TSMC Clinches Global Chip Monopoly as Intel Manufacturing Falters

Acquired

TSMC dominates the global technology ecosystem as a horizontally integrated monopoly because its aggressive capital scale and process power drove the asymptotic cost of computing toward zero.

AI-assisted summary of TSMC — original source. Check the original for full context.

Source published: 2025-01-21

Detailed notes

## Background Published on January 21, 2025, on YouTube by the channel Acquired, this remastered episode dives into the history of the Taiwan Semiconductor Manufacturing Company (TSMC). The description highlights the company's 1987 founding by Morris Chang and its rise to becoming the quiet technology giant that manufactures leading-edge chips for tech leaders like Nvidia, Apple, and AMD.

## TL;DR This episode traces the chronological history of TSMC and its founder, Morris Chang, from his early life in China and career at Texas Instruments to his revolutionary creation of the 'pure-play foundry' model in Taiwan. The narrative explores how TSMC established a horizontally integrated monopoly by heavily investing in extreme ultraviolet (EUV) lithography and capturing crucial partnerships with tech giants like Apple and Nvidia. By assuming the capital risks of manufacturing, TSMC built an insurmountable economic moat driven by scale economies and process power, effectively making it the foundational backbone of the modern global computing infrastructure.

## Atmosphere & style The episode carries an energetic, deeply analytical, and celebratory tone, blending technical precision with storytelling. The hosts share an engaging, peer-to-peer dynamic filled with moments of high-stakes tension when detailing geopolitical risks, alongside a sense of awe at the sheer scale of TSMC's manufacturing alchemy. Listening feels like an epic corporate deep-dive that balances historical reverence for Morris Chang's perseverance with sharp business frameworks and a touch of wit.

## Key takeaways - [00:00](https://www.youtube.com/watch?v=tuB3_fJyC5Q&t=0s) TSMC represents a foundational pillar of modern computing, serving as the sole manufacturer of leading-edge chips for major tech giants including Nvidia, Apple, Broadcom, Qualcomm, AMD, and Intel. - [02:08](https://www.youtube.com/watch?v=tuB3_fJyC5Q&t=128s) Founder Morris Chang launched TSMC in 1987 at age 56, utilizing his deep industry background from Texas Instruments and establishing the company's long-term operational framework. - [56:10](https://www.youtube.com/watch?v=tuB3_fJyC5Q&t=3370s) Morris Chang pioneered the 'pure-play foundry' business model, which dictates that the company strictly acts as a contract manufacturer and never competes directly by designing its own proprietary chips. - [01:07:23](https://www.youtube.com/watch?v=tuB3_fJyC5Q&t=4043s) TSMC relies on an essential monopoly partnership with Dutch equipment manufacturer ASML, which supplies the highly complex extreme ultraviolet (EUV) lithography machines required for advanced chip production. - [01:19:20](https://www.youtube.com/watch?v=tuB3_fJyC5Q&t=4760s) The semiconductor manufacturing landscape has undergone massive consolidation, shrinking from over twenty competitive leading-edge companies in the early 2000s to a near-total dominance by TSMC today. - [01:38:28](https://www.youtube.com/watch?v=tuB3_fJyC5Q&t=5908s) The critical hardware partnership with Apple to exclusively manufacture iPhone processors starting with the iPhone 6 propelled TSMC into a supreme position within consumer tech supply chains. - [01:54:25](https://www.youtube.com/watch?v=tuB3_fJyC5Q&t=6865s) TSMC's strong economic moat is characterized by massive scale economies and unique process power, resulting in incredibly high switching costs for fabless tech companies attempting to port designs elsewhere. - [01:59:08](https://www.youtube.com/watch?v=tuB3_fJyC5Q&t=7148s) TSMC faces high geopolitical risks due to its extreme geographic concentration in Taiwan, rendering it a focal point of strategic and national security tension between the United States and China. - [02:06:11](https://www.youtube.com/watch?v=tuB3_fJyC5Q&t=7571s) Intel's decision in the mid-2000s to reject the original iPhone manufacturing contract based on short-term pricing is characterized as a historic strategic mistake that fundamentally empowered TSMC's growth. - [02:11:54](https://www.youtube.com/watch?v=tuB3_fJyC5Q&t=7914s) By executing a high-investment 'if you build it, they will come' strategy, TSMC effectively built the underlying platform that reduced the cost of compute toward zero and enabled the global software boom.

## Detailed walkthrough ### [00:00](https://www.youtube.com/watch?v=tuB3_fJyC5Q&t=0s) Remastering the TSMC Story

Ben Gilbert and David Rosenthal open the episode by highlighting that their previous coverage of TSMC (Taiwan Semiconductor Manufacturing Company) remains one of the most highly praised episodes in the history of the *Acquired* podcast. Interestingly, that original recording predates their massive Nvidia series. It was taped back in 2021 when the show's audience was a mere 12% of its current size in 2026, meaning the vast majority of listeners have never actually heard it. In light of this, they decided to pull the raw audio tracks from the archives and remaster the episode from scratch.

Since that original 2021 recording, the semiconductor industry has exploded in global importance, and TSMC has cemented its status as "essentially the only manufacturer of the Leading Edge chips." Ben lists off TSMC's massive footprint, noting that they make the primary chips powering every single MacBook and iPhone shipped today. Furthermore, they are the manufacturing powerhouse driving the current AI wave, producing 100% of Nvidia's chips, alongside hardware for massive fabless companies like Qualcomm, AMD, Broadcom, and hyperscalers such as Amazon Web Services (AWS). In a surprising twist, David points out a little-known fact: TSMC even manufactures a substantial volume of chips for their historical rival, Intel.

``` 2021 Market Cap: ~ $550 Billion 2026 Market Cap: Over $1 Trillion ```

David shares a striking piece of data from their updated research: since 2021, TSMC's market cap has doubled, surging from $550 billion to well over a trillion dollars. Ben adds a jaw-dropping statistic that highlights just how much of a geopolitical and economic outlier the company is: TSMC and Saudi Aramco are the only trillion-dollar companies in the entire world that are not headquartered on the West Coast of the United States.

[02:14](https://www.youtube.com/watch?v=tuB3_fJyC5Q&t=134s) Season 9, Episode 3: The Tech Superpower

After the show's signature musical intro, Ben formally kicks off Season 9, Episode 3 of *Acquired*. He introduces TSMC as a classic example of a company that "most people have never heard of," yet stands as the ninth-largest company in the world. The narrative arc of TSMC centers heavily on its legendary founder, Morris Chang, who started the company at the age of 56, retired at 74, and then staged a dramatic comeback at age 78 to ink the monumental deal to manufacture all of Apple's silicon chips.

The scale of TSMC's ubiquity is staggering. Beyond smartphones, laptops, and AI hardware, their components are embedded in everyday cars, smart home devices, and advanced military tech like fighter jets. The hosts underscore the high-stakes geopolitical tension hovering over the business: this critical linchpin of the global economy sits entirely on an island that some consider a sovereign nation, but the People's Republic of China claims as its own territory. Ben notes that they had actually kept TSMC on their Google Doc agenda for two and a half years, but the current global chip shortage—which has disrupted everything down to halting Ford F-150 production lines—made it the absolute perfect moment to execute this deep dive.

- **Founder Track:** Morris Chang founded TSMC at 56, retired at 74, and returned at 78. - **Geopolitical Backdrop:** Operating out of Taiwan, the company sits directly in the crosshairs of US-China relations. - **The Semiconductor Miniseries:** This episode links conceptually with their previous deep dives into ARM, Sequoia Part 1, and P.A. Semi.

[04:53](https://www.youtube.com/watch?v=tuB3_fJyC5Q&t=293s) The Early Life and Wartime Flights of Morris Chang

David takes the lead on the historical narrative, tracing the origins of our protagonist, Dr. Morris Chang, back to Ningbo, China, where he was born in July 1931. To frame the timeline, David notes that Morris was born just one year after Warren Buffett was born in Omaha, Nebraska, setting up a lifetime of interesting parallels. For context on Chinese geography, David chuckles that Ningbo is a "small city" just south of Shanghai—meaning it holds a casual eight million people by today's standards. Morris’s father was a county official who later transitioned into bank management, allowing the family to enjoy a stable, well-to-do middle-class lifestyle during the pre-communist, pre-WWII era.

This stability shattered when Morris was just six years old as the Second Sino-Japanese War broke out. Morris and his mother were forced to flee mainland China for Hong Kong to escape relentless air raids and heavy fighting. However, tragedy followed them. On December 8, 1941—exactly three hours after the surprise attack on Pearl Harbor—Japanese forces attacked and invaded Hong Kong. This double-flight pattern came to define Morris's youth; they fled back to mainland China to live in Shanghai, only to be uprooted once again in 1948 when the Chinese Civil War erupted. By the time Morris turned 18, he had already survived three major wars: the Second Sino-Japanese War, World War II, and the Chinese Civil War.

- **Highest Civilian Honor:** Dr. Morris Chang holds the *Order of the Propitious Clouds with Special Grand Cordon*, the highest civilian honor in Taiwan, making him effectively a "Knight of Taiwan." - **Childhood Displacement:** Lived through three catastrophic conflicts before reaching adulthood, bouncing between Ningbo, Hong Kong, Shanghai, and back to Hong Kong.

[07:40](https://www.youtube.com/watch?v=tuB3_fJyC5Q&t=460s) "Sheer Ecstasy": Arrival in America and the MIT Pivot

In 1949, the exact same year the People's Republic of China was officially established, 18-year-old Morris Chang caught a massive break. With the help of an uncle living in Boston, he secured admission to Harvard University and emigrated to the United States. Decades later, Morris vividly recalled his arrival, describing his emotional state upon entering Harvard as "sheer ecstasy, almost disbelief." He viewed the United States at its peak of moral, political, and democratic leadership, marveling at the fact that it was the wealthiest and most stable country on Earth—a place where a person could count on the social and economic structures remaining intact a decade down the line.

However, Morris quickly encountered a practical dilemma. His parents were still stuck overseas, leaving him completely independent to fund his own life. Furthermore, racial dynamics in the early 1950s severely restricted professional mobility for Asian immigrants. Morris famously observed that respectable career tracks for Chinese individuals at the time were strictly limited to "Chinese laundrymen, Chinese restaurant owners, Chinese engineers, and Chinese professors." Because Harvard focused on turning out lawyers, politicians, and finance professionals, and lacked an undergraduate engineering program, Morris realized he was in the wrong place. After his freshman year, he packed his bags and walked down the street in Cambridge to transfer to MIT to pursue mechanical engineering.

- **Academic Acceleration:** After transferring to MIT in the fall of 1951, Morris completed both his undergraduate degree and his Master’s degree in mechanical engineering in just three years. - **The Tech Landscape of 1951:** Silicon Valley didn't exist yet; Fairchild Semiconductor hadn't launched, and early computing power relied on vacuum tubes or germanium-based transistors rather than silicon integrated circuits.

[11:21](https://www.youtube.com/watch?v=tuB3_fJyC5Q&t=681s) The Sylvania Job and Barroom Engineering Lessons

After completing his Master's degree at MIT, Morris intended to secure a PhD, but his academic ambitions hit a brick wall when he failed the rigid doctoral qualifying examinations twice. Ben and David pause to laugh about Morris’s dry, clever sense of humor, quoting how he later joked that the school was "kind enough" to let him fail a second time. Forced to find work as an entry-level engineer, legend has it that Morris received a mechanical engineering offer from Ford Motor Company for $479 a month, and a competing offer from Sylvania’s brand-new semiconductor division for $480 a month. When Ford refused to match the extra dollar, Morris took the job at Sylvania, inadvertently stepping into the nascent world of electronics.

``` Ford Motor Company Offer: $479 / month Sylvania Semiconductor Offer: $480 / month ```

David drops another piece of incredible *Acquired* lore: Donald T. Valentine, the legendary founder of Sequoia Capital, started his own career in Sylvania's semiconductor division at this exact same time. While Don was out enjoying sunny Southern California, playing water polo and falling in love with the West Coast, Morris was on a grueling grind as a junior engineer at Sylvania’s manufacturing plant in Ipswich, Massachusetts.

Because his formal background was strictly in mechanical engineering, Morris knew practically nothing about electronics. To bridge the gap, he lived out of a company-sponsored hotel room and spent his nights intensely studying William Shockley’s foundational 1950 textbook, *Electrons and Holes in Semiconductors*. Recognizing that a senior engineer at the plant was an alcoholic who frequented the hotel bar every single night, Morris developed a brilliant strategy. He would study alone in his room for hours, then head down to the bar late at night to buy drinks for his older colleague. In exchange, he would open his textbook on the counter and grill the senior engineer on complex electrical engineering concepts. Morris later credited this unorthodox setup as his primary source of schooling, noting that while the colleague "didn't solve all my problems, he solved enough so that I could move ahead." This relentless self-education routine continued quietly for three straight years.

[17:44](https://www.youtube.com/watch?v=tuB3_fJyC5Q&t=1064s) Leaving Sylvania for the Real Semiconductor Juggernaut

Morris Chang was "burning the candle at both ends," working tirelessly and spending his time learning the industry inside out. However, as he came up to speed, he realized that Sylvania was "not really the right bus to be on" if he wanted to truly advance in this new emerging industry. Future industry giants like Don Valentine came to the same conclusion, leaving Sylvania for Raytheon and later Fairchild. For Morris, the breaking point crystallized during a presentation by a senior Sylvania manager at the plant, who uttered a line that stuck with Morris for the rest of his life: "We at Sylvania cannot make what we can sell, and we cannot sell what we can make." Realizing this was an undeniable signal to move on, Morris left Sylvania for greener pastures.

Instead of heading straight to California or Silicon Valley—which the speakers note was largely a marketing phenomenon at the time—Morris went halfway across the country. In the late 1950s, the undisputed titan of the semiconductor world was not on the West Coast; it was Texas Instruments (TI) in Dallas, Texas. While younger generations know TI primarily for making graphing calculators, the company was the ultimate industry incumbent back then. Founded in the 1930s, TI actually got its start making seismic activity measurement instruments for oil exploration, effectively acting as "the TSMC" or technology provider to oil companies before transitioning into digital computing.

- TI was the absolute technology leader of the era, home to Jack Kilby, who co-invented the integrated circuit simultaneously with Fairchild's Bob Noyce and received the Nobel Prize for it. - Jack Kilby famously noted that the integrated circuit would "reduce the cost of electronic functions by a factor of a million to one," an unprecedented scale of deflation in human history.

[21:15](https://www.youtube.com/watch?v=tuB3_fJyC5Q&t=1275s) Back to Basics: Transistors, Silicon, and Integrated Circuits

To understand the trajectory of the industry, the speakers step back to establish some fundamental definitions. They describe a transistor not as the sub-microscopic element found on modern silicon dies, but as a small encased piece of circuitry with three prongs: an input, an output, and a control mechanism. Functioning essentially as a switch, it controls whether electrical voltage or current flows through, allowing equipment to register binary states of off (0) or on (1). While transistors can be made out of many materials, silicon became the standard because it is a semiconductor (a metalloid). It possesses properties that allow engineers to modify whether current flows or not, making it far easier to create an adjustable switch compared to an absolute conductor like copper or a non-conductor like wood.

The conversation shifts to why silicon triumphed over germanium, which was the dominant material early on. Germanium was rare and expensive, whereas silicon is fundamentally made from sand, making it "the second most plentiful minable element on Earth." The real magic happened with the invention of the integrated circuit (IC) by Noyce and Kilby in 1958. Before the IC, switches were manufactured one at a time and manually wired together—reminiscent of plugging vacuum tubes into old computers. The IC allowed engineers to place multiple switches onto a single substrate.

- Modern 5-nanometer processors manufactured by TSMC pack billions or even trillions of switches onto a single tiny integrated circuit. - Morris Chang joined Texas Instruments in 1958, the exact same year the integrated circuit was invented, and a full decade before the founding of Intel.

[24:28](https://www.youtube.com/watch?v=tuB3_fJyC5Q&t=1468s) The IBM Project and Morris Chang's First Big Managerial Win

Upon arriving at TI, Morris was assigned to a troubled project. TI had secured a contract with IBM, which was developing its first transistor-logic mainframe computer, the IBM 7090, to replace vacuum tubes. Anticipating massive demand, IBM broke from its tradition of manufacturing everything internally and sought TI as a "Second Source" contractor to manufacture chips based on IBM's designs. This arrangement effectively operated like an early contract manufacturing or foundry business. However, production was going terribly. IBM's own internal plant was achieving only a 10% yield (meaning 90% of the chips failed), while TI’s production line was sitting at a catastrophic 0% yield when Morris arrived. As Morris later recalled, "the supervisor was concerned, the operators were concerned, everybody was concerned."

Leveraging his background as a mechanical engineer, Morris began tinkering with the chemical and mechanical optimization of the production line. Within four months, he miraculously raised TI's manufacturing yields to 20%—doubling the efficiency of IBM’s own internal operations. This engineering feat caught the attention of TI President Pat Hagerty. When IBM sent engineers down to investigate, expecting Morris had simply gotten lucky, he walked them through the specific theories he had tested and proved why his experimental process worked.

- This success earned Morris his first major promotion: building and leading a germanium transistor development department with over 20 engineers reporting to him. - Recognizing his talent, TI sponsored Morris to get his PhD at Stanford, paying his full salary and tuition; Morris completed his doctorate in just two and a half years by remaining incredibly focused, returning to TI in 1964.

[28:17](https://www.youtube.com/watch?v=tuB3_fJyC5Q&t=1697s) Learning Curve Pricing and Global Dominance

By 1964, the industry was shifting rapidly toward silicon as manufacturers figured out how to "dope" the material to make it scale cost-effectively. The speakers emphasize that semiconductor manufacturing has always been one of the most complex scientific endeavors in human history, requiring layers of innovation across chemistry, physics, and mathematics to achieve even marginal yields. By 1967, Morris was promoted to General Manager of a division within TI's semiconductor business, where he engineered a massive breakthrough on the business side. He noticed that because building new fabrication plants (fabs) was incredibly capital-intensive, the industry standard was to charge premium prices for new product lines to recoup fixed costs immediately.

Morris realized this strategy was fundamentally flawed. Because of the manufacturing learning curve, initial yields are always low, meaning a fab needs immediate, massive volume to run at maximum capacity, accelerate learning, and drive yields up to profitable levels. High initial pricing suppressed the very demand needed to optimize the factory. To solve this, Morris hired a then-small firm called the Boston Consulting Group (BCG) to validate a radical strategy: "learning curve pricing." They decided to price chips low right out of the gate to capture massive volume, and then automatically cut prices every quarter even if the market didn't demand it.

- The strategy was highly controversial, and many competitors thought TI was being foolish by lowering prices unnecessarily, but it allowed TI to capture immense market share and block rivals. - This aggressive pricing model, combined with manufacturing excellence, built the TI integrated circuits division into the largest and most profitable IC business in the world, soundly beating competitors like Fairchild and National Semiconductor just as Intel was being founded.

***

http://googleusercontent.com/interactive_content_block/0

[34:34](https://www.youtube.com/watch?v=tuB3_fJyC5Q&t=2074s) Vanta and the Value of Automation

The episode takes a moment to highlight Vanta, a continuous monitoring and compliance platform. Reflecting on their four-year history of working with the company, the hosts recall initially viewing Vanta as a product that simply allows businesses to "focus on just making your beer taste better"—meaning companies could outsource non-differentiators to free up time. However, the platform has advanced to the point where it does compliance and real-time monitoring better than an organization could ever manage internally.

According to a recent IDC white paper, Vanta customers achieve an average of $535,000 per year in benefits, with the platform paying for itself in just three months. Acting as a single software pane of glass, it connects to external services via APIs to eliminate manual work, features new AI capabilities, and allows enterprises to build private integrations for internal systems.

* **Vanta Client Scale:** Over 9,000 customers globally, including Atlassian, Quora, and Factory. * **Acquired Listener Offer:** $1,000 of free credit by visiting `[vanta.com/acquired](https://vanta.com/acquired)`.

[36:17](https://www.youtube.com/watch?v=tuB3_fJyC5Q&t=2177s) Three Perspectives on a Career Sidetracked at TI

Following his massive successes, Morris Chang was promoted to Vice President at Texas Instruments (TI) in 1972, placing him just one level below the CEO and in charge of the entire global semiconductor business. He was the obvious leading candidate to become the next CEO, which aligned perfectly with his personal ambitions. However, "history turns on a knife point," and Morris never got the top job. The hosts analyze three distinct viewpoints explaining why his trajectory altered so dramatically.

The first viewpoint is that Morris faced ethnic discrimination. Although he was likely already an American citizen by this point, the hosts note it would not be surprising if being ethnically Chinese played a role in him being passed over, despite a lack of formal documentation.

The second perspective, which Morris himself acknowledges, involves a strategic corporate reallocation. TI was launching its consumer products division, which was struggling. In 1978, after Morris had spent six years successfully running semiconductors, Chairman and CEO Mark Shepard reassigned him to head consumer products, operating under the corporate wisdom of the era that "a good manager could manage anything."

Morris later disagreed with this philosophy, noting that consumer products required an entirely different, nebulous competency rooted in consumer appeal and marketing rather than just technology and costs. Five years later, in 1983, with the division still struggling, Morris was demoted to "head of quality and people effectiveness," a move he described as being "put out to pasture."

The third perspective blames a massive strategic oversight at TI. During the mid-1970s, the industry transitioned to the metal oxide semiconductor (MOS) process, a precursor to CMOS. Despite having top-tier engineers, TI dropped the ball. The corporate culture did not mirror the fluid, equity-driven dynamics of Silicon Valley; TI expected "company man" loyalty. Consequently, key innovators grew frustrated and left. Notably, Motorola poached a massive contingent of MOS engineers from TI.

* **LJ Sevin:** A prominent TI engineer who left to found Mostek and later became an early venture capitalist by founding Sevin Rosen Ventures.

[40:27](https://www.youtube.com/watch?v=tuB3_fJyC5Q&t=2427s) The $X86$ Off-Ramp and the IBM PC RFP

The consequences of TI's talent drain culminated in a massive historical turning point in 1980. IBM issued a secret Request for Proposal (RFP) out of a Skunk Works division in Boca Raton, Florida, for a microprocessor to power a top-secret initiative: the IBM PC. Because of Morris’s deep historical relationship with IBM, a healthy TI should have been the definitive choice for the contract.

Instead, the contract went to Intel's 8088 processor. This single decision standardized the next generation of personal computing architecture on $x86$, routing billions of dollars in enterprise value away from TI and toward Intel and Microsoft. The hosts describe this as a major missed exit on the highway of history, effectively ending Morris's aspirations at TI. At age 52, in 1983, Morris resigned from the company after 30 total years of service.

[42:43](https://www.youtube.com/watch?v=tuB3_fJyC5Q&t=2563s) The General Instrument Setback

Despite his fallout at TI, Morris remained highly respected within the broader semiconductor ecosystem. Determined to be a CEO, he narrowed his next steps down to two options: becoming a venture capitalist or joining a competitor called General Instrument (GI). He chose GI, moving from Texas to Manhattan to serve as Chief Operating Officer with the explicit corporate understanding that he would succeed the sitting CEO within a few years.

The move quickly proved to be a cultural mismatch. While TI was an R&D-focused powerhouse dedicated to pushing technology forward, GI operated like a "proto-tech private Equity Firm." Their business model centered on financial engineering—acquiring independent chip companies or corporate divisions, cleaning up their operational metrics, and flipping them for a profit. Repulsed by this approach, Morris resigned after only a year. Within an 18-month window, he had suffered two catastrophic career setbacks, concluding that his dream of leading a major US enterprise was completely dead.

[44:52](https://www.youtube.com/watch?v=tuB3_fJyC5Q&t=2692s) The Taiwanese Recruiting Mission

The narrative shifts to how a "washed-up" executive in his mid-50s ended up creating TSMC in Taiwan. During his tenure running semiconductors at TI, Morris had visited Taiwan a handful of times in the 1970s to negotiate building a TI-owned manufacturing plant. He had no deep cultural roots there, noting that Taiwan felt like a "strange land" to him; if he considered any place home, it was America. However, those early trips introduced him to key Taiwanese government officials.

By the mid-1980s, Taiwan was strictly a low-end manufacturing economy with an average corporate gross margin sitting at a razor-thin 4% to 5%. Local manufacturing consisted of cheap commodities like Barbie dolls, toys, and clothes. The hosts point out that under Hamilton Helmer's *7 Powers* framework, such low margins mean all economic surplus is arbitraged away in a race to the bottom. Desperate to escape this economic trap, the Taiwanese government established the Industrial Technology Research Institute (ITRI) in the 1970s, intending for it to serve as the nation's "Bell Labs" by executing technology transfers from the West.

In 1985, Minister KT Lee—venerated today as the "father of taiwan's economic miracle"—actively recruited Morris to move to Taiwan and run ITRI. It was a bizarre, counter-intuitive career move for a former titan of American tech. His friends, colleagues, and his wife all begged him not to take it, pointing out that top-down government innovation mandates rarely result in global commercial powerhouses. However, with his marriage falling apart and his US career stalled, Morris wanted a total change of scenery. He viewed the ITRI presidency as a comfortable, pseudo-retirement.

* **Morris's Financial State:** Secure but modest by 1985 standards, aided by the high interest rates of the era which made living off principal highly achievable. * **Cultural Clash:** Morris arrived with a hard-charging, fast-paced American corporate mindset, immediately clashing with the entrenched bureaucratic culture of ITRI's government employees.

***

*Would you like me to continue with Part 4 of the transcript walkthrough?*

[51:07](https://www.youtube.com/watch?v=tuB3_fJyC5Q&t=3067s) The RCA Technology Transfer and the Origins of UMC

The transition of Taiwan into a semiconductor powerhouse didn't happen in a vacuum, nor did it begin in a fully realized democracy. At the time, Taiwan was just emerging from martial law, meaning government roles were essentially non-democratic "jobs for life." When Morris Chang arrived at the Industrial Technology Research Institute (ITRI), the local bureaucracy viewed him as an intrusive outsider. As Morris later recalled, "they considered me a foreigner who suddenly became their boss; they were scared of me and they were right to be scared of him."

Despite the tension, the Taiwanese government had managed one major victory right before Morris arrived: negotiating a technology transfer license with RCA. While it was a step forward, the technology was far from cutting-edge. It was roughly a decade old, putting the local operations at least a generation behind American leaders like TI, Intel, and Fairchild. By the time it was deployed on the ground in Taiwan, it lagged two and a half generations behind the leading producers, meaning it could only be used for low-end manufacturing.

Even with outdated technology, the government successfully spun out United Microelectronics Corporation (UMC). UMC handled both third-party fabrication and its own chip designs. Crucially, UMC would later spin out its chip design business into MediaTek, which has grown into a massive 50 billion dollar company today. This established a paved path and acquired IP for Morris, ensuring he wasn't starting from a completely standing start.

- **The RCA Deal:** A 10-year-old semiconductor technology transfer negotiated by the Taiwanese government in the 1970s. - **The Tech Gap:** The RCA process was 2.5 generations behind bleeding-edge manufacturing by the time it was operational in Taiwan. - **UMC & MediaTek:** United Microelectronics Corporation (UMC) was created from this tech, which later birthed MediaTek, a company valued at $50 billion today. - **Legacy Fabs:** Older factories are rarely shut down; they shift from bleeding-edge smartphone chips to older form factors like Automotive or Internet of Things (IoT) hardware.

