. And of course also to Carl [236:04] D'vorak, Sumit Rana and Lee Leavon also [236:07] from Epic who spent time with us and [236:09] walked us through the history. Yeah, we [236:12] really really appreciate it. Yep. [236:14] Listeners, if you liked this episode, go [236:16] check out anything else in the acquired [236:18] back catalog. We should specifically [236:20] call out Novo Nordisk if you are into [236:22] the healthcare space. You've also got [236:25] lots of interviews available on ACQ2. [236:27] Our most recent one was a conversation [236:29] with Bill McDermott, the CEO of Service [236:31] Now. After you finish this episode, come [236:33] discuss it with other smart acquired [236:35] Slack members at [236:37] acquired.fm/slack. And with that, [236:39] listeners, we'll see you next time. [236:41] We'll see you next time. Who got the [236:44] truth? Is it you? Is it you? Is it you? [236:47] Who got the truth now? Huh? [236:52] [Music]