[54:07](https://www.youtube.com/watch?v=tuB3_fJyC5Q&t=3247s) An Offer He Couldn't Refuse

While Morris was busy transforming ITRI into a high-performing organization, a legendary political figure, K.T. Lee, blindsided him with an entirely new mandate. Lee told Morris that running Taiwan's equivalent of Bell Labs wasn't enough; he wanted Morris to personally start a brand new semiconductor company in Taiwan and scale it into a global leader. Morris was taken aback, but in the non-democratic landscape of 1980s Taiwan, pushback wasn't an option.

For Morris, the stakes were incredibly high. He was on his third job in three years, and while he had enough money to survive without a salary, getting fired from ITRI would mean he was "legit done done done" in the industry. He famously compared the high-pressure demand to a classic cinema scene: "it was like in the movie The Godfather, it was an offer I couldn't refuse." Lee gave him a week to return with a business plan, only to call him a day later and demand he present it that coming Friday—giving him a mere three days to conceptualize what would become the ninth most valuable company in the world.

- **The Ultimatum:** K.T. Lee ordered Morris Chang to build a globally dominant semiconductor firm from scratch. - **The Timeline:** Morris was originally given one week to formulate the business plan, which was abruptly cut down to 3 days. - **Career Stakes:** This was Morris's third role in three years; failure or refusal meant political and professional exile.

[56:10](https://www.youtube.com/watch?v=tuB3_fJyC5Q&t=3370s) The Pure Play Foundry Paradox and "Real Men Have Fabs"

Morris's solution to the three-day challenge was to invent the "pure play foundry"—a company that would act strictly as a contract manufacturer, making chips for other firms without designing any of its own. Hindsight makes this look like an absolute stroke of genius, but at the time, the entire global tech community thought it was a terrible idea. Every dominant semiconductor player in America, Europe, and Japan designed and manufactured everything in-house.

The prevailing industry sentiment of the mid-1980s was perfectly captured by Jerry Sanders, the colorful co-founder and CEO of AMD, who famously proclaimed that "real men have Fabs." The belief was that chip manufacturing wasn't like putting together plastic toys; it was closer to high-tech "alchemy" that required total end-to-end control. Outsourcing assembly—like Apple does today with Foxconn—is entirely different from manufacturing a next-generation silicon design, a delicate process that people believed could only be done by the original creators.

- **The Pure Play Concept:** A contract manufacturing model that builds chips exclusively for third parties and abandons proprietary design. - **Jerry Sanders Quote:** "Real men have Fabs" — a definitive mid-1980s industry belief that AMD would ironically reverse in the 2000s by spinning off GlobalFoundries. - **Manufacturing vs. Alchemy:** The hosts contrast standard electronics assembly (like the iPhone) with the complex, chemical "alchemy" of fabricating silicon wafers.

[59:00](https://www.youtube.com/watch?v=tuB3_fJyC5Q&t=3540s) Finding the Market and Raising Capital

The foundational flaw of the pure-play foundry model was an existential one: the market didn't actually exist yet. It was a classic solution looking for a problem, as every existing chip startup was still focused on building its own manufacturing facilities. However, Morris possessed a unique insight from his days at TI and General Instrument. He had watched countless brilliant integrated circuit designers want to leave corporate giants to launch startups, only to be stopped because they couldn't raise the astronomical capital required to build a fabrication plant.

To bridge the gap until those "fabless" startups emerged, Morris had to rely on short-term realpolitik. He traveled the world pitching industry giants like Intel, TI, and Motorola, offering to take over their excess capacity or manufacture their older legacy components. The Taiwanese government committed to funding 50% of the required capital but mandated that Morris raise the other half from an established industry giant to guarantee an anchor customer.

Both Intel and TI turned Morris down flat. In a last-ditch effort, he secured a meeting with the Dutch electronics giant, Philips. Morris bluntly described Philips as "the first rung of the second raiders in semiconductors," but they were his only option. Philips agreed to supply 28% of the capital, bringing the total pool to $220 million. To close the remaining 22% gap, the Premier of Taiwan personally stepped in, strong-arming local Taiwanese business leaders into investing.

- **The Funding Breakdown:** 50% from the Taiwanese government, 28% from Philips, and 22% from local business figures. - **Total Initial Capital:** $220 million USD. - **The Rejections:** Silicon Valley titans Intel and TI both rejected Morris's fundraising and anchor-customer pitches. - **Core Market Insight:** Recognizing that capital constraints, not design capability, blocked talented engineers from launching independent chip startups.

[01:02:08](https://www.youtube.com/watch?v=tuB3_fJyC5Q&t=3728s) The Zero-Dollar Valuation and the Class of 1987

Perhaps the most shocking detail of the TSMC founding story is that the company was stood up at a pre-money valuation of zero dollars. Unlike the standard equity-heavy incentives of Silicon Valley, Morris Chang received absolutely zero founding equity in TSMC. He was treated purely as a government foot soldier, operating at the whim of a regime that essentially told him, "we hired you to do this... you are our foot soldier, you know, we are the mafia." Morris kept only his standard government salary, putting everything on the line simply because his career had hit a wall.

Morris eventually built his massive wealth by using his personal cash flow to manually buy up shares of his own company before it went public on the Taiwan Stock Exchange in 1994 and the New York Stock Exchange in 1997. The year of TSMC's official launch, 1987, turned out to be an *annus mirabilis* (miracle year) for semiconductors. Simultaneously, ARM was being formed as a joint venture between Apple, Acorn, and VLSI Logic, introducing a low-power architecture that the power-hungry, x86-dominated industry initially dismissed as "pathetic."

- **Pre-Money Valuation:** $0, with Morris Chang receiving no initial equity or stock options. - **TSMC Public Offerings:** Listed on the Taiwan Stock Exchange in 1994 and the NYSE in 1997. - **The 1987 Coincidence:** The exact same year TSMC was founded, ARM was formed as a joint venture between Apple, Acorn, and VLSI Logic. - **Cultural Contrast:** Silicon Valley structures rely on massive early founder equity, while TSMC was built via top-down state mandate.

[01:05:20](https://www.youtube.com/watch?v=tuB3_fJyC5Q&t=3920s) Deconstructing the Semiconductor Value Chain

The birth of TSMC and ARM completely reconfigured how the tech world functions, shifting the ecosystem away from monolithic entities that handled everything from raw transistors to marketing. Today, the semiconductor value chain is highly specialized, deeply integrated, and split across five distinct, interlocking categories.

The baseline layer consists of Architecture IP, dominated by companies like ARM. Directly downstream sits Electronic Design Automation (EDA), which provides the critical software ecosystem. The hosts compare EDA giants like Synopsys and Cadence to "Excel or like figma for chip designers." This specialized software allows fabless design companies—such as Apple, Qualcomm, and eventually AMD—to map out complex systems-on-a-chip before handing the blueprints over to the foundries and the specialized equipment manufacturers who supply the physical factory machinery.

- **The Modern Five-Part Value Chain:** 1. **Architecture IP:** Core structural designs (e.g., ARM). 2. **EDA (Electronic Design Automation):** Design software dominated by Synopsys and Cadence. 3. **Fabless Design:** Companies that design software-integrated silicon but own no factories (e.g., Apple, Qualcomm, AMD). 4. **Foundries:** Pure-play contract manufacturers (e.g., TSMC). 5. **Equipment Manufacturers:** Firms creating the complex machinery required to run a fabrication plant.

[01:07:23](https://www.youtube.com/watch?v=tuB3_fJyC5Q&t=4043s) The Semiconductor Ecosystem and the Monopoly of ASML

The semiconductor design world relies heavily on off-the-shelf intellectual property (IP) to keep up with intense timelines. Rather than designing every single block of a chip from scratch, a company needing USB functionality can simply buy that IP from specialized providers. This architecture defines the general rules of the game, allowing compiler writers to know exactly what Assembly Language they are targeting for a future chip. The industry map can be categorized into major segments: Electronic Design Automation (EDA), specialized IP blocks, and fabless chip design companies. However, preceding the actual foundries like TSMC is a critical layer of equipment manufacturers. Historically, this space was led by companies like Lam Research and Applied Materials in the US, alongside Tokyo Electron in Japan.

Today, a single Dutch-based powerhouse commands this equipment landscape: ASML (originally standing for Advanced Semiconductor Materials Lithography). While the magical and complex process of lithography evokes nostalgic comparisons to high school photography labs, ASML's modern reality is unmatched. They are the only company in the world that manufactures the most advanced chip fabrication machinery, and their largest customer is TSMC. Interestingly, ASML was founded in 1984 as a joint venture between Advanced Semiconductor Materials International (ASMI) and Philips. This deep, historic relationship between TSMC and its primary equipment provider underpins the entire industry; TSMC’s massive operating cash flow gives them the unique capital expenditure capacity to buy more ASML equipment than anyone else on Earth.

- **Founding Date:** ASML was formed in 1984 as a joint venture between ASMI and Philips. - **Industry Structure:** The ecosystem operates across discrete horizontal layers: EDA tools, IP licensing (such as Instruction Set Architectures), Fabless design houses, Equipment Manufacturing, and Foundries.

[01:09:35](https://www.youtube.com/watch?v=tuB3_fJyC5Q&t=4175s) Sponsor Break: JP Morgan Payments

The hosts take a brief moment to thank their big partner, JP Morgan Payments. When informed that the show was remastering its TSMC coverage, the team at JPM pointed out that most listeners overlook how critical international trade finance and liquidity are to global commerce. This segment functions as pure "catnip" for the hosts' love of business mechanics. JP Morgan operates a dedicated Trade and Working Capital group tasked with helping multinational companies mitigate risks, access liquidity, and execute the complex international payments necessary to accelerate growth.

To illustrate this, they pose a scenario where a global auto parts manufacturer builds up inventory for a highly complex engine component. The manufacturer is stuck waiting for a final demand forecast from their automotive customer before shipping the units. In the meantime, they must secure credit lines and keep their own supply chain stable. JP Morgan's group steps into these exact scenarios. Backed by a legacy of global expertise dating all the way back to 1799, the institution has guided businesses through centuries of economic expansion as well as severe geopolitical and supply chain instability.

[01:11:43](https://www.youtube.com/watch?v=tuB3_fJyC5Q&t=4303s) The Secret Lineage of Apple Silicon: From StrongARM to P.A. Semi

Shifting gears to a second "sidebar" story, the hosts revisit a piece of industry lore that predates the modern iteration of their show. Back around episode twenty, the podcast focused strictly on small corporate acquisitions. During research for this remastered episode, a fascinating lineage was uncovered regarding P.A. Semi—the company whose acquisition laid the groundwork for Apple Silicon. The story traces back to ARM, which was created in 1987 as an IP design firm that licensed its architecture rather than building chips. One of ARM's original licensees was the legendary Digital Equipment Corporation (DEC).

DEC took the standard ARM architecture, tuned it extensively for performance, and branded their high-powered version "StrongARM." In a bizarre twist of corporate history, this specific ARM-based product line inside DEC was eventually acquired by Intel—the absolute champion of the rival x86 architecture. Intel marketed the technology as "XScale" before eventually shutting it down. The core group of original DEC engineers found themselves miserable under Intel's roof, realizing that alternative architectures would never flourish there. Declaring "screw this," the tight-knit team of ARM specialists quit to start their own independent firm: P.A. Semi. Apple later bought P.A. Semi, meaning the most defensible, valuable hardware advantage Apple possesses today can be traced through a wild chain from ARM and DEC, through Intel, to P.A. Semi.

- **Nvidia Founding Details:** Nvidia was started in 1993, raised a mere $20 million in total funding, and never opened its own fabrication plant, relying almost 100% on TSMC. - **The Core Chain:** The DNA of Apple Silicon moved from ARM/DEC (StrongARM) → Intel (XScale) → P.A. Semi → Apple.

[01:13:55](https://www.youtube.com/watch?v=tuB3_fJyC5Q&t=4435s) Morris Chang’s Ultimate Venture Capital Bet and the TSMC Flywheel

In its early days, TSMC survived in a highly unstable short-term market. Founder Morris Chang was forced to beg old colleagues across American, European, and Japanese semiconductor firms to let TSMC manufacture their "dregs." Integrated Device Manufacturers (IDMs) only outsourced wafers to TSMC when they temporarily lacked internal capacity or when a chip line was actively losing money. The moment an IDM brought new capacity online, they yanked their orders away, effectively transferring their manufacturing losses straight to TSMC. Because of this volatile dynamic, the consensus at the time was that a "pure-play foundry" was a fundamentally broken business model. What the titans at Intel, Texas Instruments, and Motorola failed to see was the imminent rise of the fabless chip design industry. Chang had a unique vantage point; having previously considered a career in venture capital, he realized that many talented chip designers wanted to launch startups but were completely blocked by the prohibitive cost of building a Fab.

By launching TSMC, Chang effectively became the world's ultimate semiconductor venture capitalist. Instead of picking individual winners, he took an index bet on the entire future of chip innovation by building the infrastructure platform they all required. TSMC became the "AWS of chip companies," dropping the capital requirements for chip startups to 10% or 20% of historical needs. Once startups like Qualcomm, Broadcom, Marvell, and Nvidia got moving in the 1990s, they kickstarted an unstoppable flywheel. The explosion of the fabless market fueled TSMC's revenue. Boasting 50% gross margins and 40% operating margins, TSMC plowed these massive profits directly back into capital expenditures and advanced research. They quickly erased their technological deficit, caught up to the industry leaders, and began outpacing them. Better manufacturing led to smaller process nodes, which gave their fabless customers superior product performance, generating even more market demand and revenue to reset the cycle.

---

> "The semiconductor business is like a treadmill that speeds up all the time. If you can't keep up, you fall off." — *Morris Chang*

---

[01:19:20](https://www.youtube.com/watch?v=tuB3_fJyC5Q&t=4760s) Running the Table: The Horizontally Integrated Monopoly

The brutal reality of Morris Chang's treadmill analogy is clearly visible in the consolidation of the leading-edge market. In the early 2000s, when TSMC finally caught up to the pack at roughly the 150-nanometer node, there were 22 separate companies competing at the leading edge of semiconductor manufacturing. By the late 2000s, that number shrank to 14. By the mid-2010s, only six remained. Today, at the 5-nanometer threshold, there are only two: TSMC and Samsung. Intel has repeatedly fallen behind in its attempts to match them. As the industry eyes the next frontier at 3 nanometers, Samsung's timeline has already slipped to 2024, leaving TSMC to effectively run the table alone. When Apple presents its next iPhone slides boasting about a revolutionary new node, they will take all the credit—and TSMC is perfectly happy to let them, as their core mission is to empower their customers, not to market themselves.

This shift marks the complete transition of the semiconductor industry from a vertically integrated model to a horizontally integrated one. The sheer velocity of Moore's Law makes it impossible for a single entity to be world-class at every stage—from EDA software and heavy machinery to factory operations and architecture design. To stay alive on the manufacturing treadmill, a company must generate enough profit to sustain an eye-watering level of capital investment. TSMC recently announced a staggering $100 billion capital expenditure plan spanning three years, front-loading $30 billion in a single year. While the US and Chinese governments are waking up to the geopolitical vulnerability of having this manufacturing concentrated entirely in Taiwan, throwing money at the problem cannot easily fix it. Even if a government swoops in with a "bazooka of money" or a trillion dollars, they cannot buy their way to parity by next year; TSMC's insurmountable lead is protected by a unique marriage of massive scale economies and deeply entrenched process power.

| Era | Leading-Edge Competitors | Dominant Nodes / Landscape | |---|---|---| | **Early 2000s** | 22 Companies | ~150-nanometer process; TSMC joins the leading pack | | **Late 2000s** | 14 Companies | Industry begins rapid consolidation | | **Mid-2010s** | 6 Companies | High-cap spending thins out legacy players | | **2026 (Current)** | 2 Companies | 5-nanometer node; exclusively TSMC and Samsung | | **The Horizon** | 1 Company (TSMC) | 3-nanometer node; Samsung slips to 2024; Intel lagging |

[01:22:59](https://www.youtube.com/watch?v=tuB3_fJyC5Q&t=4979s) The Hardest Engineering in the World

The engineering behind advanced semiconductor manufacturing is so difficult that its learning curve spans decades. Even if a country like China throws a billion dollars at their state-backed foundry, SMIC (Semiconductor Manufacturing International Corporation), they still struggle to match the cutting-edge tech of TSMC. As discussed on Bloomberg's *Odd Lots* podcast, a reporter covering TSMC noted that while China is actively trying to buy and build the necessary fabrication facilities, they simply "wouldn't know what to do with it." This is not a question of intelligence; it is a recognition that operating extreme ultraviolet (EUV) lithography equipment is arguably the hardest physical feat in human history, mastered by only a tiny handful of people globally.

Geopolitical maneuvering heavily dictates who can even purchase this machinery. The Netherlands recently chose not to renew its trade agreement with China regarding ASML’s most advanced equipment—a decision widely reported to be the result of heavy prodding from various US presidential administrations over the last five years. The US government recognized that ASML is the sole bottleneck for the world's most sophisticated semiconductor manufacturing, making them a key strategic point to restrict tech flow to Chinese entities like SMIC.

- **Machine Cost:** A single cutting-edge chip-making machine from ASML costs between $200 million and $300 million. - **Logistics:** It requires four Boeing 747 cargo planes just to ship a single ASML lithography machine to a buyer like TSMC. - **Integration:** ASML sends its own dedicated crew of employees on-site to help customers assemble and continuously run the machines.

[01:25:24](https://www.youtube.com/watch?v=tuB3_fJyC5Q&t=5124s) The Alchemy of Extreme Ultraviolet (EUV) Lithography

Manufacturing chips becomes exponentially harder as transistor density increases. While older chip fabrication occurred at scales around 150 nanometers, modern processes like Apple’s M1 chip utilize a 5-nanometer process. This presents a massive physics problem: the wavelength of regular white light is 193 nanometers, which is far too wide to etch such tiny features. Early human ingenuity bypassed this by shooting light through lenses and water, but that technique maxes out at roughly 11 nanometers.

To break past the 11-nanometer wall, ASML had to pioneer Extreme Ultraviolet (EUV) lithography, a process the speakers characterize as pure "alchemy." On one side of the machine, droplets of molten tin are dropped. From the other side, a highly specialized laser pulses and strikes the tin, causing it to explode into a plasma that emits extreme ultraviolet light. This laser must strike the molten tin with a level of precision "more precise than the calculations to send the Apollo missions to the moon"—and it must successfully repeat this explosion 50,000 times per second.

- **The Mirror Problem:** EUV light has such a short wavelength that it is absorbed by all standard mirrors. ASML had to invent entirely new types of mirrors to bounce the light roughly 20 times before it finally etches the silicon. - **Supply Chain Monopolies:** ASML relies on a contract with a single German company that is the only entity on Earth capable of manufacturing the specific lasers required for these machines. - **Production Scarcity:** ASML only produces about 50 of these advanced EUV machines per year, resulting in a multi-year backlog. Former competitors like Nikon completely gave up trying to compete because the engineering was too difficult.

[01:28:41](https://www.youtube.com/watch?v=tuB3_fJyC5Q&t=5321s) TSMC's Economic Moat and Morris Chang's Return

The sheer difficulty of this manufacturing paradigm gives TSMC a virtually unbreakable economic moat, allowing them to command astronomical 40% operating margins. The speakers joke that even if their own podcast, *Acquired*, wanted to leverage its brand power and invest a couple hundred million dollars to buy an ASML machine and compete, they wouldn't have a clue how to operate it despite having computer science degrees. It requires a workforce that has spent decades executing the most advanced manufacturing in human history just to figure out how to build the *next* iteration.

In 2005, thinking his life's work was complete, a 74-year-old Morris Chang stepped down as CEO of TSMC to enjoy a retirement of reading literature and spending time with his supportive second wife. He handed the reins to his longtime lieutenant, Rick Sai. However, following the chaos of the 2008 global financial crisis and rumblings that EUV technology might actually transition from a "science project" into a viable commercial reality, TSMC held a surprise press conference in the summer of 2009. They announced that the 78-year-old Morris Chang was returning as CEO, pushing Rick Sai out in a move that outsiders initially viewed as a botched executive transition.

- **Rick Sai's Second Act:** After leaving the CEO spot at TSMC, Sai went on to become the CEO of MediaTek, a successful chip-design company that originally spun out of TSMC. - **Fighting Spirit:** Upon his return, Chang famously issued a press release stating the leadership shakeup would spur "greater intensity" and that he saw "golden opportunities ahead."

[01:33:04](https://www.youtube.com/watch?v=tuB3_fJyC5Q&t=5584s) The Mobile and Cloud Tipping Points

The "golden opportunities" Chang foresaw in 2009 were driven by two massive, simultaneous computing shifts: Mobile and Cloud. On the mobile front, the July 2007 launch of the iPhone and the subsequent 2008 release of the App Store and Android completely disrupted the silicon landscape. Up to this point, computing was dominated by the PC paradigm—Intel's x86 processors plugged into a wall. Mobile required low-power ARM architecture. Early iPhones actually relied on off-the-shelf Samsung processors, which Samsung also fabricated. However, the explosion of fabless mobile chip designers like Qualcomm, Broadcom, MediaTek, and eventually Apple itself, played perfectly into TSMC's pure-play foundry model.

Simultaneously, the enterprise world shifted to the Cloud. Initially, the cloud was a massive boon for Intel because platforms like Amazon Web Services bought staggering amounts of x86 server CPUs. But as computing workloads progressed into machine learning and artificial intelligence, the traditional CPU became less important than processors capable of massive parallel processing and matrix multiplication. This propelled NVIDIA—which was a modest gaming GPU company in 2005 whose stock merely fluctuated based on whether they won Xbox or PlayStation console contracts—into a tech titan worth over $3 trillion, more than double the market cap of Intel.

- **Market Miscalculation:** In 2007, Steve Jobs displayed a slide stating Apple's modest goal was to capture just 1% of the existing smartphone market. Neither Apple nor Google anticipated how completely smartphones would swallow global computing by 2009. - **The Custom Silicon Era:** The rise of foundries lowered the barrier to entry, enabling modern tech giants to design entirely custom chips tailored for specific workloads, such as Tesla's Dojo chips. - **Manufacturing Nuance:** While many assume TSMC handles everything, Tesla has historically used Samsung to fabricate components of its custom silicon, even utilizing a Samsung Fab located in Austin, Texas. However, the speakers note that for next-generation 3-nanometer workloads, companies will ultimately have to rely on TSMC.

[01:38:28](https://www.youtube.com/watch?v=tuB3_fJyC5Q&t=5908s) The Handshake That Built the iPhone

The historic partnership between Apple and TSMC originally materialized through casual dinners and living room conversations rather than stuffy boardrooms. Apple had recently acquired PA Semi and begun designing its own proprietary chip architecture in-house, leading to the creation of the A4 chip. Driven by the conviction that the iPhone was going to explode in popularity, Apple needed a manufacturing partner to help them compete head-to-head with Samsung. Steve Jobs famously detested Samsung, viewing them as a copycat lacking clear strategic focus due to their dual roles as a consumer brand with the Galaxy line and a semiconductor foundry.

This high-stakes collaboration was a monumental gamble for both parties. Apple was tying its entire future to a foundry that was widely viewed as an "also-ran" at the time. Jeff Williams later remarked that betting heavily on TSMC meant operating with "no backup plan." For TSMC, the deal demanded an astronomical initial investment of \$9 billion. They had to dedicate 6,000 employees just to build a specialized plant for Apple in a blistering 11-month window, even though it took several years before chip production actually commenced.

- **Steve Jobs' Animous:** Jobs famously promised to wage "thermonuclear war" against Google and held deep resentment toward Samsung for allegedly stealing Apple's physical designs. - **Toxic Hell Stew:** Tim Cook would later use this phrase on stage to describe the competitive landscape and frustrations surrounding their rivals. - **The iPhone 6 Pivot:** While earlier devices like the A4 and A5 models split fabrication duties between Samsung and TSMC, the massively popular iPhone 6 became the first generation to be fabbed entirely by TSMC.

[01:41:17](https://www.youtube.com/watch?v=tuB3_fJyC5Q&t=6077s) TSMC's Monopoly and Massive Capital Expenditures

Following the monumental execution of the Apple deal, Morris Chang stepped down as CEO in 2013 but retained his position as chairman. He finally entered true retirement in June 2018 at the age of 86, leaving TSMC at the absolute peak of the semiconductor industry. By 2020, the company's financial performance was staggering, posting a \$20 billion operating profit on \$48 billion in revenue. True to form, they plowed \$17 billion of that profit straight back into capital expenditures (capex) to widen their competitive moat.

In early 2021, TSMC sent a shockwave through the financial markets by aggressively raising its capex guidance. In January, they projected a bump to between \$25 billion and \$28 billion, only to raise it again in April to a forecasted \$30 billion for the year—part of a mind-boggling \$100 billion spending plan over three years. This aggressive capital deployment effectively cornered the market, making it nearly impossible for any competitor, including Samsung, to keep up with their manufacturing capabilities.

- **Unprecedented Pricing Power:** Breaking away from Chang's second classic innovation of steady price reductions, TSMC announced a 20% price hike in 2021, illustrating their total dominance over the supply chain. - **Market Dominance:** TSMC commands over 50% of the global market share for all contract chip manufacturing and captures an estimated 95% or more of the industry's total profits. - **Leading-Edge Dominance:** In the most advanced chip segments, such as the 5-nanometer process node, TSMC commands a staggering 90% plus market share, with Samsung holding a mere 5% to 10%.

[01:45:01](https://www.youtube.com/watch?v=tuB3_fJyC5Q&t=6301s) The Unyielding Demand for the Leading Edge

TSMC's financial trajectory since its initial public offering in Taiwan in 1994 highlights an astonishing run of consistent growth. Over 27 years, the company maintained a compound annual revenue growth rate of 17.4%. Even more impressive is its valuation growth: expanding from a \$4 billion market cap at its IPO to a massive \$550 billion valuation by 2021, yielding an internal rate of return (IRR) of 19.9%. At the time of recording, TSMC stands as the ninth most valuable company on Earth, and the only one in the global top ten—aside from Saudi Aramco—that has not been previously covered on the show.

The hosts push back against bearish industry arguments suggesting that advanced processing power will eventually outpace software needs, rendering the "leading edge" obsolete. They argue that software developers invariably write code that consumes whatever computational hardware is available, meaning companies like Apple, Tesla, Google, and Amazon will always aggressively demand the fastest, smallest chips available.

- **Accelerating Growth:** TSMC grew its year-over-year revenue by 31% from 2019 to 2020, doubling its historical compound annual growth rate during that window. - **Organic Financing:** With \$28 billion in cash and cash equivalents sitting on its balance sheet, TSMC is uniquely positioned to fund its massive \$100 billion capex layout organically through operational cash flows. - **The 5-Nanometer Reality:** The term "5-nanometer" has transitioned into a marketing designation rather than a literal measurement of transistor gate length. At this scale, silicon atoms are only about five to a nanometer, meaning physical limits will soon prevent further subdivision of the substrate.

[01:50:43](https://www.youtube.com/watch?v=tuB3_fJyC5Q&t=6643s) Moore's Second Law and Seven Powers Framework

While Moore's Law regarding transistor density is widely recognized, the hosts introduce "Moore's Second Law"—also known as Rock's Law, named after legendary venture capitalist Arthur Rock. The law posits that the capital cost of a semiconductor fabrication plant doubles roughly every four years. When you synthesize Moore's Law with Rock's Law, the financial math dictates an inevitable winner-take-all monopoly because only the single most profitable player can afford the upfront capital expenditures required to achieve extreme manufacturing scale.

This dynamic serves as a textbook historical template for operating leverage. While modern venture capital and internet startups leverage software for 80% to 90% gross margins with virtually zero distribution costs, the venture capital asset class itself was originally engineered to finance the heavy fixed costs of semiconductor foundries. To analyze the depth of TSMC's structural defensibility, the hosts pivot to Hamilton Helmer’s *Seven Powers* framework, exploring the long-term differential profit margins that separate the titan from its peers.

- **Operating Leverage Reimagined:** Unlike software scale, TSMC relies on pouring capital into massive physical infrastructure, resulting in a defensive 50% gross margin moat that software companies struggle to replicate. - **Counter-Positioning Beginnings:** At its inception, TSMC utilized counter-positioning against established integrated device manufacturers. Incumbents couldn't copy TSMC's pure-play foundry model because their corporate profit centers relied heavily on bundling design and manufacturing together.

[01:54:27](https://www.youtube.com/watch?v=tuB3_fJyC5Q&t=6867s) Counter-Positioning and the Birth of Pure-Play Foundries

The discussion opens with a look back at the historical "counter-positioning" that allowed TSMC to thrive. Before TSMC pioneered the pure-play foundry model—focusing solely on manufacturing chips for other companies without designing any of their own—fabulous companies were incredibly rare. If a firm needed manufacturing capacity, they had to lease excess capacity from an Integrated Device Manufacturer (IDM). However, these IDMs leveraged their position ruthlessly: even if you had a great strategic relationship, they would strong-arm you and demand the right to market your products under their own brand.

TSMC offered an entirely alternative paradigm by promising never to compete with its customers. This strategy was a textbook case of counter-positioning because the established IDMs were fundamentally unwilling to match TSMC's neutral stance. Alongside this business model innovation, scale economies emerged as one of the top two pillars of TSMC's dominance, alongside process power.

- **Pre-TSMC Capacity Constraints:** IDMs renting out excess capacity routinely demanded co-branding rights to their clients' technology. - **Top Structural Powers:** The speakers identify scale economies and process power as TSMC's primary competitive advantages.

[01:55:25](https://www.youtube.com/watch?v=tuB3_fJyC5Q&t=6925s) The Trap of High Switching Costs and the Ecosystem Moat

Switching off of TSMC has become virtually impossible for any company that wants to remain on the "Leading Edge" of technology. The speakers debate the sheer scale of these switching costs, joking that listeners might think they have gone deep technically on this episode, but they "haven't even scratched the surface." If a company wanted to switch production to a competitor like GlobalFoundries, they couldn't just call them up and say, "hey, I'm porting over, expect my business on Monday." The integration is so deep that a transition takes years of collaborative engineering.

While traditional "network economies" like those seen in social media platforms do not apply here, TSMC possesses a powerful ecosystem moat. Electronic Design Automation (EDA) companies and intellectual property (IP) providers like ARM are so deeply integrated into TSMC's manufacturing process that it creates a collective ecosystem. TSMC itself refers to this corporate framework as something akin to "open innovation." Interestingly, this massive concentration of demand creates friction; the speakers ponder whether Apple's dominance as a primary customer actually hurts other clients, given that Apple routinely snaps up the lion's share of advanced capacity, such as the 5-nanometer process.

- **Transition Timeline:** Porting a chip design from TSMC to a competitor requires years of deep process integration. - **The Apple Effect:** Apple's massive purchasing power frequently bottlenecks access to advanced nodes (like 5nm) for smaller tech companies.

[01:57:13](https://www.youtube.com/watch?v=tuB3_fJyC5Q&t=7033s) Unassailable Process Power and the Threat of Paradigm Shifts

The speakers point to TSMC as the absolute clearest example of "process power" they have ever encountered. It has taken all 40 years of TSMC's history, accumulation of know-how, and proprietary IP to arrive at its current execution level. The organizational knowledge is so dense that if ten key engineers left to start a rival firm, duplicating what TSMC has built from scratch would be virtually impossible. No amount of raw capital can replicate their synchronized relationships with crucial supply chain partners like ASML.

The only existential threat that could unseat TSMC is a massive, fundamental computing paradigm shift—comparable to the disruption mobile inflicted on desktop computing. If a future computing architecture emerges that bypasses everything TSMC is optimized for—such as a radically cheaper form of quantum computing or an alternative substrate that completely replaces silicon—the playing field might reset. Short of an "AWS moment" that makes this hyper-advanced tech orders of magnitude cheaper and accessible to anyone, TSMC is firmly in the groove, making it "TSMC's game to lose." Because branding is antithetical to their invisible backend model, the hosts agree to skip discussing traditional branding power altogether.

- **Replication Barriers:** TSMC's 40-year accumulation of institutional knowledge and supply chain synchronization cannot be bought with money alone. - **Disruption Criteria:** Only a complete architecture reset (e.g., scalable quantum computing or non-silicon substrates) would threaten their market position.

[01:59:08](https://www.youtube.com/watch?v=tuB3_fJyC5Q&t=7148s) The Geopolitical Bear Case and Geographic Diversification

The conversation shifts naturally into a serious analysis of geopolitics, which represents the ultimate bear case for TSMC. Absent a technical paradigm shift, the most glaring risk to the business is the possibility of China asserting force to take over or annex Taiwan. In an American context, the speakers note that if a takeover happened and somehow avoided triggering a full-scale World War, China would likely divert TSMC's staggering manufacturing capacity to domestic giants like Huawei, completely cutting off the West.

When analyzing where this structural vulnerability fits within business strategy frameworks, it is classified as a distinct "weakness" or "anti-power." TSMC's geographic concentration is historically tied to the early support of the Taiwanese government, which initially "strong-armed" local business leaders to invest and ensure the foundry's success. Given this acute geopolitical risk, it is surprising how slowly TSMC has diversified its geographic footprint.

While TSMC is currently spending $12 billion to build a plant in Arizona, the hosts note that this factory will not feature their most advanced 3-nanometer or 5-nanometer manufacturing. This move, heavily subsidized by the Biden administration's semiconductor legislation, serves as a strategic first step to restore foundational manufacturing familiarity within the US. Additionally, TSMC is expanding with a new Fab in Japan and maintains some operations in China, though founder Morris Chang has openly commented that it makes little business sense to put the Leading Edge in these countries when the ideal ecosystem remains rooted in Taiwan.

- **Arizona Expansion Budget:** TSMC is deploying $12 billion for an onshore US facility, though it lags behind the company's most advanced chip generations. - **Global Footprint:** Diversification efforts include new initiatives in Japan and existing baseline operations in China, heavily incentivized by international subsidies.

[02:02:57](https://www.youtube.com/watch?v=tuB3_fJyC5Q&t=7377s) The Semiconductor Supply Chain vs. National Defense

Contemplating a future where Taiwan is annexed is a chilling exercise. While some might dismiss a corporate takeover as a minor detail in a geopolitical crisis, the speakers argue that losing access to leading-edge semiconductors is "everything." Virtually all technological progress would halt instantly; automotive lines would freeze, and manufacturers wouldn't even be able to build basic consumer vehicles like the Ford F-150.

The hosts pose a fascinating thought experiment: could TSMC's process power be physically airlifted out of Taiwan in an emergency evacuation, or does the power inherently stay with the physical geography and infrastructure? Looking at historical precedents like the NUMI plant—a joint venture where Toyota tried and largely failed to replicate its legendary production system with GM in Fremont—suggests that process power does not transplant easily.

This raises a massive question regarding national defense priorities. For decades, companies like Boeing have been propped up as vital, protected assets essential to the American way of life. Yet, because modern aerospace depends entirely on advanced silicon, having cutting-edge semiconductor capabilities onshore might actually be more critical to national defense than aircraft manufacturing itself. A similarly tense dynamic exists in South Korea with Samsung, positioned just miles away from an unpredictable North Korean border.

- **The NUMI Precedent:** Toyota's historical difficulty replicating its production system with GM illustrates the friction of moving deeply ingrained manufacturing cultures. - **Defense Interdependence:** Advanced defense supply chains, including aerospace giants like Boeing, are entirely bottlenecked by foreign semiconductor availability.

[02:06:11](https://www.youtube.com/watch?v=tuB3_fJyC5Q&t=7571s) The Greatest Strategic Error of All Time: Intel's iPhone Blunder

Moving toward their "Playbook" segment, the hosts revisit a historic "what would have happened otherwise" scenario. They read a direct quote from Bloomberg detailing a massive turning point in the mid-2000s: as Apple was secretly developing the original iPhone, Steve Jobs approached Intel's then-CEO, Paul Otellini, about manufacturing the phone's chips. Intel was already providing the processors for the Mac line, and Otellini was the very partner Jobs had triumphantly brought on stage to signal the death of the PowerPC architecture.

However, Intel viewed Apple's financial terms as a "lowball offer" and turned it down. Apple subsequently awarded the contract to Samsung, eventually began designing its own silicon internally, and ultimately outsourced its massive production runs to TSMC. Reflecting on this, the speaker visibly reacts with fists in the air, calling it the "biggest strategic error of all time." It wasn't an issue of Intel refusing to use ARM architecture; it was a choice driven entirely by short-term pricing structures.

This historical blunder gives birth to a new strategic maxim dubbed the "Rosenthal Postulate," which sits alongside their established Rosenthal Doctrine ("never bet against the internet"). The postulate dictates: **never make strategic decisions based on pricing rather than long-term economics.** Humans are profoundly prone to "cutting off their noses to spite their faces," walking away from paradigm-shifting partnerships because the initial deal structure or valuation feels slightly unfavorable.

- **The Rejected Contract:** Intel passed on manufacturing the original iPhone chip because it deemed Apple's initial financial proposal too low. - **The Long-Term Pivot:** This single rejection forced Apple to build an in-house chip design team, fueling the meteoric rise of TSMC's foundry business.

[02:09:01](https://www.youtube.com/watch?v=tuB3_fJyC5Q&t=7741s) The Slow Fade of Legacy Giants and ASML's Ascent

The final playbook theme explores how long it takes for a massive tech incumbent to truly become irrelevant. Despite Intel's strategic missteps, its fade is a slow-motion process. However, the public markets reflect a dramatic shift in power dynamics: TSMC's market cap has grown to roughly two and a half times that of Intel.

Even more striking is the ascent of ASML, the Dutch lithography manufacturer. ASML has surpassed Intel in market cap to become Europe's most valuable company. They achieved this milestone by positioning themselves as the sole-source provider of a single, irreplaceable link in the global technology value chain, selling primarily to one dominant player: TSMC. The hosts close this portion of the episode by inviting public market investors from their community to chime in on Slack, Twitter, or email regarding how they balance their investment theses between TSMC and ASML.

- **Market Cap Disparity:** TSMC's valuation sits at roughly $2.5\times$ that of legacy giant Intel. - **ASML's Corporate Stature:** By monopolizing extreme ultraviolet (EUV) lithography systems, ASML has grown larger than Intel by market capitalization.

[02:09:54](https://www.youtube.com/watch?v=tuB3_fJyC5Q&t=7794s) Intel's Inertia and the Innovator's Dilemma

The conversation transitions to speculative up-and-comers in the semiconductor space, prompting a discussion on whether it makes sense to short Intel. One speaker cautions against counted-out incumbents, noting that "it takes a long time to become irrelevant." Despite widespread criticism, Intel still controls 80% of the computer processor market and holds an even larger share in servers. The vast majority of cloud computing workloads do not run on highly specialized tech platforms like Tesla Dojo; instead, they consist of non-tech companies simply running Outlook servers on Office 365, which have no practical need for a cutting-edge 5-nanometer manufacturing process.

However, strategic indecision has severely damaged Intel over the years. Former CEO Bob Swan originally prepared to throw in the towel and outsource the manufacturing of Intel-designed chips to TSMC. In a complete reversal of that strategy, current CEO Pat Gelsinger—the former VMware chief executive who started his career at Intel and was once a top candidate to replace Steve Ballmer at Microsoft—aims to transform Intel into a standalone foundry business for fabless companies. The hosts note that if Intel runs this as a fully autonomous division, it mirrors AMD's successful but imperfect spin-out of GlobalFoundries, raising the question of why Intel should even continue to own it.

Intel's history also features a perfect textbook example of the innovator's dilemma, drawing a direct parallel to Kodak inventing the digital camera but failing to commercialize it. Intel was actually the earliest and largest investor in extreme ultraviolet (EUV) lithography, committing over \$4 billion to the technology in 2012. However, according to *The Wall Street Journal*, Intel grew skeptical that it would work and calculated it was safer to optimize existing lithography methods instead. Because they hesitated while rivals adopted EUV, Intel fell behind, proving why the industry fundamentally relies on nimble startups to commercialize breakthroughs.

* Intel maintains an 80% market share in computer processors and an even larger share in server CPUs. * Former CEO Bob Swan planned to outsource manufacturing to TSMC, while current CEO Pat Gelsinger reversed course to build an internal foundry division. * Intel invested \$4 billion into EUV lithography in 2012 but failed to adopt it in time, resulting in zero shipping Intel-manufactured chips using EUV to this day.

---

[02:14:12](https://www.youtube.com/watch?v=tuB3_fJyC5Q&t=8052s) Breaking the Rules of Startup Strategy

The hosts reflect on the abstract nature of analyzing business outcomes, arguing that you cannot reverse-engineer the true probability of success just by looking at a single positive result. TSMC founder Morris Chang successfully deployed an "if you build it, they will come" strategy by heavily investing in developing a market before it existed. This flies in the face of classical venture capital logic. For instance, legendary investor Don Valentine's first rule of investing was to back companies only when a market already exists, never to fund the development of a brand-new market.

This tension is commonly observed among founders spinning out of major tech corporations like Microsoft. Because they come from a platform ecosystem, these founders naturally want to build new platforms, often assuming demand will materialize naturally while maintaining too narrow an understanding of actual day-one customers. While Morris Chang's approach broke these conventional rules, the hosts note that even a strategy with a 90% failure rate will still succeed one out of ten times. Startups defy rigid formulas, making it difficult to distinguish an inherently flawless strategy from a risky one that simply hit a probabilistic jackpot.

Another major playbook theme centers on Jeff Bezos's famous framework for Amazon Web Services (AWS), which advises startups to outsource anything that "doesn't make your beer taste better." Derived from an old analogy about German breweries outsourcing their electricity generation to focus purely on brewing, the rule dictates that non-core competencies like finance and accounting should always be managed externally. The hidden genius of this rule is its inverse: whenever an entrepreneur spots a non-core task that thousands of companies are wastefully doing themselves, that represents a massive opportunity to build a multi-billion dollar platform company.

---

[02:16:30](https://www.youtube.com/watch?v=tuB3_fJyC5Q&t=8190s) Grading TSMC: The Ultimate B2B Moat

Rather than issuing a standard letter grade for Taiwan's historical economic decisions, the hosts evaluate TSMC's legacy through a thought experiment: Where does the company belong in the pantheon of all-time great technology giants? The scale of computing enabled by TSMC is roughly 100 times larger than Intel's peak in the year 2000. Every major technological frontier—whether machine learning, cryptocurrency, 5G, or Apple Silicon—is entirely dependent on TSMC's ability to drive the asymptotic cost of compute down toward zero. "When Mark Andreessen says software is eating the world," one host notes, "it's only eating the world because TSMC has made it so freaking cheap to manufacture silicon."

The hosts debate whether TSMC sits above or below the "FAANG" tier of consumer tech giants like Facebook, Amazon, Apple, Microsoft, and Google. While some argue that those platforms couldn't innovate without TSMC, others suggest that a foundry-and-fabless hardware model may have been structurally inevitable even if TSMC hadn't built it. Stripping away the obvious geopolitical risk of being located in Taiwan, TSMC boasts what is arguably the most impenetrable corporate moat in human history. Traditional tech monopolies rise and fall on shifting consumer trends, but TSMC's astronomical capital requirements and compounding efficiency flywheel make its market position virtually bulletproof.

Ultimately, the hosts place TSMC firmly in the top ten global tech companies of all time, though just outside the top five. Its lower consumer profile is entirely intentional; TSMC purposefully operates under the radar with no consumer branding, choosing instead to let its customers own the spotlight. Structurally, the closest historical comparison to TSMC's compounding market dominance is Standard Oil, which similarly crowded out rivals until it was impossible to compete. The hosts conclude that TSMC is undisputedly the most successful B2B hardware company in history, competing only with Microsoft for the title of the greatest business-to-business entity across any industry.

---

[02:23:11](https://www.youtube.com/watch?v=tuB3_fJyC5Q&t=8591s) Carve-outs and Closing Recommendations

Bringing the episode to a close, the hosts transition into their traditional "carve-outs" segment to share recent personal recommendations. The first host shares a pair of television recommendations discovered during a summer trip to Santa Barbara to escape the chilly San Francisco weather. Having admittedly lived "under a rock" regarding modern television, he recommends the first season of *Ted Lasso*, validating the hype surrounding its high quality despite hearing mixed reviews about the second season. On a more nostalgic note, he also recommends *Greek*, a mid-2000s comedy-drama series tracking fraternity and sorority life at a fictional university, which serves as a heartwarming period piece for anyone who attended college during that era.

The second host highlights a captivating business biography that had been recommended to him for years: *Who is Michael Ovitz?* For fans of legendary business memoirs like Nike's *Shoe Dog*, Disney's *The Ride of a Lifetime*, or the Ford-focused *An American Icon*, this book is a mandatory addition to the reading list. From roughly 1975 to 2000, Ovitz completely upended the traditional Hollywood hierarchy by founding the Creative Artists Agency (CAA). Operating with extreme secrecy and shielding his strategies from the press, Ovitz orchestrated the packages behind the era's biggest film and television hits. The book offers a crucial, multidimensional Counterpoint to the historically one-sided, negative narratives surrounding Ovitz's famously brief and tumultuous executive tenure at Disney.

[02:25:54](https://www.youtube.com/watch?v=tuB3_fJyC5Q&t=8754s) Recommendation of "Who Is Michael Ovitz?" and Hollywood Classics

The conversation winds down with a passionate book recommendation that ties directly back to themes explored in recent episodes. Reflecting on their previous coverage of Bob Iger and the Disney empire—particularly how those narratives connected back to the Creative Artists Agency (CAA)—the speakers highlight the autobiography *Who Is Michael Ovitz?* as an essential read. They describe it as an "awesome read" that will deeply resonate with anyone who enjoys classic CEO business stories, movie history, or the inner workings of the entertainment industry.

The hosts share a moment of shared nostalgia for the media they grew up on, noting how impactful these projects were during their childhoods. They reminisce about the magic of both children's films and the adult blockbusters they desperately wanted to sneak into when they were too young. It brings a unique sense of wonder to look back and hear the exact mechanics of how absolute classics like *Goodfellas* and *Jurassic Park* were packaged and brought to life behind the scenes by Hollywood's ultimate power brokers.

- *Who Is Michael Ovitz?* is highly recommended for fans of Disney, Bob Iger, movie history, and classic executive business stories. - The book provides behind-the-scenes insight into how massive cultural touchstones like *Goodfellas* and *Jurassic Park* were packaged.

[02:26:33](https://www.youtube.com/watch?v=tuB3_fJyC5Q&t=8793s) The Reality of Podcast Growth and Final Sign-Off

As the episode approaches its final moments, the hosts realize they forgot a crucial housekeeping note from the introduction and invite their audience to join their community workspace at `acquired.fm/slack`. They joke that the channel is a place to hang out with "talented, smart, good-looking people like yourselves." Amidst the banter, they offer a transparent look at the mechanics of the medium, noting that "podcasting is this weird thing" because it entirely lacks a built-in viral loop.

While shouting out the show from "social media hilltops" is always appreciated, the hosts emphasize that true growth relies entirely on old-fashioned word of mouth. They encourage listeners to share specific moments they loved, or even points they actively disagreed with, to keep the conversation going. With a warm farewell, they promise to see everyone next time, closing out the episode to the familiar strains of their theme music asking, "Who got the truth now, huh?"

- Listeners can join the community platform at `acquired.fm/slack`. - Growth in podcasting depends heavily on word of mouth rather than automated viral loops.

## Notable quotes > [00:00](https://www.youtube.com/watch?v=tuB3_fJyC5Q&t=0s) essentially the only manufacturer of the Leading Edge chips — David Rosenthal

> [07:40](https://www.youtube.com/watch?v=tuB3_fJyC5Q&t=460s) Chinese laundrymen, Chinese restaurant owners, Chinese engineers, and Chinese professors — Morris Chang

> [17:44](https://www.youtube.com/watch?v=tuB3_fJyC5Q&t=1064s) We at Sylvania cannot make what we can sell, and we cannot sell what we can make

> [17:44](https://www.youtube.com/watch?v=tuB3_fJyC5Q&t=1064s) reduce the cost of electronic functions by a factor of a million to one — Jack Kilby

> [36:17](https://www.youtube.com/watch?v=tuB3_fJyC5Q&t=2177s) put out to pasture — Morris Chang

> [54:07](https://www.youtube.com/watch?v=tuB3_fJyC5Q&t=3247s) it was like in the movie The Godfather, it was an offer I couldn't refuse — Morris Chang

> [56:10](https://www.youtube.com/watch?v=tuB3_fJyC5Q&t=3370s) real men have Fabs — Jerry Sanders

> [01:13:55](https://www.youtube.com/watch?v=tuB3_fJyC5Q&t=4435s) The semiconductor business is like a treadmill that speeds up all the time. If you can't keep up, you fall off. — Morris Chang

> [01:38:28](https://www.youtube.com/watch?v=tuB3_fJyC5Q&t=5908s) no backup plan — Jeff Williams

> [02:06:11](https://www.youtube.com/watch?v=tuB3_fJyC5Q&t=7571s) biggest strategic error of all time

> [02:06:11](https://www.youtube.com/watch?v=tuB3_fJyC5Q&t=7571s) never make strategic decisions based on pricing rather than long-term economics

> [02:11:54](https://www.youtube.com/watch?v=tuB3_fJyC5Q&t=7914s) doesn't make your beer taste better

> [02:14:30](https://www.youtube.com/watch?v=tuB3_fJyC5Q&t=8070s) When Mark Andreessen says software is eating the world, it's only eating the world because TSMC has made it so freaking cheap to manufacture silicon.

## Mentioned TSMC, Nvidia, Apple, Broadcom, Qualcomm, AMD, Intel, MacBook, iPhone, Amazon Web Services, Saudi Aramco, Morris Chang, ARM, Sequoia Part 1, P.A. Semi, Warren Buffett, Ford Motor Company, Sylvania, Donald T. Valentine, Sequoia Capital, William Shockley, Raytheon, Fairchild, Texas Instruments, Jack Kilby, Bob Noyce, IBM, IBM 7090, Pat Hagerty, Stanford, Boston Consulting Group, National Semiconductor, Vanta, Atlassian, Quora, Factory, Mark Shepard, Motorola, LJ Sevin, Mostek, Sevin Rosen Ventures, Microsoft, General Instrument, Hamilton Helmer, Industrial Technology Research Institute, KT Lee, RCA, United Microelectronics Corporation, MediaTek, Foxconn, Jerry Sanders, GlobalFoundries, Acorn, VLSI Logic, Synopsys, Cadence, Lam Research, Applied Materials, Tokyo Electron, ASML, ASMI, JP Morgan Payments, Digital Equipment Corporation, DEC, StrongARM, XScale, Marvell, Samsung, SMIC, Nikon, Rick Sai, Google, Tesla, Steve Jobs, Jeff Williams, Tim Cook, Arthur Rock, Huawei, Biden, China, Taiwan, Japan, Arizona, Boeing, South Korea, North Korea, NUMI, Toyota, GM, Fremont, Ford F-150, Paul Otellini, Mac, PowerPC, Bloomberg, Rosenthal Postulate, Rosenthal Doctrine, Bob Swan, Pat Gelsinger, VMware, Steve Ballmer, Office 365, Outlook, Tesla Dojo, Kodak, The Wall Street Journal, Don Valentine, Jeff Bezos, Mark Andreessen, FAANG, Facebook, Amazon, Standard Oil, Ted Lasso, Greek, Who is Michael Ovitz?, Michael Ovitz, Shoe Dog, Nike, The Ride of a Lifetime, Disney, Bob Iger, An American Icon, Ford, Creative Artists Agency, CAA, Goodfellas, Jurassic Park, Slack, Twitter, Ben Gilbert, David Rosenthal

[Source](https://www.youtube.com/watch?v=tuB3_fJyC5Q)

中文简报

英特尔制造受挫，台积电确立全球芯片垄断地位台积电作为横向整合的垄断者主导着全球科技生态系统，因其激进的资本规模与工艺实力推动计算的渐进成本趋近于零。本期内容按时间顺序回顾了台积电（TSMC）及其创始人张忠谋（Morris Chang）的发展历程，从其早年在中国的生活、在德州仪器（Texas Instruments）的职业生涯，到他在台湾开创性的“纯晶圆代工”模式。叙述探讨了台积电如何通过重金投资极紫外（EUV）光刻技术，并锁定与苹果（Apple）、英伟达（Nvidia）等科技巨头的关键合作关系，从而建立起横向整合的垄断地位。通过承担制造业的资本风险，台积电依靠规模经济和工艺实力构建了不可逾越的经济护城河，切实成为现代全球计算基础设施的核心支柱。台积电是现代计算的基础支柱，作为先进制程芯片的独家制造商，服务于英伟达、苹果、博通、高通、AMD和英特尔等主要科技巨头。 (0:00) 创始人张忠谋于1987年以56岁之龄创立台积电，凭借其在德州仪器深厚的行业背景，确立了公司的长期运营架构。 (2:08) 张忠谋开创了“纯晶圆代工”商业模式，明确公司仅严格作为合同制造商运作，绝不通过设计自有芯片开展直接竞争。 (56:10) 台积电依托与荷兰设备制造商ASML至关重要的垄断性合作，后者提供先进芯片生产所需的高度复杂的极紫外（EUV）光刻机。 (67:23) 半导体制造格局经历了大规模整合，从2000年代初拥有20多家具备先进制程竞争力的企业，演变为如今台积电近乎完全主导的局面。 (79:20)

Reading summary

In their deep dive on TSMC, Acquired hosts Ben Gilbert and David Rosenthal argue that Taiwan Semiconductor Manufacturing Company built the world's most critical manufacturing monopoly by pioneering the pure-play foundry model and compounding unmatched process power. However, they emphasize that this foundational pillar of global computing remains structurally vulnerable due to its extreme geographic concentration in Taiwan. Inventing the Pure-Play Foundry Model (56:10) When Morris Chang founded TSMC in 1987 with backing from the Taiwanese government and Philips, semiconductor manufacturing was dominated by integrated device manufacturers (IDMs) like Intel and Texas Instruments that designed and fabricated chips internally. The prevailing industry consensus, captured by AMD CEO Jerry Sanders's quip that 'real men have fabs,' held that chip fabrication was an inseparable, proprietary craft. Startup chip designers faced massive capital requirements to build fabrication plants or had to rely on rival IDMs that often demanded co-branding rights and offloaded manufacturing losses onto external clients. TSMC countered this paradigm by promising never to design chips or compete with its customers. The hosts describe Chang's initiative as an index bet on the entire semiconductor sector, effectively transforming TSMC into the 'AWS of chip companies.' By removing the requirement for startups to finance physical foundries, TSMC enabled the rise of the fabless chip design industry—creating anchor demand for customers like Qualcomm, Broadcom, and Nvidia while securing massive, long-term manufacturing volume for itself. Capital Moats and Compounding Process Power (79:20) The hosts explain TSMC's dominance through the interaction of Moore's Law and 'Moore's Second Law' (Rock's Law), which posits that the cost of building a chip fabrication facility doubles roughly every four years. This dynamic triggered severe industry consolidation: while 22 companies competed at the leading manufacturing edge in the early 2000s, that figure dwindled to six by the mid-2010s, and eventually narrowed to just TSMC and Samsung at the 5-nanometer threshold. TSMC sustained this position by reinvesting extraordinary amounts of operating profit directly into capital expenditures—allocating $17 billion of its $20 billion profit in 2020 and initiating a three-year, $100 billion spending blueprint in 2021. Beyond raw capital deployment, Gilbert and Rosenthal argue that TSMC's most defensible asset is its accumulated 'process power.' Operating advanced extreme ultraviolet (EUV) lithography machinery—monopolized by Dutch equipment maker ASML—represents some of the most intricate engineering in history. The institutional knowledge required to run these tools at high manufacturing yields took four decades to develop and cannot be easily purchased or replicated by competitors, allowing TSMC to achieve 40% operating margins and rare supply chain pricing power. Mobile Inflections and Legacy Misses (98:28) TSMC's modern scale accelerated sharply following Morris Chang's return as CEO in 2009 at age 78 to capitalize on shifts toward mobile and cloud architectures. A decisive catalyst was Intel's refusal to manufacture chips for the original iPhone because Intel viewed Apple's financial terms as a lowball offer. Gilbert and Rosenthal characterize this rejection as one of the biggest strategic errors in corporate history, formulated around short-term pricing margins rather than long-term platform economics. After initially relying on Samsung, Apple struck a monumental manufacturing partnership with TSMC to produce proprietary Apple Silicon. TSMC committed $9 billion in upfront capital and 6,000 dedicated personnel to erect a dedicated facility in an 11-month timeframe. Securing Apple's exclusive leading-edge business guaranteed the massive volumes TSMC required to drive down costs along the manufacturing learning curve, permanently outpacing legacy competitors who remained tied to declining PC architectures. Geopolitics as an Existential Anti-Power (119:08) Despite its competitive strengths, the hosts identify TSMC's acute geographic concentration on Taiwan as its primary structural weakness, or 'anti-power.' With China claiming sovereignty over Taiwan, any military conflict or blockade presents an immediate existential risk to the world economy. The hosts highlight that losing access to TSMC's leading-edge manufacturing would halt technological progress globally, freezing production across commercial computing, consumer electronics, and automotive manufacturing lines. Mitigating this vulnerability is exceptionally difficult because process power resists physical relocation. TSMC's subsidized $12 billion facility in Arizona represents a step toward geographic diversification, but it lacks the capability to produce the company's most advanced nodes. The hosts suggest that cutting-edge domestic semiconductor fabrication has effectively surpassed traditional defense hardware like aerospace manufacturing as the single most critical asset for national security. The hosts conclude that TSMC has established an almost unassailable enterprise moat that drives down the marginal cost of computing worldwide. However, because its non-replicable manufacturing infrastructure and specialized workforce remain heavily centered in Taiwan, the continuity of the entire global technology ecosystem hinges on regional geopolitical stability with no viable near-term substitute.

精炼阅读摘要

《Acquired》节目指出，台积电（TSMC）通过开创不与客户竞争的“纯晶圆代工”模式，颠覆了半导体垂直整合的传统格局，并在数十年的工艺积累与巨额资本支出的推动下，构建起全球领先芯片制造的近乎垄断地位；然而，其极度集中的地理位置也使其成为全球经济中最关键且脆弱的地缘政治支点。纯代工模式颠覆垂直整合 (56:10) 在20世纪80年代中期的半导体行业，主流观念遵循AMD首席执行官Jerry Sanders所谓“真正的男人拥有晶圆厂”的准则，芯片设计与制造高度绑定在垂直整合制造商（IDM）内部。张忠谋受命创立台积电时，提出了激进的反定位策略——专注于合约制造，承诺绝不推出自有芯片设计。在创立初期，该模式被普遍质疑，因为当时芯片初创企业几乎必须自建产线，外部晶圆代工市场几乎不存在。台积电最初只能承揽英特尔、摩托罗拉等巨头的过剩产能以求生存，甚至在成立时被评估为零美元估值。节目主讲人认为，台积电本质上扮演了整个芯片创新领域的终极风险投资人和基础设施平台（类似于云计算领域的AWS）。这一模式彻底打破了芯片设计的高资本门槛，直接催化了英伟达、高通、博通等无晶圆厂（Fabless）设计巨头的崛起，并将原本垂直一体化的半导体产业链重塑为高度专业化的水平分工格局。学习曲线定价与资本飞轮 (28:17) 张忠谋早年在德州仪器（TI）担任高管时便发现，由于半导体制造的学习曲线效应，新晶圆厂初始良率极低，传统通过高定价回收巨额固定成本的做法会抑制产能爬坡所需的订单需求。为此，他引入波士顿咨询公司的“学习曲线定价法”，即在投产初期主动调低价格以迅速抢占规模，进而通过海量生产拉升良率并持续降价。这一理念后来深刻嵌入台积电的运营哲学，形成了通过产能规模驱动良率、良率换取现金流、再将现金流全额再投资的飞轮。随着制程微缩推进，“摩尔第二定律”（洛克定律）显示晶圆厂建造成本大约每四年翻一番。巨额资本开支迫使行业加速淘汰：2000年代初处于前沿节点的22家制造商，到5纳米节点时仅剩台积电与三星两家，英特尔则屡屡掉队。主讲人指出，台积电2020年实现480亿美元营收与200亿美元营业利润，并将其中的170亿美元重新投入资本支出；到2021年，台积电更规划了三年1000亿美元的资本开支，凭借40%左右的高营业利润率构筑了竞争对手难以企及的规模经济壁垒。顶尖工艺壁垒与关键客户战役 (82:59) 在《7 Powers》商业框架下，节目主讲人将台积电视为“工艺能力（Process Power）”的终极体现。半导体制造融合了物理、化学与精密工程的极限挑战，台积电40年累积的组织隐性知识和工程师协同网络，构成了即使投入数百亿美元也无法在短时间内复制的护城河。此外，台积电与全球唯一极紫外光刻机（EUV）垄断供应商ASML形成了深度的共生关系。在美国政府推动荷兰限制对华出口先进设备的背景下，台积电稳居ASML最核心的客户位置，进一步锁定了先进制程的供给侧垄断。而在需求侧，关键的历史转折点出现在2000年代末至2010年代初。英特尔曾因短视的定价条款拒绝为初代iPhone制造芯片，被主讲人评价为商业史上最严重的战略失误之一。随着张忠谋在2009年以78岁高龄重返CEO职位，台积电果断抓住移动计算与云计算的双重浪潮，投资90亿美元与6000名员工为苹果定制专属产线。苹果将全部A系列自研芯片代工交予台积电，不仅彻底改变了与三星的竞争格局，也为台积电提供了无可替代的顶级先进制程验证平台与确定性现金流。地缘集中风险与工艺迁移困局 (119:08) 节目强调，台积电最大的基本面风险并非来自商业竞争对手，而是其高度集中在台湾本土所带来的极端地缘政治脆弱性。在战略分析框架中，这种地理集中性被定性为严重的“反权力（Anti-Power）”。如果发生台海冲突导致先进制程产能中断，全球绝大部分关键科技产业链（从汽车生产线到云计算中心）都将陷入停摆。主讲人分析指出，制造工艺与复杂的工程师生态具有深度物理绑定，极难通过单纯的海外建厂迅速移植，类似于丰田早年在加州NUMMI工厂难以完全复制其生产系统。尽管台积电在《芯片法案》补贴下投资120亿美元赴美国亚利桑那州设厂，但海外工厂在进度与成本上面临重重挑战，且规划中并不包含最顶尖的3纳米或更先进制程。先进制程在地理上的严重依赖，使台积电成为关乎国家安全但短期内难以去中心化的战略焦点。节目总结认为，台积电通过极致的纯代工专注与数十年的复合资本投入，奠定了近现代计算工业的技术基石，其B2B护城河的稳固程度超越了绝大多数依靠消费趋势变动的软件巨头；然而，这项人类历史上最精密的制造奇迹，未来仍将持续受制于硅基物理极限的潜在范式转移，以及大国地缘博弈带来的供应链脱钩压力。

Detailed notes — zh

## Background 本期节目由知名商业科技播客《Acquired》（主持人为 Ben Gilbert 和 David Rosenthal）于 2025 年 1 月 21 日在 YouTube 平台重新母带重制并发布。节目简介指出，本期深度剖析了全球最隐秘却举足轻重的科技巨头——台积电（TSMC）不可思议的崛起史：从 56 岁的行业传奇张忠谋在 1987 年破局创立，到如今为英伟达、苹果、高通乃至英特尔代工几乎所有先进制程芯片，带出了一段交织着创新、坚韧以及尖端激光光刻技术的商业史诗。 ## TL;DR 本期节目详尽复盘了张忠谋从战乱流离赴美求学、在德州仪器开创学习曲线定价法，到 56 岁赴台创办台积电并首创纯晶圆代工（Pure-play Foundry）模式的波澜壮阔历程。台积电通过与无晶圆厂设计公司（Fabless）深度捆绑并形成自我造血飞轮，在极高的资本开支与学习曲线壁垒下逐步超越英特尔与各路先驱，更在 2009 年张忠谋复出后拿下苹果独家大单，全面统治移动与云计算时代。两位主播结合《7 Powers》等商业战略框架，深入剖析了台积电以流程优势（Process Power）和规模经济构筑的万亿帝国护城河，并探讨了它在科技万神殿中的历史地位及其面临的唯一软肋——地缘政治宿命。 ## Atmosphere & style 节目的基调兼具商业考古的史诗感与极客拆解的澎湃热情。两位主持人配合默契且张力十足，既有面对“Rosenthal 公理”等战略失误时的拍案叫绝与挥拳兴奋，又有面对 EUV 光刻机“工业炼金术”时的由衷敬畏。整场叙事从早年个人际遇的戏剧性转折自然过渡到冷峻的商业经济学与地缘博弈，听来跌宕起伏、干货极密，宛如一场酣畅淋漓的现代工业登顶全景史诗。 ## Bulletin - [00:00](https://www.youtube.com/watch?v=tuB3_fJyC5Q&t=0s) 台积电市值已突破 1 万亿美元，成为除沙特阿美之外全世界唯一一家总部不在美国西海岸的万亿美元巨头，制造了全球绝大多数先进制程芯片。 - [56:09](https://www.youtube.com/watch?v=tuB3_fJyC5Q&t=3369s) 56 岁创立台积电的张忠谋开创了“纯晶圆代工”（Pure-play Foundry）模式，将自己打造为半导体界最强超级平台，以极低门槛催生了英伟达、高通等无晶圆厂（Fabless）设计生态。 - [01:17:14](https://www.youtube.com/watch?v=tuB3_fJyC5Q&t=4634s) 先进制程制造门槛急剧收窄，全球具备尖端代工能力的企业从 2000 年代初的 22 家缩减至仅剩台积电与三星两家，台积电垄断了 5 纳米等先进制程 90% 以上的市场份额及全行业 95% 以上的利润。 - [01:21:35](https://www.youtube.com/watch?v=tuB3_fJyC5Q&t=4895s) 台积电宣布三年内资本开支达 1,000 亿美元，凭借高达 40% 的营业利润率与 40 年沉淀的“流程优势”（Process Power），构筑起外部资金即便以万亿美元计也难以短期逾越的绝对制造壁垒。 ## Key takeaways - [12:42](https://www.youtube.com/watch?v=tuB3_fJyC5Q&t=762s) 张忠谋青年时代历经抗战、二战及内战三次战乱，赴美转入 MIT 机械工程，因博士资格考失利以一美元薪资之差进入半导体行业，在德州仪器通宵苦读开启技术生涯。 - [31:20](https://www.youtube.com/watch?v=tuB3_fJyC5Q&t=1880s) 在德州仪器期间，张忠谋通过机械工程实验将 IBM 芯片良率翻倍，并与 BCG 合作开创颠覆性的“学习曲线定价法”，助德仪 IC 业务登顶全球半导体霸主。 - [41:18](https://www.youtube.com/watch?v=tuB3_fJyC5Q&t=2478s) 因调任消费品业务受阻、技术路线转轨迟缓及遭遇天花板，张忠谋错失德仪 CEO 宝座并目睹 IBM PC 大单旁落英特尔，52 岁离开德仪并在短暂停留通用仪器后赴台掌舵工研院。 - [45:13](https://www.youtube.com/watch?v=tuB3_fJyC5Q&t=2713s) 面对台湾早期仅 4% 毛利且无设计、无研发、无知识产权的贫瘠底色，张忠谋用极致排除法独创只做制造、绝不与客户竞争的“纯晶圆代工”（Pure-play Foundry）模式。 - [01:00:57](https://www.youtube.com/watch?v=tuB3_fJyC5Q&t=3657s) 在“真男人才拥有晶圆厂”的 IDM 鼎盛时代，台积电以 0 元投前估值、张忠谋 0 创始人股权起步，靠承接巨头亏损及溢出产能度过最初艰难岁月。 - [01:13:55](https://www.youtube.com/watch?v=tuB3_fJyC5Q&t=4435s) 台积电扮演了半导体行业造雨者（AWS 角色），让初创芯片设计团队免除自建晶圆厂的天文开支，直接孕育了英伟达、博通、高通等无厂芯片巨头。 - [01:50:43](https://www.youtube.com/watch?v=tuB3_fJyC5Q&t=6643s) 摩尔定律（运营成本）与洛克定律（晶圆厂建造成本每四年翻倍）协同作用，使尖端芯片制造注定成为赢家通吃的重资本、高利润垄断赛道。 - [01:24:00](https://www.youtube.com/watch?v=tuB3_fJyC5Q&t=5040s) 极紫外光刻（EUV）已进入物理极限的“工业炼金术”范畴，每秒轰击液态锡滴 5 万次，ASML 设备的绝对垄断与稀缺产能进一步筑高了台积电的先进制程护城河。 - [01:32:48](https://www.youtube.com/watch?v=tuB3_fJyC5Q&t=5568s) 2009 年 78 岁的张忠谋在金融危机中“乔丹式”复出，果断押注移动计算与云端 AI 范式转移，并在 2012 年以 90 亿美元建厂豪赌一举锁定苹果 A 系列处理器独家代工。 - [01:44:53](https://www.youtube.com/watch?v=tuB3_fJyC5Q&t=6293s) 上市 27 年间台积电营收年化复合增速达 17.4%，年化 IRR 达 20%，具备强大的逆周期全线提价能力，将周期性重资产制造做成了全球最深护城河。 - [01:57:20](https://www.youtube.com/watch?v=tuB3_fJyC5Q&t=7040s) 基于《7 Powers》分析，台积电拥有不可复制的“流程优势”（Process Power）与“逆向定位”，其工艺良率与组织资产是任何对手仅靠资金砸不出来的壁垒。 - [02:06:45](https://www.youtube.com/watch?v=tuB3_fJyC5Q&t=7605s) 英特尔曾因价格考量拒绝为初代 iPhone 代工芯片（Rosenthal 公理反面教材），并在放弃制造与 IDM 2.0 之间摇摆，陷入了“创新者的窘境”。 - [01:59:08](https://www.youtube.com/watch?v=tuB3_fJyC5Q&t=7148s) 台积电最大的系统性脆弱性并非商业竞争，而是其产能高度集中于台湾本土所面临的地缘政治风险（Anti-Power），其制造生态在海外难以被简单“物理空运”复刻。 - [02:21:10](https://www.youtube.com/watch?v=tuB3_fJyC5Q&t=8470s) 播客最终将台积电评定为全球科技万神殿前十名，并无可争议地列为有史以来第一大 B2B 硬件企业，其推动算力成本渐进趋零，掌控着支撑现代数字文明的“新石油”。 ## Detailed walkthrough ### [00:00](https://www.youtube.com/watch?v=tuB3_fJyC5Q&t=0s) 重制经典：站在 2025 年回望芯片霸主台积电 Ben 和 David 在节目开篇回顾道，听众们一直反馈当年关于台积电（TSMC）的这期节目是《Acquired》有史以来最精彩的单集之一。值得一提的是，这期节目甚至早于他们备受关注的英伟达（Nvidia）特辑，最初录制于 2021 年。当时《Acquired》的听众规模只有现在的 12% 左右，意味着绝大多数听众未曾听过原版。而到了 2025 年，半导体在整个人类世界中的重要性与日俱增，台积电几乎成了全球先进制程芯片的“唯一工厂”——每一台出货的 MacBook 和 iPhone 里的核心处理器、驱动当今 AI 革命的英伟达芯片，以及高通（Qualcomm）、AMD、博通（Broadcom）等芯片设计公司，甚至是 AWS 这样的超大规模云厂商自研芯片，乃至英特尔（Intel）的大量芯片，全都在台积电代工制造。台积电乘着智能手机浪潮登上了令人难以置信的巅峰，而今在 2025 年开启的 AI 新时代中，它依然是这一切背后的制造超级大国。为此，两位主持人决定找出原始录音音轨，从头进行全面母带重制（Remaster）。David 查阅数据发现，自 2021 年首次录制以来，台积电的市值已经翻了一倍，从 5500 亿美元暴增至突破 1 万亿美元。如今放眼全球，除沙特阿美（Saudi Aramco）之外，台积电是全世界唯一一家总部不在美国西海岸的万亿美元巨头，这一极端的“离群值”地位充分展现出这家公司的不可思议。 ### [02:32](https://www.youtube.com/watch?v=tuB3_fJyC5Q&t=152s) 第九大巨头的传奇序幕与地缘政治作为《Acquired》第九季第三集，本期聚焦的主角正是名震天下却又行事低调的台积电。这又是一期典型的“许多普通人从未听过，它却位列全球最有价值公司前列”的故事。它的创始人张忠谋（Morris Chang）的人生轨迹堪称传奇：他在 56 岁时创立台积电，74 岁退休，78 岁又重新出山挂帅，一举敲定代工苹果全部芯片的划时代世纪协议。无论是如今市面上每一部 iPhone、即将推出的每一台 Mac，还是英伟达、AMD、高通芯片，乃至亚马逊、微软、Meta 自研的专用加速芯片，亦或是汽车、智能家居设备和战斗机中的核心半导体，几乎都打着台积电代工的烙印。然而，这个支撑整个现代世界运转的枢纽，却坐落在一座地位极其微妙的岛屿之上——一方面它拥有独特的政治实体，另一方面中国大陆则视其为不可分割的一部分。本期故事不仅交织着一家初创公司向科技超级大国的跨越、一位不被看好的非典型创始人的逆袭，还夹杂着错综复杂的地缘政治博弈。录制当时恰逢全球芯片大短缺，连福特汽车都因缺少芯片不得不暂停其王牌皮卡 F-150 的生产线，更凸显了这期历时两年半筹备的半导体专题的深远意义。 - **全球万亿市值俱乐部特征**：除沙特阿美与台积电外，其余成员几乎全部集中于美国西海岸（如苹果、微软、英伟达、谷歌、亚马逊、Meta）。 - **张忠谋关键履历节点**：56 岁创业建立台积电，74 岁功成身退，78 岁再度出山执掌大局并拿下苹果超级大单。 - **产业牵动面**：从消费级电子（iPhone、Mac）到 AI 算力中心（英伟达 GPU、AWS 云端芯片），再到传统汽车与军事国防装备。 ### [04:53](https://www.youtube.com/watch?v=tuB3_fJyC5Q&t=293s) 战火中颠沛流离的少年张忠谋故事的起点在 1931 年 7 月的中国宁波。巧合的是，这距离沃伦·巴菲特（Warren Buffett）在内布拉斯加州奥马哈出生仅仅过去了一年，两人在日后的漫长生涯中有着诸多奇妙的对照。张忠谋后来荣获了台湾地区的最高文职荣誉——一等卿云勋章（Order of the Propitious Clouds with Special Grand Cordon），在当地地位尊崇得宛如一位封勋骑士。但在 1931 年，他出生的地方是位于上海以南的沿海城市宁波（今天已是拥有约 800 万人口的繁荣大都市）。张忠谋的父亲曾是一名县政府官员，后来成为一名银行经理。因父亲工作调动，张家在战前的中国各地频繁搬迁。在张忠谋幼年时，家里生活水准属于中产阶级，在当时的中国社会相对优渥平稳。然而平静在他 6 岁时被彻底打破——抗日战争（第二次中日战争）爆发，为了躲避频繁的空袭和战火，幼年的张忠谋跟随母亲逃往香港避难。 ```mermaid flowchart TD accTitle: 青年张忠谋辗转求学与迁徙路线 accDescr: 展示张忠谋从 1931 年出生于宁波，历经抗日战争与内战流离辗转，最终于 1949 年赴美求学哈佛与麻省理工的历程。 subgraph S1["战乱迁徙 (1931-1948)"] A["1931 宁波出生"] --> B["1937 避难香港"] B --> C["1941 逃往上海"] C --> D["1948 重返香港"] end subgraph S2["赴美求学 (1949-1954)"] D --> E["1949 哈佛大学"] E --> F["1951 转学 MIT"] F --> G["机械工程本硕连读"] end ``` ### [06:27](https://www.youtube.com/watch?v=tuB3_fJyC5Q&t=387s) 亲历三场大战：从香港沦陷到内战流离张忠谋对战争的残酷记忆刻骨铭心。1941 年 12 月 8 日，也就是日军偷袭珍珠港仅仅 3 个小时之后，日军便对香港发动了全面进攻与入侵。张忠谋日后常常提到，全世界的人都知道 1941 年 12 月 7 日的珍珠港事件，但鲜有人提及在短短几小时之后，同一幕惨剧就在香港上演。香港沦陷后，一家人被迫再度踏上逃难之路，穿越封锁线返回中国内地，辗转定居上海并度过了几年相对平静的求学时光。然而到了 1948 年，二战虽已结束，国共内战却迎来了决战阶段。伴随着剧烈的社会动荡，一家人不得不再度逃回香港。至此，张忠谋在不满 18 岁的人生里，竟然接连完整经历并逃离了三场重大战争：抗日战争、第二次世界大战以及国共内战。 ### [07:40](https://www.youtube.com/watch?v=tuB3_fJyC5Q&t=460s) 踏上自由大陆：哈佛岁月与现实考量 1949 年，中华人民共和国宣告成立，18 岁的张忠谋也迎来了改变一生命运的转折点。在波士顿一位叔叔的资助与协助下，他成功获得了哈佛大学（Harvard University）的录取通知书，独自一人远赴重洋前往美国求学。刚到哈佛时的体验让张忠谋震撼不已。他多年后深情回忆道：“踏入哈佛校门的那一刻，我心中的感觉是纯粹的狂喜，甚至难以置信。那时的美国正处于其道德领导力与政治民主威望的顶峰，同时也是世界上最富裕、最稳定的国家。”对于一个前 18 年都在战火与流亡中度过的人来说，那种“确信十年后这里的经济与社会制度依然完好如初”的稳定感，是潜心做学问和追求创新的无价基石。他终于不用再担心为了躲避空袭而被迫中断学业了。然而初到异国的兴奋很快迎来了现实的冷水。张忠谋的父母仍在香港无法赴美，他必须靠自己自力更生，独立谋生。在 20 世纪 50 年代初的美国，种族天花板极为严苛。正如张忠谋后来所言： > “在 50 年代初的美国，华人的出路只有洗衣工、中餐馆老板、工程师以及大学教授。对华人而言，只有这几个算得上受人尊重的职业。这里没有华人律师，没有华人会计师，更不可能有华人政客。” ### [09:48](https://www.youtube.com/watch?v=tuB3_fJyC5Q&t=588s) 转战剑桥彼岸：MIT 的工程救国与半导体黎明前夜当时的哈佛大学以培养文科精英、政客、法律和商界人才著称，本科甚至没有设立独立的工程学位体系。深谙生存法则的张忠谋清醒地认识到，要在这个社会立足，工程技术才是华人最扎实可靠的晋升阶梯。于是在哈佛读完大一后的 1951 年秋天，张忠谋做出了果断决定：离开哈佛，沿着剑桥的街道搬到了不远处的麻省理工学院（MIT），转读机械工程（Mechanical Engineering）。尽管换学校导致他比同届工程系学生落后了一年进度，但他展现出惊人的专注力与智力，仅仅用了三年时间，就将本科学士学位以及机械工程硕士学位全部攻读完毕。彼时的科技版图与今天不可同日而语——二战后的军事国防预算正源源不断流入美国西海岸，斯坦福大学的学术与产业孵化也刚刚萌芽。但那个时候，仙童半导体（Fairchild Semiconductor）尚未创立，甚至连半导体行业的黄埔军校肖克利半导体实验室（Shockley Semiconductor Laboratory）都还在孕育之中。一个即将掀起全球硅晶圆革命的年轻人，正站在现代科技巨变的最前端，静静等待着属于他的时代发令枪响。 ### [11:00](https://www.youtube.com/watch?v=tuB3_fJyC5Q&t=660s) 早期半导体拓荒期、博士资格考失利与张忠谋的冷幽默这一时期整个半导体行业还处在绝对的史前洪荒时代。肖克利半导体（Shockley Semiconductor）可能才刚刚起步，世人仍在依赖真空管，像贝尔实验室那样摸索前行；硅材料的应用还要等上许多年，晶体管方兴未艾，集成电路（IC）尚未诞生，一切甚至都还在使用锗（germanium）而非硅来进行制造。在麻省理工学院（MIT）花费三年拿下机械工程硕士学位后，张忠谋原本打算继续攻读博士，彻底完成自己的技术学术训练。然而，命运跟他开了个残酷的玩笑——MIT 的博士资格考试规定每人只有两次机会，而他连续考了两次都没能通过。 ```mermaid flowchart TD accTitle: 张忠谋早年学术受挫与转折 accDescr: 展示张忠谋从 MIT 硕士、两次博士资格考失利到被迫进入职场的路径 subgraph MIT["麻省理工学院 (MIT)"] A["完成机械工程硕士学位 (耗时 3 年)"] --> B["申请攻读博士学位"] B --> C["第一次博士资格考试：失败"] C --> D["第二次博士资格考试：再次失败"] end D --> E["学术道路受阻 转入工业界求职"] ``` 主播们提到，为了准备本期节目，他们看遍并听遍了张忠谋所有公开发表的视频与录音资料，发现他是一个极为幽默风趣、拥有极干练“冷幽默”（dry humor）的人。谈起这段考试失败的经历，张忠谋曾风趣地自嘲：“不幸的是，阻碍我继续前行的最大绊脚石是我挂掉了博士资格考试；但幸运的是，他们非常仁慈地给了我第二次补考机会，而我不幸地又挂了一次。”在斯坦福的一次公开访谈中，台下有观众问他当年究竟是如何戒掉烟瘾的，如今已经九十多岁高龄的张忠谋淡定回答：“我从来没戒过，我现在依然在抽。”他不仅为自己抽烟斗进行辩护，还一本正经地搬出理论：虽然抽烟对肺不好，但对他的精神世界和思考大有裨益，所以他深信这反而延长了自己的寿命；甚至他还去查阅过统计数据，坚称“数据显示抽烟斗的人比不抽烟的人活得更长”。这番妙语连珠生动勾勒出了张忠谋独具魅力的个人性格。 ### [12:42](https://www.youtube.com/watch?v=tuB3_fJyC5Q&t=762s) 福特与希凡尼亚：一美元月薪之差引发的人生转向资格考试失败意味着张忠谋无法顶着博士光环进入研发机构，他必须以普通硕士工程师的身份去市场上找一份真正的基层入门级工作。在当时的年代，科技产业还根本没有成型，更不在一般大学生的择业雷达之上；张忠谋学的是机械工程，而非当时已经存在的电气工程，因此他最初的心仪之选是传统重工业巨头。很快，他拿到了梦寐以求的福特汽车（Ford Motor Company）录取通知，福特开出的起薪是每月 479 美元。与此同时，他还拿到了另一个备选方案——来自希凡尼亚公司（Sylvania）新成立的半导体部门的录用信，月薪为 480 美元。坊间传说，虽然两份薪水仅仅相差象征性的 1 美元，但张忠谋当时主动联系福特，希望对方能跟进加薪哪怕 1 美元、匹敌希凡尼亚的待遇。然而福特傲慢地拒绝了，张忠谋一气之下便接下了希凡尼亚的橄榄枝。尽管这段“1 美元改变半导体历史”的故事可能带有后人加工的演义色彩，但它确凿无疑地将这位年轻的机械工程师推向了一片完全未知的半导体热土。 ### [14:17](https://www.youtube.com/watch?v=tuB3_fJyC5Q&t=857s) 与唐·瓦伦丁的奇妙交集，以及酒吧里的半导体夜校有趣的是，红杉资本（Sequoia Capital）传奇创始人唐·瓦伦丁（Don Valentine）在同一时期也恰好在希凡尼亚的半导体部门开启了自己的职业生涯。不过两人的境遇天差地别：瓦伦丁身处阳光明媚的南加州，打着水球、享受着西海岸的迷人气候，立誓此生再也不离开加州；而张忠谋则被分配到了马萨诸塞州伊普斯威奇（Ipswich）的一处偏僻工厂，过着毫无光环、极其枯燥的倒班与钻研生活。张忠谋对电气工程一无所知，工厂甚至没给他租正式公寓，他只能长期借住在公司指定的快捷旅馆里。为了快速补齐专业短板，他每天下班回到旅馆房间，就啃读当时唯一能找到的半导体权威圣经——由威廉·肖克利（William Shockley）于 1950 年撰写的《半导体中的电子与空穴——附晶体管电子学应用》（*Electrons and Holes in Semiconductors with Applications to Transistor Electronics*）。面对没有大学教授指导的自学困境，张忠谋展现出了惊人的求生意志与应变智慧： - 他发现厂里有一位资深工程师嗜酒如命，几乎每晚都会准时泡在旅馆一楼的酒吧里。 - 张忠谋制订了一套严格的作息：下班先在房间里自学苦读数小时，标记所有卡壳看不懂的物理机制与线路图。 - 深夜等那位资深工程师微醺现身时，张忠谋便带上厚厚的教科书直奔吧台，通过不停请对方喝酒，换取前辈为自己答疑解惑。 - 张忠谋后来深情回忆道：“他并没有帮我解开所有难题，但他解答的已经足够让我继续前进。他是我在电气工程方面最重要的启蒙导师。” 这样的生活持续了整整三年，张忠谋像蜡烛两头烧一样拼命工作、刻苦补课。然而，随着对业务与产业认知的加深，他敏锐地察觉到希凡尼亚并不是一条能载他驶向远方的巨轮。就像唐·瓦伦丁迅速跳槽去雷神（Raytheon）随后转投仙童半导体（Fairchild）一样，张忠谋也意识到必须离开。压垮骆驼的最后一根稻草来自厂里一位高管在全体大会上的奇葩发言，那句名言令张忠谋记了一辈子：“在希凡尼亚，凡是我们能造出来的东西，我们都卖不掉；凡是我们能卖出去的东西，我们都造不出来。”听到公司最高层如此坦承供应链与商业化的彻底脱节，张忠谋心里明白：必须立刻走人。 ### [18:41](https://www.youtube.com/watch?v=tuB3_fJyC5Q&t=1121s) 错位偏见：早期的半导体霸主并非硅谷，而是德州仪器离开希凡尼亚的张忠谋并未追随当时的潮流前往加州淘金，而是将目光投向了位于中间地带的德克萨斯州达拉斯。后世在大众传播中过度神化了仙童半导体与叛逆八人组（Traitorous Eight）对硅谷的开创，但主播指出，在那个年代，“硅谷是中心”纯属后来的公关神话：当时全球在数字与模拟芯片领域体量最大、技术最先进的半导体霸主，不是任何加州公司，而是盘踞在达拉斯的德州仪器（Texas Instruments, TI）。 ```mermaid flowchart LR accTitle: 德州仪器从石油勘测到半导体霸主的演进 accDescr: 展现德州仪器由石油勘测仪器设备商转型为半导体与计算技术巨头的历程 A["1930 年代起步 勘测设备制造"] --> B["为德州石油公司提供 地震波勘测与数据处理"] B --> C["自主研发计算与数字电子技术 支撑石油业务"] C --> D["技术外溢与突破 崛起为全球半导体巨头"] ``` 德州仪器的发家史同样出人意料：它始建于 1930 年代，最初是专门为德州蓬勃发展的石油公司制造测量地震活动仪器的设备商，本质上扮演着石油巨头背后核心技术服务商的角色（类似半导体业界的台积电）。为了支撑庞大的地质数据处理与勘探计算，他们顺理成章地跨界杀入计算和数字电子技术，最终一跃成为半导体工业的统治级巨擘。正是在德州仪器的实验室里，杰克·基尔比（Jack Kilby）几乎与仙童半导体的罗伯特·诺伊斯（Robert Noyce）在同一时期独立共创了集成电路，而基尔比凭借这项发明最终荣获诺贝尔物理学奖。基尔比曾对集成电路与半导体的颠覆性意义留下过一段极具穿透力的论断： - “当时我们在发明它时并未完全意识到，集成电路将把电子功能的成本降低到原本的一百万分之一。” - 主播感慨道，在整部人类文明史上，从来没有任何一项技术发明能够将某种关键资源的消耗成本，如此戏剧性地实现百万倍级别的骤降。而张忠谋即将踏入的，正是由德州仪器引领的这波颠覆全球生产力的滔天巨浪。 ### [21:10](https://www.youtube.com/watch?v=tuB3_fJyC5Q&t=1270s) 晶体管与集成电路：现代半导体的基本构造在深入探讨台积电的崛起之前，有必要理清半导体世界中最核心的几个基本概念。首先是晶体管（Transistor）。人们今天常常将其想象为硅片上极其微小的结构，但理解晶体管最好的方式，是把它看作一个伸出三个引脚的小型封装元件：这三个引脚分别对应输入端、输出端以及控制端。本质上，晶体管就是一个由外部电信号控制的开关，用于决定电流究竟是顺畅通过还是完全阻断，从而在二进制的世界中表达“0”（关）与“1”（开）。晶体管的材料选择决定了其性能与商业可行性。木头是绝缘体，完全无法导电；而铜等优良导体导电能力太强，无法被轻易阻断。半导体（如准金属“硅”）则拥有介于两者之间的独特物理特性，允许人们人为调节其导电状态，因此是构建开关的绝佳介质。在行业早期，锗（Germanium）曾是主要的半导体材料，但它昂贵且稀有；相比之下，硅在自然界中以沙子的形态广泛存在，是地球上储量第二丰富的可开采元素，具备极高的经济优势。在此基础上的另一项革命性飞跃是集成电路（Integrated Circuit, IC）。在集成电路诞生之前，工程师们必须像摆弄老旧的真空管那样，制造单个晶体管并手工将其逐一连接起来。直到 1958 年，杰克·基尔比（Jack Kilby，德州仪器）与罗伯特·诺伊斯（Robert Noyce，仙童半导体）各自独立发明了集成电路，人类才得以在同一块底板上集成大量开关。这一发明奠定了现代计算的基石——如今台积电的 5 纳米先进工艺已经可以在极微小的芯片上塞入数百亿乃至数万亿个晶体管开关。 ```mermaid flowchart LR accTitle: 电子计算元件演进路径 accDescr: 展示从分立式真空管与晶体管发展到集成电路的历史演进过程。 subgraph Discrete["分立元件时代"] A["真空管时代 物理插拔与手动布线"] --> B["单个分立晶体管 体积减小但仍需手工互连"] end subgraph Integration["集成化时代（1958年起）"] B --> C["微型集成电路 IC 1958年基尔比与诺伊斯发明"] C --> D["先进制程芯片 台积电制造百亿级晶体管"] end ``` ### [24:08](https://www.youtube.com/watch?v=tuB3_fJyC5Q&t=1448s) 德仪时期的首战告捷：拯救 IBM 订单与代工雏形 1958 年，集成电路被发明的那一年，27 岁的张忠谋（Morris Chang）恰好加入了德州仪器（Texas Instruments, TI）。此时距离英特尔（Intel）成立还有整整十年时间，德仪在当时的技术地位堪比今天的 Google 加上 Meta，而外界对其颠覆性创新尚未充分察觉。张忠谋加入后接手的第一个项目，便是德仪内部的一个棘手难题——与 IBM 的芯片代工合作。 IBM 当时正在开发其划时代的首款晶体管大型机 IBM 7090，意图彻底淘汰真空管逻辑架构。由于预见市场需求极大，一贯坚持垂直整合、自主制造一切零部件的 IBM 罕见地决定寻找“第二供应商”（Second Source）。IBM 将完整的芯片设计方案交给德仪，委托其补充制造——这种“客户提供设计、德仪负责代工生产”的模式，事实上正是后来晶圆代工（Foundry）商业模式的早期雏形。然而，该芯片的生产制造极其艰难。IBM 自己的晶圆厂良率（Yield）仅有约 10%（即每生产 100 颗芯片只有 10 颗可用，90% 报废）；而德仪的生产线良率更是惨遭滑铁卢，几乎跌至 0%，下线的芯片全军覆没。正如张忠谋后来回忆道：“主管在发愁，操作工在发愁，所有人员都在发愁。” - **专业背景介入**：张忠谋拥有麻省理工学院机械工程专业背景，他敏锐地意识到半导体制造在本质上是一套复杂的机械与化学流程，可以运用机械工程的实验优化方法加以改良。 - **良率突破**：经过四个月的精密调控与试验，张忠谋奇迹般地将德仪生产线的良率拉升到了 20%，直接达到 IBM 自身产线良率的两倍之多。 - **名震管理层**：《IEEE Spectrum》杂志在后来的专题报道中写道，这一战绩让德仪总裁帕特·哈格蒂（Pat Haggerty）直接记住了张忠谋的名字。IBM 起初怀疑这只是运气，特意派工程师团队南下德州实地调查，张忠谋从容地向他们拆解了自己所验证的工程理论与实验流程。 - **职业跃升**：凭借该突破，张忠谋晋升为其人生中第一个管理岗位，掌管包含 20 多名工程师的锗晶体管开发部门。 ### [27:08](https://www.youtube.com/watch?v=tuB3_fJyC5Q&t=1628s) 斯坦福读博、工艺极限与学习曲线定价策略德仪管理层意识到张忠谋是一位罕见的工程与管理奇才。为了留住并培养他，公司做出了极为优厚的承诺：全额资助他前往斯坦福大学攻读博士学位，并在读书期间继续向其发放全额工资。对于张忠谋而言，半导体制造正是他的天命所在。他在斯坦福展现出极高的专注度，仅耗时两年半便斩获博士学位。前斯坦福校长约翰·轩尼诗（John Hennessy）后来在对谈中向他请教快速毕业的秘诀，张忠谋淡淡地回答：“我只是非常专注，几乎不干别的。” 1964 年张忠谋重返德仪时，半导体行业正迎来历史性转折：产业界逐步从锗转向更便宜、更充裕的硅材料。然而，早期的硅掺杂与规模化制造工艺极其繁复，涉及化学、物理和数学等多重学科突破的层层叠加。半导体制造开始成为人类工程史上最复杂的科学实践之一。这亦为台积电未来的商业价值埋下了伏笔：即便到今天，许多人误以为代工厂只是低技术含量的“代工制造”，但台积电凭借其在制造技术上的深度 IP 积累，能够长期维持高达 40% 的营业利润率。芯片设计固然困难，但在物理层面上将数以亿计的晶体管制造出来并保证极高良率，同样是人类工程科技的巅峰。 1967 年，张忠谋被提拔为德仪半导体事业部旗下部门的总经理。在此期间，他迎来了自己在商业策略层面的重大突破。以往当德仪及行业其他厂商投资建设耗费巨资的新晶圆厂时，惯常的做法是针对新品制定高昂售价，以尽快收回沉没成本。但张忠谋基于自己在 IBM 产线上的实战经验敏锐地洞察到：半导体制造并非静态生产，其良率与成本深受“学习曲线”（Learning Curve）的驱动，急于在高位定价反而会压制规模的迅速形成。 ### [31:20](https://www.youtube.com/watch?v=tuB3_fJyC5Q&t=1880s) 学习曲线定价法：德州仪器的制胜法宝芯片制造存在一条极其残酷的良率曲线（Yield Curve）：在开发新工艺初期，制造良率通常极低。对于晶圆厂而言，最理想的运营状态是新厂一旦上线，就必须立刻满负荷运转（Max Capacity）。这样不仅能以最快速度积累数据、摸索工艺，推动良率跨越临界点进入盈利区间，而且由于建厂的高昂固定成本早已沉没，维持满产是摊薄固定成本、释放巨大经营杠杆的唯一途径。当时业内传统做法是在初期定下极高售价，试图直接将高昂的启动成本转嫁给客户，结果抑制了市场需求，导致产能无法及时爬坡。张忠谋看透了这一点：与其被动等待，不如主动承担早期的产能爬坡成本。他在 1960 年代末聘请了当时规模尚小的波士顿咨询公司（BCG），共同制定了极具颠覆性的“学习曲线定价法”（Learning Curve Pricing）。张忠谋曾如此回顾这项开创性的策略： > “那是在 60 年代末，当时的波士顿咨询公司（BCG）还只是一家很小的机构。我们结合了海量数据、理论模型并付出了巨大心血。最终的成果就是所谓的‘学习曲线定价法’：起步就定低价，随后每个季度主动且连续地降价，哪怕市场根本没有提出降价要求。这项举措非常成功，尽管在当时引发了巨大争议——很多人觉得我们疯了：‘市场又没逼你降价，你为什么平白无故主动降价？’但我们坚持做下去，因为我们笃信这个底层逻辑。结果，我们的市场份额持续狂飙。” 这项激进的定价策略结合其他打法，直接让德州仪器（Texas Instruments, TI）的集成电路（IC）业务一跃成为全球规模最大、盈利能力最强的半导体霸主。彼时恰逢 Intel 刚刚创立，仙童半导体（Fairchild）与国家半导体（National Semiconductor）横行，但德州仪器在张忠谋的操盘下彻底击溃了所有竞争对手。 - **半导体晶圆制造的经济逻辑**：固定成本极高、边际成本极低，必须依靠海量产能拉升良率，实现巨大的经营杠杆。 - **与硅谷早期生态对比**：硅谷依托斯坦福大学以及政府国防订单迅速膨胀，但德州仪器凭借扎实的成本与规模优势独占鳌头。 ```mermaid flowchart TD accTitle: 学习曲线定价机制 accDescr: 展示德州仪器如何通过主动定低价推动良率与规模优势的正向飞轮 subgraph 飞轮起点 A["制定低价 (Learning Curve Pricing)"] --> B["刺激海量市场需求"] end subgraph 制造壁垒 B --> C["晶圆厂迅速实现满负荷运转"] C --> D["加速摸索制造工艺，良率快速飙升"] end subgraph 市场垄断 D --> E["单位边际成本断崖式下跌"] E --> F["持续主动季度降价"] F --> G["挤压竞对并夺取全球最大市场份额"] G --> A end ``` --- ### [34:14](https://www.youtube.com/watch?v=tuB3_fJyC5Q&t=2054s) 赞助商插播：Vanta 节目中间穿插了赞助商 Vanta 的宣传。Vanta 是领先的信任管理平台，能够自动化处理各类安全审查与合规框架，包括 SOC 2、ISO 27001、GDPR 以及 HIPAA 等，提供实时持续监控而非静态点状报告。根据 IDC 白皮书数据，Vanta 每年可为客户带来约 53.5 万美元的综合效益，投资回报周期仅需三个月。 --- ### [36:17](https://www.youtube.com/watch?v=tuB3_fJyC5Q&t=2177s) 命运的转折点：错失 CEO 宝座与三大诱因凭借在半导体业务上的辉煌战功，张忠谋于 1972 年升任德州仪器副总裁（VP），仅次于 CEO，统领整个半导体事业部。他成为下一任 CEO 最顺理成章的头号接班人。这也是他多年来在达拉斯酒吧里通宵攻读半导体教科书所追求的终极目标。然而历史往往充满变数，张忠谋最终与 CEO 之位失之交臂。主播们深入梳理并剖析了导致张忠谋职场受挫的三大核心视角： 1. **隐性种族歧视**：尽管缺乏直接的文字实证，但身为第一代华人移民，即使早已加入美国国籍，在 1970 年代的美国传统大企业晋升到最高权力中枢，极有可能遭遇了无形的“天花板”与种族偏见。 2. **调任消费品业务的战略误判**：1978 年，在执掌半导体部门六年之后，管理层因消费品部门（如计算器业务）陷入困局，决定调任以管理见长的张忠谋出任消费品事业部 VP，试图力挽狂澜。然而消费业务依赖的是品牌调性、广告营销与变幻莫测的消费者心理，与比拼技术演进和生产成本的 B2B 半导体完全不同。张忠谋未能成功扭转局面，并于 1983 年被降职为形同虚设的“质量与人员效能主管”，被边缘化并“打入冷宫”。 3. **技术路线错失与人才流失**：在 1970 年代中叶，半导体行业迎来从双极型工艺向金属氧化物半导体（MOS，CMOS 的前身）的重要技术代际变迁。坐拥顶尖工程实力的德州仪器却在这场关键转型中反应迟钝，大量核心技术骨干纷纷流失至摩托罗拉（Motorola）或出走创业。例如传奇人物 L.J. Sevin 离开德州仪器创立了 Mostek，后来还成为著名半导体风投机构 Sevin Rosen Funds 的创始人。德州仪器刻板的忠诚文化无法阻挡硅谷的崛起，技术优势逐渐被加州同行蚕食。张忠谋对调任消费品业务曾做出极为深刻的评价： > “当时的德州仪器董事长兼 CEO 马克·谢泼德（Mark Shepherd）深信当时盛行的一套管理信条，即‘一个出色的管理者能管好任何业务’。但在这次实践中，我认为他完全错了。我发现消费品业务截然不同：客户群体完全不同，市场逻辑完全不同，要想脱颖而出所需要的核心能力也完全不同。在半导体业务中，本质就是技术与成本；而在消费市场，技术虽有帮助，但更核心的是迎合消费者的喜好——那是一种极其玄妙、难以捉摸的东西。” - **1972 年**：张忠谋升任德州仪器副总裁，统领半导体事业群，位列 CEO 最热门人选。 - **1978 年**：被平调至挣扎中的消费品事业部，遭遇水土不服。 - **1983 年**：被明升暗降为质量与人员效能负责人（Head of Quality and People Effectiveness），职业生涯跌入低谷。 --- ### [40:27](https://www.youtube.com/watch?v=tuB3_fJyC5Q&t=2427s) 历史的刀锋时刻：IBM PC 绝密招标德州仪器在技术转轨上的失误，最终在 1980 年酿成了半导体工业史上最具标志性的错失。当时张忠谋已被调离半导体部门，IBM 位于佛罗里达州博卡拉顿（Boca Raton）的臭鼬工厂（Skunk Works）秘密团队启动了一项代号绝密的个人电脑项目，即日后改变整个计算机行业格局的 **IBM PC**。作为一直以大型机（Mainframe）闻名业界的巨无霸，IBM 破天荒地决定为这款微型计算机寻找外部供应商，并发放了绝密的 CPU 处理器竞标邀请（RFP）。如果在数年前张忠谋全权执掌半导体部门的鼎盛时期，德州仪器绝不可能让这份改变产业版图的历史级大单从手中溜走，然而此刻的德州仪器正面临战略脱节与人才失血，硅谷的对手即将彻底颠覆格局。 ### [41:18](https://www.youtube.com/watch?v=tuB3_fJyC5Q&t=2478s) 德州仪器的致命错失与张忠谋的三十年谢幕在个人计算机历史的关键十字路口，德州仪器（Texas Instruments, TI）不慎驶离了时代的主干道。当时 TI 与 IBM 拥有长达数年的深厚渊源，TI 理应是初代 IBM PC 微处理器的首选供应商；然而，IBM 最终却选用了英特尔（Intel）的 8088 芯片。这一抉择彻底引爆了连锁反应：整个计算世界的架构迅速被锁定在 x86 之上，开启了由 IBM、英特尔处理器、微软操作系统以及日后成千上万的 IBM PC 兼容机所主导的“Wintel”黄金时代。下一代计算浪潮的王冠，就这样从 TI 头上滑落，稳稳戴在了英特尔的头上。这一重大战略错失是否该归咎于张忠谋个人？答案并不清晰。TI 内部盛行将核心骨干横向轮岗的文化，高层派张忠谋去收拾消费电子业务的烂摊子，但他不仅难以挽回消费电子的溃败，还被迫远离了半导体核心战场。作为众人眼中 TI 下一任 CEO 最无可争议的人选，张忠谋的上升天梯在此时戛然而止。1983 年，52 岁的张忠谋在担任了几年虚职幕僚长后正式递交辞呈，结束了自己在 TI 长达近三十年的辉煌生涯。 - **IBM PC 的架构抉择**：IBM 选用 Intel 8088 而非 TI 芯片，奠定了 x86 架构统治数十年的基石。 - **职业生涯拐点**：张忠谋在 1983 年离开 TI，当时年仅 52 岁，从备受瞩目的明日之星沦为半导体行业内的“前权力玩家”。 --- ### [42:43](https://www.youtube.com/watch?v=tuB3_fJyC5Q&t=2563s) 通用仪器的短暂过渡与美国 CEO 之梦的破灭尽管在 TI 黯然落幕，张忠谋在全美半导体行业依然享有极高的威望，猎头与商业机会纷至沓来。他渴望执掌一家大型企业，最终将选择缩小至两个方向：要么转型去投身风险投资（VC）行业，要么接受纽约老牌半导体企业通用仪器（General Instrument, GI）的邀请，出任公司二把手（COO），并附带一条默契条款——若几年内运转顺利，他将接替现任 CEO 执掌全局。出于未能在 TI 登顶的遗憾与好胜心，张忠谋决定接受挑战，举家搬往曼哈顿加入 GI。然而，GI 的企业基因与 TI 截然相反。TI 是典型的技术至上主义温床，强调自研、核心技术突破与工程创新；而位于纽约的 GI 本质上更像是一家早期的科技私募股权（PE）投资机构，其商业模式主要是低价收购各类独立半导体公司或大集团旗下的边缘业务部门，通过财务与运营整理改善报表，再伺机高价转手卖出。这种资本运作与金融工程导向彻底背离了张忠谋的产业初心，仅仅一年后，他便毅然辞职。在不到 18 个月的时间里连续遭遇两次职业挫折，他在事后坦言：“在经历 TI 和 GI 的双重打击后，我认为自己成为一家美国大型公司 CEO 的志向，已经彻底化为泡影。” - **职业路径分水岭**：当时张忠谋在风险投资人与 GI 首席运营官之间做抉择，最终对 CEO 宝座的渴望让他选择了 GI。 - **GI 的商业本质**：GI 缺乏对半导体底层研发的投入，本质上以并购、财务重组和业务转售（资本运作）为导向。 --- ### [45:13](https://www.youtube.com/watch?v=tuB3_fJyC5Q&t=2713s) 4% 毛利的代工孤岛：台湾产业底色的现实审视一个年过半百、在美国登顶无望的技术老兵，如何奇迹般地在太平洋彼岸创立台积电（TSMC）？这绝非顺理成章的归乡故事。张忠谋出生于中国大陆，早在青年时期便赴美深造并加入美籍，此前仅在 1970 年代代表 TI 赴台考察过建厂事宜。对他而言，台湾完全是一片陌生的土地。彼时 1980 年代中期的台湾，完全处于全球产业链的最底层。在西方消费者的记忆中，“Made in Taiwan”（台湾制造）代表的是廉价的塑料芭比娃娃、玩具与初级纺织品。台湾本土企业的平均毛利率仅有可怜的 4% 到 5%，尚未扣除行政管理与运营费用。用商业战略家汉密尔顿·赫尔默（Hamilton Helmer）在《7 Powers》中的理论解释，这种毛利水平意味着行业内的所有超额利润都已被完全套利，所有企业都在同质化的恶性价格战中竞相探底，根本无法沉淀出真正的企业价值。面对这样贫瘠的工业底色，张忠谋凭借深刻的洞察力做出了极致的排除法： ```mermaid flowchart TD accTitle: TSMC Pure Play Foundry Strategic Formulation accDescr: 张忠谋通过排除台湾在研发、设计、营销和知识产权上的绝对劣势，推导出纯晶圆代工模式的过程 subgraph Weaknesses["台湾产业的全面劣势（绝不涉足）"] W1["无研发实力 (No R&D)"] W2["无芯片设计能力 (No Circuit Design)"] W3["无销售与市场渠道 (No Sales & Marketing)"] W4["缺乏知识产权储备 (No IP)"] end subgraph Strength["仅存的潜在比较优势"] S1["半导体制造与晶圆加工潜力"] end subgraph Solution["唯一可行的商业模式"] M1["纯晶圆代工模式 (Pure-Play Foundry)"] end Weaknesses -->|全面规避劣势| M1 Strength -->|最大化单点突破| M1 ``` 正如张忠谋后来那段精辟的论断：“我们在研发上毫无优势，在电路设计与产品设计上毫无优势，在市场营销上毫无优势，在知识产权上更是几乎一无所有。台湾唯一可能具备的优势——而且还只是一种潜在优势——就是半导体制造、晶圆制造。那么，什么样的公司能够匹配这一优势并避开所有劣势？答案就是纯晶圆代工厂（Pure-play Foundry）。” --- ### [48:07](https://www.youtube.com/watch?v=tuB3_fJyC5Q&t=2887s) 李国鼎的召唤与工研院的“半退休”之旅台湾当局深知依靠 4% 毛利的代工组装无法立足，必须依靠技术升级摆脱贫困。早在 1970 年代，台湾便设立了工业技术研究院（ITRI，工研院），旨在打造“台湾版贝尔实验室”，从美国引进技术转让并孵化本土企业。时任台湾核心决策层的重要官员李国鼎（后来被尊称为“台湾经济奇迹之父”）向张忠谋递出了橄榄枝，力邀这位美国半导体工业领袖赴台出任工研院院长。从外界来看，接受这个职位近乎不可思议。世界各地由政府自上而下主导“打造下一个硅谷”的宏伟规划，绝大多数最终都沦为平庸的政绩工程。张忠谋的前同事与妻子均极力反对，而此时他的婚姻也正走向破裂。面对职业与情感的双重瓶颈，张忠谋抱着“换个环境、换个心境”的念头接下了聘书。对张忠谋而言，这份工作更像是一种体面的“半退休”。他当时虽然谈不上巨富，但在 1985 年较低的财富门槛以及高利率环境下，其积累的资产足以支撑他按照自己简朴的生活方式安度晚年，无须再为生计奔波。然而，当这位习惯了美国科技巨头高强度、雷厉风行作风的行业统帅，在 1985 年踏入由台湾公务员体制主导的工研院时，一场剧烈的文化冲突拉开了帷幕。 - **李国鼎的角色**：作为台湾高科技转型的关键操盘手，力邀张忠谋回台掌舵工研院。 - **工研院定位**：承担技术引进与转化的国家级科研机构，也是后来联电（UMC）与台积电的孵化摇篮。 - **赴台初始心态**：张忠谋最初并未计划创办台积电，而是将其视为摆脱美国生活低谷的“准退休差事”。 ### [51:05](https://www.youtube.com/watch?v=tuB3_fJyC5Q&t=3065s) 工研院的技术底子与联电的诞生张忠谋初到工研院（ITRI）时面临的环境非同寻常。当时台湾仍处于戒严体制末期，工研院内部的人事架构和管理体系近乎官僚机构的“铁饭碗”，大家过着终身雇佣式的生活。张忠谋后来回忆说，院里的员工都把他看作一个“突然空降成为顶头上司的外国人”，对他充满了戒备与恐惧——而事实证明，他们确实有理由害怕这位以严苛治军著称的半导体铁血老将。不过，在张忠谋到来之前，台湾当局已经走对了一步关键的棋：他们成功与美国工业巨头 RCA 谈判，达成了一项半导体技术转让许可。这项技术转让虽然让台湾迈出了第一步，但实际上只引进了落后行业领先水平近十年的老旧工艺。当德州仪器（TI）、英特尔（Intel）和快捷半导体（Fairchild）在最尖端的先进制程上激烈厮杀时，RCA 自身的技术本就落后一代；等到技术在台湾真正落地投产时，已经比国际顶尖制程落后了两代半之多。在半导体行业，这种老工艺只能生产极低端的芯片，但即便在今天，半导体产业也遵循着类似的逻辑：当晶圆厂退居非先进制程后，它们并不会被关闭，而是继续为汽车芯片、物联网（IoT）设备或家电等不需要极限制程的产品供货。工研院利用从 RCA 引进的技术，于1980年分拆成立了台湾第一家半导体公司——联华电子（UMC，联电）。联电最初并非技术领导者，但长期来看发展得相当不错。它不仅为第三方客户代工制造，也从事自主芯片设计。多年后，联电将自身的芯片设计部门独立分拆，演变成为当今市值高达数百亿美元的芯片设计巨头联发科（MediaTek）。正因为工研院和联电已经踩出了一条路，拥有了初步的技术积累和产业雏形，张忠谋到来时并非一切从零开始。 - **RCA 技术转让**：为台湾半导体产业提供了最初的火种，虽然落地时已落后行业先锋两代半，但完成了基础设施与产业工人的初步积累。 - **联华电子（UMC）**：工研院孵化的第一家半导体公司，后来将其芯片设计业务剥离，诞生了当今的无晶圆厂芯片巨头联发科。 --- ### [54:07](https://www.youtube.com/watch?v=tuB3_fJyC5Q&t=3247s) 李国鼎“无法拒绝的条件”与三天诞生商业计划张忠谋在工研院大刀阔斧地推行改革，全力将其打造成一个高绩效的科研机构。就在他全力投入之际，时任主管科技政策的李国鼎突然找到他，直接提出了一个震撼的要求：“你在工研院做得很好，现在我要你出来创办一家公司。”张忠谋起初以为是让他选拔院里的某个人去负责，但李国鼎直截了当地打断他：“不，我要你张忠谋亲自去办一家半导体公司，而且我要你把它做成世界一流的行业领袖。” 面对这一突如其来的使命，张忠谋几乎没有退路。当时他刚从美国辞职，这是他三年内的第三份工作；虽然他早已实现财务自由，不需要薪水维生，但如果在工研院也以失败告终，他的职业生涯将彻底走到尽头。张忠谋后来幽默而精辟地评价这次谈话：“这就像电影《教父》里的经典情节，‘这是一份我无法拒绝的提议’（an offer I couldn't refuse）。”李国鼎作为政府官员，初衷只是希望台湾能拥有一家像英特尔或 IBM 那样成功的半导体大厂，至于具体怎么做，他给张忠谋留出了极大的裁量权。李国鼎给张忠谋定下的期限原本是一周：“回去拿一份商业计划书来见我，告诉我你需要多少资源，我们全力把它促成。”然而仅仅过了一天，李国鼎就改口通知：“周五就来向我汇报。”张忠谋手中仅剩区区三天时间。常言道“需要是发明之母”，正是这紧迫的三天极限构思，催生了后来高居全球科技之巅的巨无霸企业。 ```mermaid flowchart TD accTitle: 台积电商业构想诞生脉络 accDescr: 展现李国鼎下达创业指令及张忠谋在约束下提出纯晶圆代工模式的推演流程。 subgraph Directives["政府指令与外部约束"] A["李国鼎下达指令： 创办一家世界领先的半导体公司"] B["极限时间约束： 由一周缩短至仅剩3天汇报"] C["自身退路封闭： 三年内第三份工作，只能成功不许失败"] end subgraph Evaluation["现实资源评估"] D["台湾优势： 出色的工程师与生产制造纪律"] E["致命短板： 无芯片设计能力、无全球销售渠道、无知识产权"] end subgraph Breakthrough["战略突破"] F["独创商业模式： 纯晶圆代工（Pure-Play Foundry）"] end A --> D A --> E B --> Breakthrough C --> Breakthrough D --> F E --> F ``` --- ### [56:09](https://www.youtube.com/watch?v=tuB3_fJyC5Q&t=3369s) 纯晶圆代工的疯狂：“真男人才拥有晶圆厂” 张忠谋交出的方案堪称惊世骇俗：创办一家专门从事合约制造的“纯晶圆代工”（Pure-Play Foundry）公司，只替别人代工制造芯片，绝不涉足芯片设计与自主品牌。在几十年后的今天，人们习惯顺着结果向后推演，称赞这是无与伦比的远见卓识；但在1980年代中期的半导体产业界，所有人都会一致认为这是一个极其愚蠢的主意。当时全球半导体行业奉行的是完全垂直整合的 IDM 模式（Integrated Device Manufacturer）。从英特尔、摩托罗拉到日本电气（NEC），所有正统半导体巨头都牢牢掌控从芯片设计、掩模制造、晶圆加工到封装测试的全产业链。AMD 联合创始人兼首席执行官杰里·桑德斯（Jerry Sanders）曾有一句响彻硅谷的傲慢名言：“真男人才拥有晶圆厂”（"Real men have Fabs"）。在主流认知中，半导体生产绝非组装芭比娃娃那样可以随意剥离的低端代工，而是整个技术体系中最核心、最精密的生命线。两位主持人在此深入探讨了芯片制造与现代电子代工之间的本质差异。人们常常将苹果将制造外包给台积电与富士康相提并论，但这两者有着天壤之别：在装配线上组装一部 iPhone 属于精密制造，而将下一代先进制程的芯片设计蓝图在纳米尺度的硅晶圆上化为实体，其复杂程度堪称现代“炼金术”。这种炼金术只有真正的炼金术士才能驾驭。在当时，既然没有一家正规芯片公司会放弃自己的晶圆厂，一家只负责代工的企业注定会面临最致命的灵魂拷问——正如传奇风投家唐·瓦伦丁（Don Valentine）最关心的那样：“市场究竟在哪里？” - **“真男人才拥有晶圆厂”**：AMD 掌门人杰里·桑德斯的名言代表了当时全行业的共识；讽刺的是，AMD 在2000年代后也撑不住自建晶圆厂的烧钱压力，将制造部门拆分为格罗方德（GlobalFoundries），自己转型成为纯芯片设计公司。 - **纯晶圆代工模式的先驱困境**：在没有现成客户群体的年代，这一方案完全是“拿着锤子找钉子”——因为台湾当时唯一的优势就是制造执行力，张忠谋只能退守纯制造。 --- ### [59:02](https://www.youtube.com/watch?v=tuB3_fJyC5Q&t=3542s) 张忠谋的远见洞察与寻找第一批客户尽管从表面上看，纯晶圆代工模式是一个缺乏买家的危险方案，但张忠谋内心深处却洞察到一股正在酝酿的技术洋流。凭借他在德州仪器和通用仪器数十年的高管生涯，他亲眼目睹了无数极具天赋的芯片设计工程师渴望离开大公司自主创业，但横亘在他们面前难以逾越的天堑，正是建厂所需的惊人资本。当时人们认为芯片公司必须自备晶圆厂，而晶圆厂恰恰是半导体产业链中资金最密集、折旧最残酷的重资产环节，设计人才根本无法筹措到足以建厂的天文数字资金。张忠谋坚信：只要有人能够提供高质量、可信赖的独立代工制造服务，降低创业门槛，市场上必然会如雨后春笋般涌现出大批无晶圆厂的芯片设计公司（Fabless）。这正是“如果你建好了，他们自然会来”（If you build it, they will come）的构想。然而，现实是残酷的——长远的市场虽然看得到，但眼下这些设计公司根本尚未诞生，台积电必须先在短期的夹缝中活下来。为了解决生存问题，张忠谋必须施展纯熟的商业现实主义（Realpolitik）。他将目光锁定在那些传统半导体巨头身上。正如他当年在德州仪器早期掌管晶圆生产线时所经历的那样，大芯片厂在推出新产品时经常面临突发性的产能不足，或者在生产线向新一代工艺升级过渡时，仍需维持旧款成熟元件的供给。张忠谋计划通过承揽这些巨头的溢出产能或老旧成熟产品，作为新公司的生存底盘。此时，台湾当局也为他开出了注资条件：政府愿意出资一半，但张忠谋必须从英特尔、德州仪器这类跨国半导体巨头手中融到另一半资金，以便将未来的潜在客户直接绑定为台积电的股东与背书人。怀揣着这一设想，张忠谋踏上了游说全球半导体巨头的漫漫征程。 > “在德州仪器和通用仪器时，我看到很多集成电路设计师想出来自立门户，但阻碍他们的最大障碍，就是他们筹集不到足够的钱来创办公司……因为在当时，大家都认为每家半导体公司都必须拥有晶圆制造能力，而制造是资本最密集的部分。我看到那些想离开的人，被无法筹集巨资建造晶圆厂的现实所阻拦。” —— 张忠谋 ### [01:00:57](https://www.youtube.com/watch?v=tuB3_fJyC5Q&t=3657s) 苦求投资与零股权创立：台积电的资本起步张忠谋（Morris Chang）为了筹集创办台积电（TSMC）所需的资金，先后拜访了老东家德州仪器（Texas Instruments）以及英特尔（Intel），但遭到了两家巨头一致婉拒：“Morris，我们很尊敬你，但这事行不通。”在几乎穷尽所有可能的情况下，他抓住了最后的救命稻草——荷兰飞利浦公司（Philips）。虽然飞利浦当时也有半导体业务，但张忠谋对其实力并不算十分推崇，后来曾留下一句辛辣的评价，称飞利浦充其量不过是“二流阵营里的第一名”（the first rung of the second-raters）。但在别无选择的现实面前，飞利浦成了唯一愿意掏钱的外国行业玩家。台积电创办时的总资本定为 2.2 亿美元。其中，台湾当局出资 50%（1.1 亿美元），远超他们最初设想的掏腰包额度；飞利浦注资 28%；至于剩下的 22%，则由当时的台湾行政院院长亲自出面，以带有强硬行政色彩的手段，“按头”要求台湾本土的大企业界领袖们分摊认购。更为不可思议的是台积电当时的估值安排与张忠谋的个人待遇：台积电的投前估值（Pre-money valuation）竟然是 0 美元。在硅谷创业逻辑中不可思议的是，身为灵魂创立者的张忠谋没有拿到哪怕 1% 的创始干股，整家公司的股权全部归投资者所有（当局占 50%，飞利浦及其他投资者占 50%）。当时张忠谋的身份更接近一名奉命行事的“政府雇员”与公职先锋，仅拿着固定薪水度日。面对这种极度苛刻、几近“黑帮马仔”式的安排，张忠谋之所以接受，也反映了他当时职业生涯的低谷与背水一战的处境。日后台积电获得巨大成功、张忠谋坐拥数十亿美元身家，全凭他在公司运转起步后，用个人的全部薪金与多余现金流一点一滴去自购台积电股份（台积电随后于 1994 年在台湾证交所上市，1997 年在纽交所挂牌）。 - **初始出资结构（总资本 2.2 亿美元）**： - 台湾当局：出资 50%（1.1 亿美元） - 荷兰飞利浦：出资 28% - 台湾本土私营企业集团：出资 22%（行政高层直接施压认购） - **张忠谋的初始持股**：0 股权，仅作为公职管理人领取薪水，后期个人财富几乎全靠二级市场与私下自有现金增持积累。 --- ### [01:03:35](https://www.youtube.com/watch?v=tuB3_fJyC5Q&t=3815s) 1987 年的奇迹巧合：ARM 诞生与低功耗架构台积电于 1987 年正式设立并以 0 元投前估值完成融资。而在同一年的半导体世界版图上，还诞生了另一家改写计算历史的公司——ARM。ARM 最初由苹果（Apple）、Acorn 计算机以及负责制造代工的 ASIC 厂商 VLSI Logic 联合投资成立。 1987 年堪称半导体产业的“奇迹之年”（Annus Mirabilis）。彼时的半导体界正完全被 x86 架构统治，台式个人电脑始终连着电源插座，整个行业奉行的是“真男人就要有晶圆厂，真男人就要使劲耗电”（Real men have fabs and real men use power）的狂放信条。在当时看来，指令集极精简、主打低功耗的 ARM 显得羸弱不堪甚至派不上用场。但谁也没有料到，这两个诞生于同一年、起初看似边缘的技术力量，日后将深刻绑定在一起，共同撑起从智能手机、平板电脑到现代移动与服务器计算的庞大帝国。 --- ### [01:05:20](https://www.youtube.com/watch?v=tuB3_fJyC5Q&t=3920s) 拆解现代半导体价值链：五大环节与 ASML 伏笔在半导体行业早期，IDM（垂直整合制造）模式是绝对主流——一家公司包揽晶体管发明、芯片设计、晶圆制造以及封装营销的全流程。然而，随着产业复杂度的飙升，价值链在 20 世纪 80 年代开始裂变，最终演进为如今高度分工的五大核心环节： ```mermaid flowchart LR accTitle: 现代半导体产业分工流程 accDescr: 展示从芯片IP、EDA软件设计，经由无晶圆设计公司，再结合设备制造厂商，最终由纯晶圆代工厂制造的产业全链路。 subgraph Design["设计生态"] direction TB IP["底层 IP 架构 (Arm / 接口 IP)"] EDA["EDA 设计软件 (Synopsys / Cadence)"] Fabless["Fabless 无厂设计商 (Apple / Qualcomm / AMD)"] IP --> Fabless EDA --> Fabless end subgraph Manufacturing["制造生态"] direction TB Equip["前道设备制造商 (ASML / Lam / Applied Materials)"] Foundry["晶圆代工厂 (Foundry) (TSMC 台积电)"] Equip --> Foundry end Fabless -->|"交付芯片版图 (GDSII)"| Foundry ``` 1. **底层架构与 IP（Architecture & IP）**：以 ARM 为代表的指令集架构掌控者，以及提供标准化现成功能模块（如片上 USB 接口单元）的 IP 供应商。现代片上系统（SoC）无需从零手搓每个组件，直接外购成熟 IP 授权已成惯例。 2. **电子设计自动化（EDA）工具**：以 Synopsys 和 Cadence 为主导的专业工业级芯片设计软件，堪称半导体工程师的“Excel 与 Figma”。EDA 软件必须在底层对代工厂的物理制造规则具有深刻感知。 3. **无晶圆厂芯片设计公司（Fabless）**：如苹果、高通（Qualcomm）、后期的 AMD，以及早期的 P.A. Semi 和如今的人工智能算力新秀 Cerebras，专注研发与架构实现，自身不再拥有高昂且笨重的制造工厂。 4. **半导体制造设备巨头（Equipment Manufacturers）**：制造晶圆厂内部精密设备的厂商，包括美国的泛林集团（Lam Research）、应用材料（Applied Materials）以及日本的东京电子（Tokyo Electron）。而在这其中，站着无可替代的荷兰光刻机垄断者——阿斯麦（ASML）。 5. **纯晶圆代工厂（Pure-play Foundry）**：以台积电为核心，承接全球无晶圆设计商订单并全力攻坚前沿工艺制造。值得注意的是，ASML 的原名为 ASM Lithography，于 1984 年作为飞利浦与先进半导体材料国际（ASMI）的合资企业而成立。飞利浦不仅在 1987 年为台积电的诞生注入了关键资本，更是 ASML 孕育成立的共同亲手缔造者。正是这层隐秘而深厚的“荷兰纽带”，为日后台积电凭借丰沛的经营现金流大举投入资本支出（CapEx）、率先锁定 ASML 最尖端光刻机产能埋下了极其关键的战略伏笔。 --- ### [01:09:38](https://www.youtube.com/watch?v=tuB3_fJyC5Q&t=4178s) 赞助插播：摩根大通的全球贸易与营运资本服务播客随后切入赞助商环节，介绍了长期战略合作伙伴摩根大通支付（J.P. Morgan Payments）。针对半导体与尖端制造业这类错综复杂的全球化产业链，摩根大通旗下不仅负责处理交易清算与支付系统，其专属的“贸易与营运资本集团”（Trade and Working Capital Group）还为跨国企业在面对冗长复杂的生产流程与供应链库存周期时，提供风险对冲手段及流动性资金支持。 ### [01:10:27](https://www.youtube.com/watch?v=tuB3_fJyC5Q&t=4227s) 商业插播：J.P. Morgan 与长周期全球业务支持在深入半导体产业的演进之前，节目首先带来了一段赞助内容。如果一家企业要给汽车制造商等大客户供货，往往必须先拿到对方明确的实际需求预测才能出货；但在漫长的等待期内，企业仍需获取信贷额度并管理复杂的供应链。J.P. Morgan 的贸易与营运资金业务团队正是为应对这类复杂局面而设立。除了在支付产品领域的行业领先地位外，J.P. Morgan 拥有跨越数个世纪的全球深厚积累（其历史可追溯至 1799 年），帮助企业穿透经济增长与市场波动的周期。正如台积电（TSMC）数十年如一日穿越宏观经济、地缘政治与全球供应链的风浪一样，发展中的企业同样需要像 J.P. Morgan 这样在幕后协助企业推进结算、化解风险并获取充裕流动性的长期金融伙伴。 --- ### [01:11:43](https://www.youtube.com/watch?v=tuB3_fJyC5Q&t=4303s) 支线故事：从 ARM、DEC、StrongARM 到 P.A. Semi 与苹果芯片的传承随后，主持人 David 讲起了一个在调研本期节目时意外发现的历史彩蛋——P.A. Semi（帕罗奥图半导体公司）的真正身世。P.A. Semi 曾是 Acquired 在第 20 多期探讨过的一家初创公司，如今它的技术血脉却构成了当今 Apple Silicon 的基石。其渊源最早可追溯至 1987 年诞生的 ARM。作为纯 IP 芯片架构授权公司，ARM 早期的核心授权客户之一是老牌计算巨头 DEC（数字设备公司，Digital Equipment Corporation）。DEC 获得授权后，针对极致性能优化了 ARM 架构，开发出著名的“StrongARM”微处理器系列。令人啼笑皆非的是，x86 架构的霸主 Intel 后来收购了 DEC 的 StrongARM 业务线，将其重新包装为 XScale，但最终并未将其发扬光大，甚至将其关闭淘汰。这让当年 DEC 内部专注 ARM 的核心工程师团队感到极度愤怒和失望——在崇尚 x86 垄断的 Intel 内部，另类架构根本毫无出路。于是这群顶尖工程师选择离开，共同创办了 P.A. Semi。之后，苹果为了自主掌握核心计算硬件，收购了 P.A. Semi，并以此为班底打造出今天苹果体系中最深厚、最坚固的技术护城河——Apple Silicon。从 ARM（本身最初也有苹果的合资背景）、DEC、Intel、P.A. Semi 再到今天的 Apple Silicon，整个技术演进完成了一个跨越数十年的奇妙闭环。 ```mermaid flowchart LR accTitle: Apple Silicon 的技术演进脉络 accDescr: 展示从早期的 ARM 与 DEC 合作，经过 Intel 收购后出走创立 P.A. Semi，最终被苹果收购成为 Apple Silicon 的历史过程。 subgraph Origin["技术起源与演进"] direction TB A["ARM (1987年成立)"] --> B["DEC (数字设备公司)"] B --> C["StrongARM 架构"] end subgraph Corporate["并购与流转"] direction TB C --> D["Intel 收购 StrongARM (后改名 XScale 并遭冷落)"] D --> E["核心工程师离职创业 成立 P.A. Semi"] end subgraph AppleEcosystem["终局归宿"] direction TB E --> F["Apple 收购 P.A. Semi"] F --> G["Apple Silicon (A系列 / M系列芯片)"] end Origin --> Corporate Corporate --> AppleEcosystem ``` --- ### [01:13:55](https://www.youtube.com/watch?v=tuB3_fJyC5Q&t=4435s) 靠巨头的“残羹冷炙”起步与张忠谋对无厂半导体（Fabless）的豪赌镜头拉回创办初期的台积电。在最初几年，台积电根本没有任何稳定的市场，张忠谋只能动用自己在美、欧、日半导体巨头的所有老关系和私人人脉，到处恳求前同事们把最边缘、最不要的业务分给台积电，用张忠谋的话说就是“啃残羹冷炙（drags）”。在当时，IDM（垂直整合制造商）只有在两种情况下才会把晶圆代工订单交给台积电： - **自建晶圆厂产能告急时**：作为临时缓冲区救急；一旦他们自己的新产能上线，代工订单就会立刻被无情撤回，因此订单极不稳定。 - **产品严重亏损、自己不想再做时**：这些芯片之所以被抛出来，是因为 IDM 自己造会亏本，因此他们实质上只是把生产亏损转嫁给了台积电。当时的半导体行业普遍断言：“纯晶圆代工（Pure Play Foundry）根本没有市场空间”，这项业务因此被主流视作无稽之谈。然而，张忠谋却看到了当时几乎所有人都没有注意到的趋势：Fabless（无厂芯片设计）模式即将迎来大爆发。张忠谋坦承，当时他无法百分之百确信这一定会发生，但他拥有比 Intel、德州仪器（TI）或摩托罗拉高管更强烈的期盼与嗅觉。在德州仪器与通用仪器（General Instrument）任职期间，他亲眼目睹了无数天才芯片设计师渴望自立门户、成立创业公司，但都被自建晶圆厂高达数千万甚至数亿美元的天文数字成本死死困住。张忠谋原本在考虑转型成为一名风险投资人（VC），但他创办台积电实际上完成了一次无与伦比的战略迂回：他把自己变成了“半导体行业有史以来最强大的超级 VC”。台积电通过提供代工制造这一底层平台，为整个芯片创业与创新市场做底层兜底和指数级赋能。如果说 Y Combinator 孵化了互联网创业群，或者微信为应用开发提供了生态分发体系，那么台积电就是半导体界的造雨者——它让芯片初创团队只需要筹集以往 10% 到 20% 的启动资金，即可把芯片投入量产。凭借张忠谋在半导体界的泰斗级声望，没有任何一名芯片工程师会拒绝与这位德州仪器的前二号人物面谈。他亲自奔走布道，告诉所有顶尖工程师：“从大公司出来创业吧，你们不再需要筹集巨资造晶圆厂，台积电就是你们的专属工厂！” --- ### [01:17:14](https://www.youtube.com/watch?v=tuB3_fJyC5Q&t=4634s) 飞轮启动：芯片界的 AWS 与先进制程大逃杀经历了最初几年的苦熬之后，无厂芯片设计的创业大潮终于在 1990 年代轰然爆发，一批未来叱咤风云的芯片巨头相继在台积电的襁褓中诞生：高通（Qualcomm）、博通（Broadcom）、迈威尔（Marvell）以及英伟达（Nvidia）。 - 以 1993 年成立的英伟达为例，黄仁勋在整个创业历程中累计仅通过股权融资募集了约 2,000 万美元，并且终生未建一座自己的晶圆厂，几乎 100% 的制造都依托台积电完成。 - 正是凭借这种类似芯片界“AWS”的赋能机制，英伟达才能用仅仅 2,000 万美元的初始融资，一步步成长为如今数千亿美元市值的巨无霸。这一生态为台积电构建了一个超乎张忠谋最初想象的超级增长飞轮： 1. **赋能催生 Fabless 客户群**：台积电为初创芯片公司承担制造重担，推动无厂半导体市场蓬勃壮大。 2. **高毛利与自由现金流**：随着客户出货量激增，台积电的营收快速攀升，长期保持约 50% 的毛利率与 40% 的营业利润率。 3. **利润反哺资本开支（CapEx）与工艺研发**：丰厚的利润被源源不断地重新投入到下一代晶圆厂建设和更先进制程技术的研发中。 4. **制程实现超越并驱动客户性能跃升**：台积电最初在技术上大幅落后于全球同行，但在大约 10 年内迎头赶上，随后在工艺节点上反超全球竞争对手。台积电制造的制程线宽更小、性能更优的芯片，使得其客户的产品竞争力更强、攻占更多下游应用领域，从而反哺给台积电更加庞大的订单与利润。这项飞轮启动时虽然缓慢而艰难，但在随后的二十年间彻底清洗了全球半导体先进制程的竞争格局： - **2000 年代初（约 150 纳米节点）**：全球尚有 **22 家** 处于技术最前沿的先进制程晶圆制造公司，台积电此时刚刚追赶上梯队，成为这 22 家之一； - **2000 年代末**：随着制程微缩的成本呈指数级暴增，先进制程玩家急剧锐减至 **14 家**； - **2010 年代中叶**：全球仅剩 **6 家** 具备先进制程制造能力； - **节目录制时的 5 纳米节点时代**：全球仅剩 **2 家**——台积电与三星（Samsung），而原本的半导体霸主 Intel 此时在先进制程上已显露掉队疲态。 ### [01:19:57](https://www.youtube.com/watch?v=tuB3_fJyC5Q&t=4797s) 3nm 竞赛与半导体行业横向分工的大势所趋在先进制程的残酷角逐中，竞争对手们正逐渐掉队。随着三星将 3nm 制程推迟至 2024 年，台积电（TSMC）虽然也将进度稍稍推迟了六个月，但依然将在次年率先量产 3nm。这意味着在下一代顶级工艺节点上，台积电极有可能成为市场上唯一的赢家。当苹果（Apple）在发布会的主题演讲中展示最新一代 iPhone 芯片的卓越性能并包揽所有赞誉时，台积电非常乐意退居幕后——他们的职责从来不是做终端营销，而是全力“赋能客户”（empower their customers）。台积电的业务飞轮一旦转动起来，便在整个芯片代工行业势如破竹、通吃全盘（run the table）。半导体行业的格局经历了从 IDM 垂直整合（即单家公司包揽从设计到制造的所有环节）到横向专业化整合的根本性转变。推动这一变革的核心驱动力在于摩尔定律惊人的演进速度。技术迭代如此之快，以至于没有任何单一实体能够全盘精通所有环节：从 EDA 电子设计自动化软件、精密制造设备、芯片制造工艺，到芯片本身的微架构设计乃至指令集架构（ISA）。整个产业链被迫拆解，交由各个细分领域最顶尖的公司分工协作。张忠谋对此曾有一个极为形象的比喻：“半导体行业就像一台不断加速的跑步机，如果你跟不上，就会掉下去。”（"The semiconductor business is like a treadmill that speeds up all the time. If you can't keep up, you fall off."）在先进制程从 22nm、2nm 一路狂奔向 1nm 的过程中，即使竞争对手只专注于纯代工这一件事，一旦落后半步，等待他们的也只有出局。 ### [01:21:35](https://www.youtube.com/watch?v=tuB3_fJyC5Q&t=4895s) 千亿美元资本支出与不可逾越的工艺壁垒要在半导体这台高速运转的跑步机上保持不掉队，代价是极其昂贵且不断飙升的资本支出（CapEx）。为了给下一代先进制程建设新晶圆厂，代工厂必须拥有惊人的盈利能力来提供资金血包。台积电刚刚宣布，未来三年内的资本支出将高达 1,000 亿美元——仅当年就达到 300 亿美元，随后两年将追加 600 亿美元，且这一数字未来只增不减。这种高密度的制造能力高度集中在台湾这一地缘政治极度敏感脆弱的区域，引发了全球范围内的战略焦虑。无论是美国政府还是中国大陆，都迫切希望将先进制程供应链迁回本土或实现自主可控。然而，半导体制造绝非光靠政府砸出“资金重炮”（bazooka of money）就能迅速砸出结果的行业。即便美国政府当年拿出 1 万亿美元，也无法在一年内追赶上台积电。台积电的核心护城河是“规模经济”（scale economies）与“工艺实力”（process power）的深度融合： - **学习曲线的不可跃迁性**：工程制造的极高难度和工艺学习曲线需要数十年的一线积累，没有任何捷径可走。 - **设备操作 Know-How 的稀缺性**：正如彭博社《Odd Lots》播客常驻大中华区的科技记者所分析的，即使像中芯国际（SMIC）这样的中国晶圆厂获得巨额投资并尝试自主突围，哪怕把最尖端的制造设备买回来，全球也没有几个人真正懂得如何操作调试并把良率提升到商业化水平。这不仅关乎设备，更关乎全世界最精尖、最难掌握的制造工艺工程体系。 ### [01:24:00](https://www.youtube.com/watch?v=tuB3_fJyC5Q&t=5040s) ASML 与 EUV 极紫外光刻的现代“工业炼金术” 在最尖端设备的供给端，荷兰阿斯麦（ASML）占据着绝对的垄断地位。由于荷兰政府在中美科技博弈以及美国政府的施压下，未续签与中国大陆的贸易许可协议，导致中芯国际（SMIC）等企业根本无法买到最先进的光刻机。但光刻机本身的物理造价与制造难度，就已经超越了普通工业机械的范畴： - **单机成本与运输难度**：一台最先进的光刻机售价高达 2 亿美元（下一代甚至将升至 3 亿美元）。单台机器重达数百吨，需要动用整整 4 架波音 747 货机分批运送。机器送达晶圆厂后，还需要由 ASML 的专业技术工程团队长期常驻现场负责组装、调试并协同运行。 - **从深紫外到极紫外的物理极限**：常规白光的波长为 193 纳米，过去人类通过复杂的光学透镜以及浸润式光刻（在水介质中折射光线），最多只能把制程推进到约 11 纳米。当芯片向 5 纳米、3 纳米逼近时，传统光学路径彻底走到了尽头。 - **工业炼金术：EUV 的生成**：为了获得波长仅为 13.5 纳米的极紫外光（EUV），机器内部必须制造出高能等离子体。在真空腔体内，熔融液态锡滴以极高速度落下，高能脉冲激光瞬间精确击中锡滴将其轰爆为等离子体，从而激发出 EUV 光芒。这一激光命中液滴的过程必须以每秒 50,000 次的频率连续进行，且命中精度要求比阿波罗登月计划的轨道计算还要严苛。 - **光学镜面的全新发明**：由于 EUV 光的波长极短，几乎会被包括常规玻璃镜在内的所有已知物质吸收，ASML 必须联合德国蔡司（Zeiss）发明采用数百层特殊涂层的全新布拉格反射镜，并在极高精度的脉冲激光器上深度依赖德国通快公司（Trumpf）。激光束在打到硅晶圆之前需要经历约 20 次反射。 ```mermaid flowchart TD accTitle: EUV 光刻机极端光源生成与晶圆刻蚀流程 accDescr: 展示从熔融锡滴下落、脉冲激光高频轰击到等离子体发射 EUV 光线并经特殊反射镜聚焦至硅晶圆的物理过程 subgraph LightGen["EUV 光源生成 (每秒重复 50,000 次)"] A["高频微滴喷射器: 滴落熔融锡滴 (Molten Tin)"] B["高能脉冲激光器 (极高精度制导)"] C["激光精准轰击锡滴"] D["瞬间爆碎产生高能等离子体 (Plasma)"] A --> C B --> C C --> D end subgraph Optics["光学反射与芯片光刻"] E["激发出 13.5nm 极紫外光 (EUV)"] F["全新特殊多层膜反射镜 (经历多次折射，常规镜片无法反射)"] G["高精密光罩刻蚀硅晶圆 (实现 5nm/3nm 先进制程)"] D --> E E --> F F --> G end ``` 这种被主持人叹为现代“工业炼金术”（true alchemy）的制造设备，ASML 全年产量仅有大约 50 台。过去在光刻机市场叱咤风云的日本巨头尼康（Nikon），最终也因 EUV 的研发难度过于超现实而被迫放弃退场。如果不借助这种全球最前沿的工业设备，苹果未来的 M2、M3 或下一代 A 系列仿生芯片根本无法诞生。本和戴维幽默地打趣道：即使像 Acquired 播客这样经营出色、有社区有声誉、随手掏出几亿美元买机器，或者政府直接扔出千亿美元巨资，想要亲自下场和台积电竞争先进制程，也依然是完全不可能实现的妄想。 ### [01:29:02](https://www.youtube.com/watch?v=tuB3_fJyC5Q&t=5342s) 台积电难以逾越的制造护城河与40%营业利润率哪怕你拥有天文数字般的现金、能够圈地建厂，甚至有渠道买到光刻机——现实中这根本不可能，因为当时阿斯麦（ASML）的设备订单已经积压了五十多台，未来数年的产能全被抢购一空——即便把这些顶级机台摆在面前，也没有人知道第一步该怎么开工。半导体制造的尖端工艺不是照着说明书就能搭建的乐高积木，必须由那些真正在人类历史上完成过最极限、最复杂精密制造的工程团队亲手探索，才能推导出下一代制程节点的工艺。正是这种绝对的技术与经验断层，让台积电（TSMC）坐拥高达 40% 的营业利润率（Operating Margin），展现出令行业为之惊叹的护城河。两位主持人感叹，台积电从最初接手大厂弃之不顾的残羹剩饭（Dregs），到伴随无晶圆厂（Fabless）芯片设计公司的崛起而滚大飞轮，这中间跨越了一道巨大的鸿沟。回顾 2005 年，台积电的发展其实已经远超张忠谋（Morris Chang）最初的设想。当时以英伟达（Nvidia）为代表的无晶圆厂设计公司正崭露头角，玩家们自己组装游戏 PC 时都渴望装上英伟达的显卡。但当时的英伟达还远不是全球市值前二十的巨无霸，整个半导体行业甚至还残留着“真男人就该有自己的晶圆厂（Real men have fabs）”的偏见。那时的英伟达股价仅仅取决于它是否拿下了索尼下一代 PlayStation 或是微软 Xbox 的游戏机芯片合同，这与后来的机器学习、加密货币热潮完全不可同日而语，但对台积电而言，这已经是源源不断且利润丰厚的业务。 - 尖端制造壁垒：ASML 先进机台常年处于数十台订单积压状态，全球顶尖工艺 Know-how 完全集中于极少数工程师手中。 - 业务飞轮：从承接廉价边缘订单起步，通过先进制程研发反哺，最终铸就近 40% 的高营业利润率。 --- ### [01:30:08](https://www.youtube.com/watch?v=tuB3_fJyC5Q&t=5408s) 2005年张忠谋交棒退休，与2009年危机中的“乔丹式”复出到了 2005 年，74 岁高龄的张忠谋萌生退意。他多年来持续用个人资金买入台积电股票，投资收益丰厚，自认已经完成了职业使命，可以功成身退。于是，他将台积电的帅印正式移交给了长期副手蔡力行（Rick Tsai），自己开启了退休生活。在随后的闲暇时光里，张忠谋沉浸于他热爱的文学阅读，安享晚年；他将其第二任妻子（张淑芬）视为重新点燃其生命活力与激情的关键。彼时在技术前沿，极紫外光刻（EUV）也才刚刚从纯科学设想逐步显露出落地的可行性雏形。然而，2008 年席卷全球的金融危机彻底打破了平静。半导体产业陷入混乱与动荡，这场危机让台积电面临前所未有的严峻考验。2009 年夏天，台积电召开了一场出人意料的新闻发布会，投下一枚重磅炸弹：蔡力行卸任 CEO，78 岁的张忠谋重新披挂上阵，回归执掌大局。这被外界戏称为如同迈克尔·乔丹脱下 45 号球衣重返赛场，甚至比乔丹的复出更为惊人——这是张忠谋职业生涯的“第三幕”。虽然蔡力行随后开启了自己的另一段辉煌并出任联发科（MediaTek）CEO，但这次突如其来的换帅在当时被广泛视为一次并不完美的接班过渡。 - 退休与交接：2005 年，74 岁的张忠谋正式退休，将 CEO 权柄交接给蔡力行。 - 危机时刻的复出：2009 年夏，经历金融海啸冲击后，78 岁的张忠谋宣布重返台积电担任首席执行官。 --- ### [01:32:48](https://www.youtube.com/watch?v=tuB3_fJyC5Q&t=5568s) 黄金机遇：智能手机浪潮与纯晶圆代工模式的全面爆发面对外界的疑虑，张忠谋在当年的新闻通稿中掷地有声地宣称：这次人事调整绝不会削弱台积电的“战斗精神”，反而将激发更大的冲劲；更重要的是，他敏锐地看到了台积电前方浮现出的“黄金机遇”（Golden Opportunities）。张忠谋所指的黄金机遇，正是当时刚刚破晓的移动智能手机时代。 2007 年 7 月初代 iPhone 震撼问世，2008 年搭载 App Store 与开放 SDK 的 iPhone 3G 登场，吸引了海量开发者涌入生态，紧接着 Android 系统也相继发布。在此之前，苹果（Apple）虽然施展了“神级魔法”，将基于 Unix、面向 x86 架构精简后的 OS X 系统成功移植到低功耗设备上，但其初代芯片并未自主设计，而是直接采购三星（Samsung）的成品 ARM 芯片，并由三星代工制造。在旧有的 PC 时代，整个计算范式完全建立在插在墙上供电的台式机、英特尔（Intel）和 x86 架构之上；而移动计算对超低功耗的极致苛求，促成了 ARM 架构的崛起。最关键的转折在于：几乎所有引领移动革命的芯片设计公司——高通（Qualcomm）、博通（Broadcom）、联发科（MediaTek），乃至后来自研芯片的苹果——全都是没有制造工厂的 Fabless 公司。张忠谋与这些设计公司的掌舵人均保持着深厚的私交，台积电成了这些巨头将蓝图变为现实的唯一桥梁。2007 年乔布斯在发布会上还谨慎地表示苹果的目标仅仅是拿下智能手机市场 1% 的份额，但到了 2009 年（正值 iPhone 4 筹备发布之际），整个科技界终于如梦初醒：智能手机不是 PC 的附属玩具，而是下一个计算时代的全新范式。 ```mermaid flowchart TD accTitle: 计算范式转变与半导体代工爆发 accDescr: 展示从传统PC时代到移动与云时代的转变，以及台积电在无晶圆厂浪潮中的核心枢纽位置。 subgraph PC_Era ["PC 时代（旧计算范式）"] PC_Arch["x86 架构 / 插电桌面设备"] Intel_Dom["英特尔垂直整合垄断（IDM）"] PC_Arch --> Intel_Dom end subgraph Mobile_Era ["移动与现代时代（新计算范式）"] Mobile_Arch["低功耗 ARM 架构 / 智能手机"] Fabless_Giants["Fabless 阵营崛起 (Apple, Qualcomm, Broadcom, MediaTek)"] Mobile_Arch --> Fabless_Giants end subgraph Foundry_Engine ["台积电制造引擎"] TSMC["TSMC 纯晶圆代工 先进工艺制造与交付"] end Intel_Dom -.->|"范式失守"| Mobile_Arch Fabless_Giants ==>|"全部制造需求交付"| TSMC ``` --- ### [01:35:14](https://www.youtube.com/watch?v=tuB3_fJyC5Q&t=5714s) 移动与云端双浪潮：算力范式转移与英特尔霸权的终结智能手机仅是新范式的一半，与移动浪潮并驾齐驱的是企业端的“云计算”（The Cloud）。移动与云计算两大技术浪潮在接下来的十年中撬动了数万亿美元的市值增长。云计算初期是英特尔的巨大利好，亚马逊（Amazon）等云服务商为了构建数据中心采购了海量单价极高的 x86 至强服务器 CPU。然而，随着云端计算负载向大规模并行处理演进，尤其是在人工智能（AI）与机器学习兴起之后，通用 CPU 的主导地位开始动摇。面对矩阵乘法与矢量数学等高并发负载，GPU 和专用芯片（ASIC）展现出无可比拟的优势。英伟达凭借在并行计算上的积累实现了市值的爆炸式增长，其三四千亿美元的市值已经达到英特尔的两倍左右。更重要的是，晶圆代工模式打破了巨头对尖端半导体的制造垄断，开启了“定制芯片”（Custom Silicon）时代——从科技大厂到特斯拉（Tesla）打造 Dojo 超算芯片，各类企业都能自行定义硬件架构。虽然当时特斯拉自动驾驶芯片部分采用了三星位于美国得州奥斯汀的工厂代工，但两位主持人强调，若想追求极致的 3 纳米及以下先进制程，行业最终都无法绕开台积电。这场计算重心的转移标志着英特尔统治地位的落幕，而台积电则成为了承载全人类底层算力演进的最纯粹“指数基金”（Index）。正是带着这样的宏大格局，重出江湖的张忠谋在年近八旬之时亲自操盘，着手在 2012 年敲定那笔彻底重塑科技版图的历史性协议——拿下苹果 A 系列处理器的独家代工大单。 - 双轮驱动：消费端的移动设备（低功耗 ARM）与企业端的云计算/AI（高并发 GPU），共同构成了过去十年科技资本的最大引擎。 - 架构分工：CPU 负责复杂的串行操作系统任务，GPU 与定制 ASIC 则统治了深度学习、图形渲染与矢量矩阵运算。 - 权力更迭：无晶圆厂模式与台积电先进工艺互相成就，彻底瓦解了英特尔依托 IDM 建立的长达数十年的产业霸权。 ### [01:38:21](https://www.youtube.com/watch?v=tuB3_fJyC5Q&t=5901s) 苹果的豪赌：客厅定局与台积电的专属晶圆厂台积电与苹果的结盟源自一次极为关键的高层私密会晤。据传苹果首席运营官杰夫·威廉姆斯（Jeff Williams）与台积电高管在客厅共进晚餐，在看似家常的氛围中敲定了改变半导体格局的合作。在此之前，苹果收购了芯片设计初创公司 P.A. Semi，启动了自主芯片架构的研发，准备推出初代自主设计的 A 系列处理器（A4）。当时苹果预见到未来全球消费者对 iPhone 的庞大需求，而其现存的代工伙伴三星既是代工商，又通过 Galaxy 系列手机与苹果直接竞争。史蒂夫·乔布斯（Steve Jobs）对三星深恶痛绝，不仅因设计抄袭提起多起法律诉讼，还曾在安卓阵营崛起时放出要发起“热核战争（thermonuclear war）”的狠话；蒂姆·库克（Tim Cook）后来也在公开场合将双方的纠葛形容为“有毒的地狱炖汤（toxic hell stew）”。对于双方而言，这场合作是一场毫无退路的孤注一掷。根据彭博社的报道，当时的台积电在许多外人眼里还只能算是一家二线代工厂。威廉姆斯后来回忆道：“如果我们重注台积电，就没有任何备用方案（no backup plan）。”而对台积电来说，接下订单意味着必须预先砸下 90 亿美元建造晶圆厂，并在短短 11 个月内调动 6000 名员工为苹果建造专属生产厂房，且数年后才能真正投产芯片。早期诸如 A4、A5 处理器仍存在台积电与三星分工代工的情况，但到了搭载更大屏幕、引爆全球抢购潮的 iPhone 6 时代，台积电已几乎成为苹果独家代工商。这种为单一大客户怒掷 90 亿美元产能的“押上身家之举”，充分体现了创始人张忠谋无与伦比的行业威望与胆魄——即便他作为创始人并未持有台积电无偿赠予的创始人股权，其所有持股均是真金白银买来的。 - 合作契机：苹果通过收购 P.A. Semi 自研芯片架构，亟需摆脱与竞争对手三星在代工上的利益冲突。 - 风险代价：苹果毫无替补供应商退路；台积电单项斥资 90 亿美元并派遣 6000 名员工在 11 个月内建厂。 - 市场爆发：iPhone 6 凭借大屏幕设计成为现象级爆款，台积电由此稳坐苹果核心处理器独家制造宝座。 ```mermaid flowchart TD accTitle: 苹果与台积电合作决策流程 accDescr: 展示苹果因与三星竞争而将自研芯片转交给台积电，台积电斥巨资建厂并最终垄断代工的历程。 subgraph AppleSide["苹果的战略转向"] A["收购 P.A. Semi\n自研 A 系列架构"] --> B["摆脱竞争对手三星\n寻找纯晶圆代工伙伴"] B --> C["Jeff Williams 秘密会晤\n无退路押注台积电"] end subgraph TSMCExecution["台积电的超常规投入"] D["张忠谋拍板 90 亿美元资本支出"] --> E["投入 6,000 名员工\n11 个月建成专属厂房"] end subgraph Outcome["市场结果"] C --> D E --> F["早期混用三星与台积电代工"] F --> G["iPhone 6 爆火并转向台积电全包"] end ``` ### [01:41:15](https://www.youtube.com/watch?v=tuB3_fJyC5Q&t=6075s) 张忠谋功成身退与惊人的资本开支军备竞赛促成与苹果的世纪联姻后，张忠谋于 2013 年卸任台积电 CEO，但继续留任董事长把控战略大局。直到 2018 年 6 月，台积电不仅彻底站上世界之巅，张忠谋也在 86 岁高龄正式宣布彻底退休，辞去董事长一职。随着台积电在尖端工艺上的完全领跑，其资本开支规模演变成令所有竞争对手绝望的护城河。2020 年，台积电实现营收 480 亿美元，营业利润高达 200 亿美元；令人惊骇的是，公司将其中整整 170 亿美元的营业利润再度全额投入到资本支出（capex）中。进入 2021 年 1 月，台积电发布业绩指引，将当年的资本支出大幅提升至 250 亿至 280 亿美元；到了同年 4 月，管理层再次上调资本支出预期至 300 亿美元，并宣布未来三年累计资本开支将达到 1000 亿美元。这笔天量预算犹如直接打向整个半导体工业的信号弹，金融市场彻底意识到，台积电正在垄断最先进工艺的市场，其资本重压甚至连三星都无法跟上节奏。 ### [01:42:31](https://www.youtube.com/watch?v=tuB3_fJyC5Q&t=6151s) 赞助商插播：现代产品基础设施平台 StatSig 在探讨台积电的统治级数据前，节目插入了赞助商 StatSig 的介绍。StatSig 从早期的小型创业团队，迅速成长为包括 OpenAI、Figma、Atlassian、Vercel 等硅谷主流科技公司竞相使用的明星平台。公司将此前大厂（如 Facebook、Netflix）内部标配的产品底层架构进行独立化与商业化，提供包括高阶实验工具、A/B 测试、功能开关（Feature Flags）、产品分析与会话回放（Session Replays）等全套组件。如今，StatSig 不仅服务初创公司，还进入了 Bloomberg、微软和 Electronic Arts（EA）等传统与成熟企业，日处理事件量已暴增至 2 万亿次，支持数十亿终端用户的产品迭代体验。无论是基于现有数据仓库的原生部署，还是完全托管方案，StatSig 都能帮助企业以数据驱动决策。他们还提供优厚的初创支持计划（包含价值 5 万美元的 10 亿次免费事件额度）及企业级优惠。 ### [01:44:53](https://www.youtube.com/watch?v=tuB3_fJyC5Q&t=6293s) 二十七载年化 20%：最强护城河与全行业定价权回归台积电的核心数据，主持人指出了一组极其震撼的复利奇迹：自 1994 年在台湾证券交易所上市以来的 27 年间，台积电的营收复合年均增长率（CAGR）高达 17.4%；其市值从上市初期的 40 亿美元飙升至录制时点的 5500 亿美元，年化内部收益率（IRR）接近 20%（达 19.9%）。台积电不仅已成为全球市值第九大的超级巨头，也是全球前十公司中仅次于沙特阿美的非美国老牌企业。从 2019 年到 2020 年，台积电营收增速达 31%，在庞大的体量基数上实现了复合增速的翻倍爆发。面对 3 年 1000 亿美元的超大额资本开支，台积电不仅账上坐拥 280 亿美元的现金与现金等价物，更具备强大的内生造血能力——这得益于其近年来前所未有的“定价权（pricing power）”。张忠谋过去倡导的“主动降价让利”策略已被技术代差带来的溢价所取代；台积电不仅停止了传统的周期性降价，甚至在当年直接宣布全线涨价 20%。通过这种极致的技术垄断，台积电成功将原本属于周期性重资产的大宗制造业务，塑造成商业史上最深厚的护城河之一。 - 资本与市场垄断：台积电占据了全球晶圆代工制造总市场 50% 以上的份额，并攫取了全行业 95% 以上的利润。 - 尖端制程统治力：在 5 纳米等最先进前沿制程上，台积电市场份额高达 90%，而唯一能勉强跟随的三星仅占约 5% 到 10%。 - 战略闭环：将充沛的经营现金流源源不断地回填至尖端产能，使所有追赶者在物理与技术节奏上均落后其五年以上。 ### [01:48:12](https://www.youtube.com/watch?v=tuB3_fJyC5Q&t=6492s) 前沿节点的不可替代性与物理极限挑战台积电在先进制程上的掌控力极其惊人，占据了 90% 甚至趋近 100% 的份额。主持人将台积电比作“半导体界的苹果”（the Apple of semiconductors）——虽然并不占有全部出货量市场份额，却掠走了全行业最暴利的利润池；客户依然能从其他厂商处买到前几代或较差的技术，但最前沿的技术独此一家。针对彭博社旗下的 Odd Lots 播客曾讨论过看空台积电的论点（即处理器性能已经足够强大，未来不再需要顶尖前沿制程），主持人坚决反驳并认为其极其站不住脚。从特斯拉自动驾驶到苹果设备，顶尖硬件的需求从未停止；软件永远会不断膨胀，以榨干最先进硬件的算力。人们常开玩笑抱怨“苹果故意写代码变慢我的旧电脑”，真相是所有软件工程师在开发时都默认用户拥有市面上最先进的处理器，绝不会去专门为六七年前的老旧笔记本优化性能。诸如谷歌、亚马逊等云巨头也绝不可能对算力满足现状。随后探讨深入到技术底层细节：当下的 5 纳米工艺在某种意义上已经变成了一个“营销术语”（marketing name），不再指代晶体管栅极的物理长度，性能提升也不再单纯依靠缩小栅极。更关键的挑战在于物理极限：晶体管之间的距离已经无法再无限拉近。1 纳米的间隙中仅能容纳大约 5 个硅原子，进入 3 纳米等更微观尺度后，硅材料本身已无法再无限细分。因此，未来的行业创新要么彻底替换底层基底材料，要么从多核架构（multicore）及系统级封装等架构层面寻找突破，让芯片以更快速度处理更多任务。这种探索将变得越来越艰难且昂贵，而台积电几乎是全球唯一有能力在此类前沿阵地上持续跟进的芯片制造企业。 --- ### [01:50:43](https://www.youtube.com/watch?v=tuB3_fJyC5Q&t=6643s) 摩尔第二定律（洛克定律）与资本壁垒垄断在众所周知的摩尔定律之外，半导体行业还存在一条“摩尔第二定律”，即由著名风险投资先驱阿瑟·洛克（Arthur Rock）提出的“洛克定律”（Rock's Law）。该定律指出：半导体芯片制造晶圆厂（Fab）的建造成本大约每四年翻一番。现实中，如今建造一座先进晶圆厂已飙升至 150 亿到 200 亿美元。台积电的资本支出（CapEx）甚至在两三年内从 170 亿美元猛增至 300 亿乃至 600 亿美元，实际资本消耗速度远超四年翻倍的传统假设。 ```mermaid flowchart TD accTitle: 摩尔定律与洛克定律形成赢家通吃垄断 accDescr: 展示制程缩减与高额资本支出协同作用，最终导致台积电赢家通吃局面的飞轮模型。 subgraph CostDynamics ["成本与规模双重飞轮"] ML["摩尔定律（规模生产下的运营成本曲线）"] RL["洛克定律（晶圆厂前期天量资本支出 CapEx）"] end subgraph ScaleExecution ["全球超大规模制造能力"] FC["巨额资本转化为专用固定成本与最先进制程"] OP["巨量产能摊销固定成本 维持约 50% 高毛利"] end Monopoly["赢家通吃垄断 （成为不可替代的 Pure Play 巨头）"] ML --> OP RL --> FC FC --> OP OP --> Monopoly Monopoly -.->|"巨额利润再投资"| RL ``` 这两条定律的结合导向了一个不可逆转的经济学结论：半导体制造行业注定演化为一个“赢家通吃”（winner-take-all）的高利润垄断市场。摩尔定律本质上决定了大规模量产时的运营成本（OpEx），而洛克定律决定了使生产成为可能的前期资本支出（CapEx）。这正是商业竞争中极致的固定成本摊销逻辑——通过全球最大规模的产出，将庞大的前期投资稀释为单位优势。风险投资正是为了满足这种巨额固定成本、高杠杆（带来后续高额利润率）的半导体生产而诞生的产物。尽管互联网软件行业后来展现了高达 80% 到 90% 的毛利率（相比台积电 50% 的毛利率更高），但软件公司缺乏像台积电这样通过把数十数百亿美元资本持续投入硬核制造设备所构筑的超级护城河。 --- ### [01:53:01](https://www.youtube.com/watch?v=tuB3_fJyC5Q&t=6781s) 七大战略维度解析台积电的商业护城河随后主持人运用汉密尔顿·赫尔默（Hamilton Helmer）在商业著作《7 Powers》（七大战略维度）中的框架，全面剖析台积电长期维持超额利润的防御性壁垒。 * **逆向定位（Counter-Positioning）**：台积电创立之初作为纯代工厂（Pure Play Foundry），当时的垂直整合制造巨头（IDM）无法模仿其模式。传统 IDM 的利润核心在于设计与制造一体化的高毛利，且早期无晶圆厂设计公司租用 IDM 闲置产能时，IDM 往往仗势要求将产品打上自身品牌进行销售。张忠谋以代工立誓“不与客户竞争”，IDM 巨头因商业模式冲突无法跟进。 * **规模经济（Scale Economies）**：与制程能力并列为台积电最强劲的两大支柱之一，庞大出货量分摊了先进制程数以百亿计的固定折旧成本。 * **转换成本（Switching Costs）**：客户脱离台积电的代价高得无法想象。若一家公司想要生产顶尖手机芯片，除了台积电别无选择；即使转向格芯（GlobalFoundries）等少数竞争对手，也无法简单做到“周五打个电话、下周一交付晶圆”，芯片设计与台积电复杂的底层制造工艺流程深度绑定，迁移产线需要耗费数年研发与调校时间。 * **网络效应（Network Economies）与生态效应**：半导体代工不具备类似社交网络的传统网络效应（甚至像苹果包揽 5 纳米全部初期产能，客观上挤压了其他客户的配额）。但台积电构建了庞大的开放创新生态（OIP，Open Innovation Platform），EDA 工具开发商与半导体 IP 提供商（如 ARM）都以台积电的制程规范作为第一优先级进行深度协同与优化，形成了独特的产业生态引力。 * **流程权力（Process Power）**：在随后的分析中，主持人认定台积电具备甚至超越皮克斯级别的“流程权力”（Process Power），其制造工艺的良率控制、运营协同与工程组织能力，是其他对手砸钱也买不来的深厚组织资产。 ### [01:57:20](https://www.youtube.com/watch?v=tuB3_fJyC5Q&t=7040s) 流程优势的终极壁垒与潜在颠覆风险在分析 Hamilton Helmer 的“七大力量”（7 Powers）时，台积电展现出商业史上最清晰、最极致的“流程优势”（Process Power）。台积电如今所处的技术巅峰和良率护城河，凝结了其整整 40 年的技术积累与实践经验。即便此时有十位顶尖高管离职创办一家新企业，也不可能从零复制台积电——其数以万计的核心专利、无可替代的技术工匠与科学家团队、海量工艺 know-how，以及与 ASML（阿斯麦）等设备供应商数十年结成的紧密同盟，是无论投入多少资金都无法在短时间内复刻的。在现有的半导体演化轨道上，台积电几乎没有对手，这场比赛只能是“台积电自己输给自己”。能够真正动摇甚至瓦解台积电优势地位的，唯有类似移动互联网颠覆 PC 桌面生态那样的彻底范式转移（Paradigm Shift）。例如，如果未来人类所需的计算形式发生了本质转变，彻底跳出台积电目前擅长的物理极限（比如量子计算取得突破性进展，使 ASML 极其复杂昂贵的极端紫外线光刻、激光轰击熔融锡液等工艺变得不再必要），整个竞争维度才会被重新洗牌。此外，若行业经历类似云计算的“AWS 时刻”，出现某种能够将制造尖端芯片的技术与认知门槛降低几个数量级的技术通道，才可能剥夺台积电的流程护城河；单纯更换硅基底材料等渐进式演进，台积电完全有能力从容应对。 - **品牌与独占资源（Branding & Cornered Resource）的局限性**：台积电的商业模式本质上与终端品牌建设背道而驰；终端消费者认同的是 Apple、Nvidia 等品牌，而非台积电本身。 - **流程优势不可复制性**：即便资本毫无上限，缺乏数十年的试错演化与产业协同网络，也无法凭空再造一个台积电。 ### [01:59:08](https://www.youtube.com/watch?v=tuB3_fJyC5Q&t=7148s) 终极“看跌情形”：地缘政治风暴与抗力量风险剥离了计算范式转移的假设后，台积电面临的唯一、且最具毁灭性的下行风险（Bear Case）来自于地缘政治。作为位于地缘争议岛屿上的企业，台湾海峡的局势始终是一柄悬顶之剑。如果中国大陆采取果断行动统一台湾或行使对其主权，假设这种剧变并未引发世界大战，全球芯片供应链也将被立刻切断重构。西方企业依赖的尖端产能将直接转向中国本土企业，包括此前被美国制裁而无法在台积电代工的华为等。这种地缘政治宿命在战略分析框架中很难被简单定义为某种“护城河”，它更像是一种“抗力量”（Anti-Power）或系统性脆弱性。早期台积电确实因台湾当局的全力扶植而具备特殊优势——当年台当局如同强硬的商业推手，勒令本土商业巨头出资入股，提供一切政策与用地资源保证台积电落地；然而发展到今天，全球对半导体芯片的极度依赖与台积电产能高度集中在单一地理位置的现实，已经让整个现代工业体系如履薄冰。 ```mermaid flowchart TD accTitle: TSMC Geopolitical Risk Flow accDescr: 展现台积电从极度集中的生产体系到地缘政治危机带来的全球冲击传导机制 subgraph Concentration["台积电核心竞争力与现实"] A["40年积累的流程优势 (Process Power)"] --> B["超90%先进先进制程集聚台湾"] end subgraph Trigger["外部地缘冲击"] C["地缘冲突 / 台海危机"] end subgraph Impact["全球连锁反应"] D["全球先进制程芯片断供"] E["汽车制造、消费电子与航空业停摆"] F["美国/盟友试图产能外迁与技术空运面临水土不服"] end B -. 暴露于脆弱节点.-> C C --> D D --> E C --> F ``` ### [02:01:24](https://www.youtube.com/watch?v=tuB3_fJyC5Q&t=7284s) 海外建厂与能否“空运”流程优势的世纪难题由于面临各方巨大的供应链韧性压力，台积电启动了一定程度的跨国建厂布局，包括在美国亚利桑那州投资 120 亿美元新建晶圆厂、在日本熊本建设晶圆厂，以及在中国大陆保留的部分运营设施。然而，这些海外工厂大多获得了当地政府的巨额补贴（如拜登政府推动的半导体芯片法案），且引入的均非最顶尖的 3 纳米甚至 5 纳米先进制程。张忠谋曾公开直言，在商业逻辑上，将先进制程移出台湾毫无道理；台积电的极致效率深深依赖于台湾本土数十年磨合出来的完整上下游配套生态系统。面对极端情况，业界存在一个残酷的思维实验：假如台海发生冲突，台积电的“流程优势”是否可以通过人员疏散、将 ASML 设备移交别处的方式“物理空运”到其他国家复刻？历史经验对此并不乐观。以汽车工业为例，丰田（Toyota）当年曾试图与通用汽车（GM）在加州佛里蒙特建立 NUMMI 合资工厂，向美国输出精益生产系统（TPS），但受制于文化与现场管理，始终未能真正重现其日本本土的威力（该厂址后来转手成为特斯拉的工厂）。在战略防卫层面，现代社会乃至波音等国防巨头的战机建造都必须依赖先进制程半导体；如果失去台湾的晶圆制造，全球几乎所有技术进步都将面临休克。 - **亚利桑那州工厂投资**：耗资 120 亿美元，但其制程落后于台湾本土主力先进制程，象征意义与试验意义大于直接替代本土先进制程。 - **战略资产的对比思考**：过去美国将波音等重工业视为必须本土掌控的立国之本，但在智能化时代，先进制程半导体制造对国家命脉的决定性已全面超越传统军工载具。 ### [02:05:04](https://www.youtube.com/watch?v=tuB3_fJyC5Q&t=7504s) “大乱斗”之下的全球捆绑与《彭博社》未竟的历史秘辛整个半导体产业链展现出高度纠缠的全球化特征。各方利益深度捆绑：全球繁荣基于分工协作，只要没有一方彻底打破规则，体系就能持续运转；这类似于韩国与三星电子长期生活在朝鲜三八线火炮威胁之下的常态。虽然近一年来地缘局势的火药味显著升温，但在缺乏外部统一冲击（比如发现外星文明降临使全人类纷争显得微不足道）的前提下，全球商业界只能寄希望于这种脆弱的平衡能维持下去。在进入商业战略剖析（Playbook）阶段之前，节目重新翻开了一个至关重要的历史“假想时刻”（What-If Scenario）。援引《彭博商业周刊》（Bloomberg）的独家披露：在 2000 年代中期，当苹果公司（Apple）紧锣密鼓地筹备推出第一代颠覆性智能手机 iPhone 时，史蒂夫·乔布斯（Steve Jobs）曾亲自找到时任英特尔（Intel）CEO 的保罗·欧德宁（Paul Otellini），力邀英特尔为尚未发布的 iPhone 开发并供应核心处理器芯片…… ### [02:06:45](https://www.youtube.com/watch?v=tuB3_fJyC5Q&t=7605s) 战略决策的致命陷阱：Rosenthal 公理 David 抑制不住内心的激动，甚至半开玩笑地要求把播客录像放出来，好让大家看到他兴奋得挥舞双拳的模样。这源于科技史上最令人扼腕的一场错失：当年史蒂夫·乔布斯（Steve Jobs）在 Mac 放弃 PowerPC 架构、全面转向 Intel 处理器时，曾把时任 Intel CEO 的保罗·欧德宁（Paul Otellini）请上台，共同宣布这是计算的未来与坚不可摧的盟约。然而，当乔布斯随后带着初代 iPhone 芯片的代工意向找到这位亲密伙伴时，欧德宁却认为苹果给出的报价过低（lowball offer），直接拒绝了这一合同。结果，苹果转头将订单交给了三星，紧接着开启了轰轰烈烈的芯片自研之路，并最终将所有前沿先进制程代工合同尽数交给了台积电。这一戏剧性的历史转折引发了主播们对商业博弈底层的思考：Intel 拒绝苹果，并不是因为技术路线之争（比如嫌弃 ARM 架构），纯粹是因为觉得价格不够优厚，这堪称“史上最严重的商业战略失误”。David 据此提出了一个如同几何学公理般的核心商业法则——“Rosenthal 公理”（Rosenthal Postulate）：**永远不要仅仅基于价格或估值（pricing）去否决重大的战略决策**。这种“因小失大”的人类天性在科技史上屡见不鲜： - 风险投资机构因纠结于估值而错失世纪巨头，例如 Andreessen Horowitz（a16z）当年在 Uber 估值达到 3 亿美元时被吓退、不敢跟进； - 福特汽车（Ford）在早年间仅仅因为区区 1 美元的薪资分歧，就放弃了聘用半导体传奇人物张忠谋； - Intel 在自身架构能够捕获巨大上升空间的前提下，仅仅因定价数字偏低而把划时代的移动芯片合同拒之门外。 ```mermaid flowchart LR accTitle: Apple Chip Sourcing Evolution accDescr: 苹果从早期找 Intel 碰壁，到三星代工，再到自研并全面由台积电制造的历程。 subgraph S1["1. 错失良机"] A["苹果寻求 iPhone 芯片合作"] -->|"低价合同被拒"| B["Intel 拒绝代工"] end subgraph S2["2. 转折与自研"] A -->|"早期代工"| C["三星生产初代芯片"] C --> D["苹果启动自研 Apple Silicon"] end subgraph S3["3. 终极赢家"] D -->|"全外包先进制程"| E["台积电 (TSMC) 独家制造"] end ``` ### [02:09:04](https://www.youtube.com/watch?v=tuB3_fJyC5Q&t=7744s) 走向平庸的漫长过程与 ASML 的市值逆袭在商业竞争中，一家庞大巨头的衰退往往需要经历极为漫长的时间（"It takes a very long time to become irrelevant"）。尽管 Intel 在二级市场已经失宠——彼时台积电的市值已达到 Intel 的约 2.5 倍，就连半导体供应链上游、仅向极少数公司提供单一设备的光刻机巨头 ASML（阿斯麦），市值也已经超过了 Intel，跃居为欧洲市值最高的科技上市公司。两位主播向收听节目的公开市场投资者抛出话题：如果看好半导体先进制程这一长线逻辑，大家究竟是如何在台积电与 ASML 之间分配筹码的？是否也在通过做空 Intel 来构建对冲组合？然而，Intel 绝非毫无还手之力的空壳。时至今日，Intel 依然牢牢把控着全球个人电脑处理器 80% 的市场份额，在数据中心服务器领域的占有率甚至更高。世界上绝大多数在云端运行的企业级工作负载，根本不是特斯拉 Dojo 那种需要极限制程的尖端 AI 训练，而只是普通跨国公司在 Office 365 上运行的日常 Outlook 邮件服务。这些庞大的基本盘完全不需要 5 纳米或 3 纳米先进制程，却源源不断地为 Intel 输送着巨额现金流。 - **市场对比**：ASML 作为半导体光刻环节独家供应商，凭借 EUV 设备的绝对垄断，市值成功超越老牌垂直整合制造霸主 Intel。 - **存量市场优势**：Intel 在 PC 芯片市场依然掌握约 80% 份额，服务器 CPU 垄断优势依然深厚。 - **现实需求分层**：绝大多数传统商业工作负载（如企业云端服务器）不需要最前沿的先进制程，成熟制程依然构筑了极厚的利润护城河。 ### [02:10:40](https://www.youtube.com/watch?v=tuB3_fJyC5Q&t=7840s) 摇摆不定的战略与“创新者的窘境” Intel 真正的致命软肋在于管理层长期的战略犹疑与决策反复。前任 CEO 鲍勃·斯旺（Bob Swan）眼看制造技术掉队，已经开始筹备认输，计划将 Intel 自主设计的先进制程芯片外包给台积电代工；然而当技术老将帕特·基辛格（Pat Gelsinger，曾任 VMware CEO，最早在 Intel 奠定行业声誉，并曾是接任微软 CEO 的热门候选人）回归挂帅后，公司政策彻底调头，宣布全面进军代工业务（IDM 2.0），试图将 Intel 自身打造成一个能接纳外部芯片设计公司的晶圆代工厂。这种在“放弃制造”与“开放制造”之间的剧烈摇摆，不仅未能解套，反而让巨轮越陷越深。更为讽刺的是，Intel 正在上演科技史上最纯粹的“创新者的窘境”（The Innovator's Dilemma），其境遇与当年在实验室率先发明数码相机却不敢商业化推广的柯达如出一辙。早在 2012 年，Intel 就深刻意识到极紫外光刻（EUV）代表着半导体的未来，甚至是 EUV 最早、最大的早期投资人，累计砸下了超过 40 亿美元的研究资金。然而，面对现有光刻工艺带来的丰厚即期回报，Intel 变得举棋不定，认为改良现有技术路线是更稳妥的下注，从而放缓了拥抱新技术的步伐。 - **战略摇摆的恶果**：从外包台积电到自主大举建厂做代工，反复的路线调头让 Intel 错失调整窗口。 - **EUV 投资的反讽**：Intel 最早押注并巨资注资 EUV 光刻，但截至讨论当时，Intel 自身甚至没有量产出货过任何一款由自身工厂使用 EUV 制造的芯片。 - **创业公司的价值**：在既得利益和成熟业务面前，大公司天然难以向自身开刀，这也从侧面印证了颠覆性技术为什么往往需要通过初创公司来彻底释放。 ### [02:14:12](https://www.youtube.com/watch?v=tuB3_fJyC5Q&t=8052s) 幸存者偏差与“专注核心能力” 从商业方法论的角度，David 深入剖析了张忠谋创办台积电时的底层逻辑：如果仅从最终发生的辉煌结果出发，人们根本无法客观反推其最初成功的概率。张忠谋当年所奉行的“如果你把工厂建好了，客户自然就会来”（"If you build it, they will come"）策略，在常规商业逻辑中其实是一个极度危险的劣质策略。例如红杉资本的传奇投资人唐·瓦伦丁（Don Valentine）就极其厌恶这种做法，红杉的第一铁律永远是“只投资于已经存在的成熟市场，绝不投钱去开拓培育一个不存在的市场”；而许多从西雅图微软离职创业的人，也常常陷入这种自建平台而无人问津的泥潭。从统计学上看，这种在没有现成需求下豪赌先进制程产能的模式，大概率有 90% 的概率会惨败收场；但一旦碰上那 10% 的极小概率契机，它就缔造了如今的代工帝国。这证明在商业世界里，所有的既定规则最终都可以被打破，并不存在一劳永逸的成功公式。最后，主播们重申了杰夫·贝佐斯（Jeff Bezos）关于 AWS 的经典论述——“做任何事都要问，这是否能让你的啤酒尝起来更甘醇”。这源自早期德国啤酒厂自行发电的历史典故：随着公用电网的普及，自建发电机不再是啤酒厂的竞争优势，只会分散精力。无论是做初创公司还是做科技巨头，企业必须把非核心的职能（如财务、后勤甚至非核心制造）彻底外包，将全部兵力聚焦于真正能让产品变得无可替代的本质能力上。 ### [02:16:08](https://www.youtube.com/watch?v=tuB3_fJyC5Q&t=8168s) 评级思考：台积电在科技万神殿中位列何处？节目的评级环节并未沿用传统的打分模式。如果只是对台湾当年出资并持有近 50% 股权的决策打分，毫无悬念会是“毫无悬念且无趣的 A+”。因此，两位主持人展开了一场思维实验：在有史以来所有伟大科技公司的“万神殿”中，台积电究竟应该排在什么位置？它是 FAANG 级别的巨头，还是能跻身历史前五、前十或前二十？在整个人类半导体产业的五层价值链中，该如何定义台积电所拥有的统治级力量？ ```mermaid flowchart TD accTitle: TSMC Scale and Moat Flywheel accDescr: 台积电通过规模与制造能力构建的极深飞轮护城河 subgraph Design["顶层创新与设计生态"] A["无晶圆厂设计商与巨头 (Apple, Nvidia, AMD)"] B["前沿计算需求 (AI/ML, 5G, 加密运算)"] end subgraph Foundry["台积电代工平台核心"] C["台积电先进制程制造 (TSMC Advanced Nodes)"] D["计算成本渐进趋零 (Compute Approaches Zero)"] E["庞大营收与自由现金流 (Reinvestment / CapEx)"] end B --> A A -->|"交由制造"| C C --> D C --> E E -->|"巨额研发与扩产"| C ``` 一种参照系是将它与 2000 年巅峰时期的英特尔相比，但台积电当前代工的产品规模是当年英特尔的百倍以上。放眼当今每一次令全行业沸腾的计算浪潮——无论是机器学习（AI/ML）、加密货币还是 5G，背后无一例外都依赖台积电制造的芯片。当马克·安德森（Marc Andreessen）说出著名的“软件正在吞噬世界”时，现实中正是因为台积电将半导体制造的边际成本压得极低，才让算力成本在渐进意义上接近于零，使得任何软件都能在这些硅片上肆意生长。虽然很难将功劳严格在台积电、阿斯麦（ASML）以及以苹果内部 600 人芯片设计团队为代表的顶级设计商之间清晰切分，但两位主持人一致认为：台积电毫无疑问是有史以来最成功、最重要的 B2B 硬件企业。 - **平台型公司的底层逻辑**：任何时候当你看到大量企业在处理与自身核心产品不直接相关的重资产苦活（“无法让啤酒味道更好”的无效重复劳动）时，那就是建立平台型巨头的巨大机遇。 - **产业赋能广度**：台积电的客户涵盖从英伟达、苹果到高通、AMD 乃至英特尔自身的所有核心算力主导者。 - **算力成本曲线**：台积电以一己之力推动了全球算力成本大幅下降，成为现代数字经济与软件创新真正的物理底座。 ### [02:18:46](https://www.youtube.com/watch?v=tuB3_fJyC5Q&t=8326s) 护城河与地缘风险：超越英特尔的B2B硬件巅峰断言台积电已经超越英特尔是一个极具分量的结论。英特尔不仅定义了硅谷，代表了“叛逆八人组”的历史传承，更是摩尔定律几十年的掌旗者。然而在复利规律下，所有的超额价值都会集中在末期爆发：台积电在随后两三年内创造的经济价值，很可能超过此前整个半导体工业历史的总和。台积电在庞大体量下依然能实现单年 30% 的营收爆发式增长，这是传统时期的英特尔完全无法企及的奇迹。在对比苹果、微软、谷歌、亚马逊和 Facebook 等科技巨头时，虽然这些消费级霸主在纯粹的终端价值创造上依然庞大，但如果作为“反方辩友”，完全可以论证：缺少了台积电，这些公司现有的所有硬件与云端创新根本无法成立。除非代工制造（Foundry）与无晶圆厂（Fabless）模式在历史演进中是不可避免的宿命，否则台积电就是独一无二的基石。在护城河的深度上，台积电几乎坚不可摧。大科技公司即便强如微软也曾经历低谷与重启，而台积电凭借其商业模式和动辄百亿级资本开支的高昂门槛，构筑了铜墙铁壁般的防御体系。两位主持人感叹，台积电“除了物理防弹以外，防得住商业世界的一切”——它唯一的深层软肋就是台湾的地缘政治风险。就像美元的信用背后有美军支撑一样，一家企业如果没有军队护卫，它的安全完全建立在各方遵守规则的基础之上。 - **复利终局效应**：半导体产业演进数十年的价值积累，在先进制程向极小纳米逼近的今天，以前所未有的密度向掌握终极良率的台积电倾斜。 - **极致壁垒**：台积电的护城河比传统科技巨头更具结构性排他性，竞对即便投入天量资金也极难在良率与产能上形成替代。 - **地缘脆弱性**：台积电是世界上极少数价值已超越主权国家级保障界限的企业，其最大的系统性风险并非源自市场竞争，而是外部地缘局势。 ### [02:21:10](https://www.youtube.com/watch?v=tuB3_fJyC5Q&t=8470s) 历史对标与终极定论：半导体即新时代的石油如果展望未来 30 年台积电是否依然能称霸，核心在于其当前形成的垄断飞轮：在结构上没有任何现存对手能够追赶，除非出现某种不可预见的外部范式转移。历史上与之最相似的企业是洛克菲勒的标准石油（Standard Oil）——它们都通过结构性设置挤压了所有竞争对手的生存空间，让富者愈富、强行筑高行业门槛，甚至其拆分后的庞大谱系在百年后依然深刻主导着能源产业。关于台积电在公众心智中的排位，两位主持人探讨了“品牌定位”带来的认知偏差：台积电刻意保持低调，不建立面向大众的消费品牌，而是将聚光灯完全让给客户，让客户享受聚光灯下的辉煌，这也导致普通大众对其规模和影响力的感知远低于 FAANG 等公司。经过反复推敲，两人给出了一致的定论： | 评估维度 | 最终讨论评定 | 关键核心依据 | |:--- |:--- |:--- | | **科技万神殿总排位** | **历史前十（Top 10）** | 暂未稳坐历史前五（Top 5），但无可争议地跻身人类科技商业史上排名前十的最伟大公司。 | | **B2B 硬件赛道** | **有史以来第一（No. 1）** | 彻底超越巅峰时期的英特尔，成为有史以来规模最大、壁垒最坚固的纯硬件代工/B2B 企业。 | | **全行业 B2B 横向对比** | **与微软并列争顶** | 唯一能在企业服务和关键基础设施影响力上与微软相提并论的绝对巨头。 | 半导体已经全面替代传统能源，成为支撑人类现代文明运转的“新石油”，而台积电正是这项关键要素的真正掌舵者。 ### [02:23:11](https://www.youtube.com/watch?v=tuB3_fJyC5Q&t=8591s) Carve Outs 分享：《足球教练》、《希腊同盟》与迈克尔·奥维茨自传在漫长的台积电深度复盘之后，节目进入了轻松的保留栏目“Carve Outs（好物推荐）”。David 分享了自己和妻子 Jenny 近期在圣巴巴拉（Santa Barbara）租 Airbnb 度假的惬意经历，借此躲避旧金山寒冷多雾的夏季，并希望将其变成每年夏天的家庭传统。在度假期间，平时极少看电视的两人开启了“刷剧”时光： - **《足球教练》（Ted Lasso，第一季）**：虽然外界普遍认为第二季口碑下滑，但 David 的逆向思维是“正因为大家对第二季期望落空后反应如此强烈，反而证明第一季写得极为出色”。两人一口气看到了第八集，盛赞其温暖治愈的高水准。 - **《希腊同盟》（Greek）**：由 Jenny 提议重温的一部 2000 年代中期的经典美剧，讲述虚构大学中兄弟会和姐妹会的生活。剧集充满了怀旧色彩，恰好贴合了两人当年身处大学校园的时代记忆。 Ben 带来的推荐则是一本多位好友向他推介长达两三年的重磅商业传记——《谁是迈克尔·奥维茨？》（*Who is Michael Ovitz?*）。Ben 将其与耐克创始人菲尔·奈特的《鞋狗》（*Shoe Dog*）、迪士尼前 CEO 鲍勃·艾格的《一生的旅程》（*The Ride of a Lifetime*）以及讲述福特转危为安的《美国偶像》（*American Icon*）并列。迈克尔·奥维茨在 1975 年至 2000 年间创立并带领 CAA（创新精英文化经纪有限公司），彻底颠覆了好莱坞传统的经纪人体系与打包交易模式（Packaging）。CAA 在极少接受媒体采访、对外极度隐秘的情况下重塑了整个好莱坞影视生态。Ben 坦言，此前自己通过《一生的旅程》和流媒体纪录片了解迪士尼历史时，对奥维茨短暂且失败的迪士尼总裁生涯抱有片面先入为主的偏见；读完这本自传后，才真正体会到奥维茨在商业战略、行业重组以及组织掌控力上的深远天才。 ### [02:25:50](https://www.youtube.com/watch?v=tuB3_fJyC5Q&t=8750s) 推荐书目《Who is Michael Ovitz?》与好莱坞打包时代的回忆在节目的个人推荐（Carve Outs）环节尾声，两位主持人探讨了自传《Who is Michael Ovitz?》的精彩之处。这本书与《Acquired》此前做过的多期节目存在着千丝万缕的奇妙联系，特别是关于迪士尼（Disney）以及鲍勃·艾格（Bob Iger）的商业篇章。阅读这本书的契机正是录制那些内容时，了解到了好莱坞顶级经纪公司 CAA（Creative Artists Agency，录音中被误识别为 CIA）是如何建立起统治地位的。对于喜欢商业巨头故事、迪士尼商业史或本身就是电影迷的听众来说，迈克尔·奥维茨（Michael Ovitz）的这本回忆录堪称绝佳读物。这本自传勾起了主持人童年时代的强烈共鸣。两人回忆道，他们成长过程中所看过的无数经典影视作品——无论是小时候年纪太小按理不能看、却千方百计想偷看的成人大片，还是经典的合家欢儿童电影——其幕后运作几乎都有 CAA 的身影。从马丁·斯科塞斯的黑帮经典《好家伙》（Goodfellas，此前节目中曾被推荐过）到风靡全球的《侏罗纪公园》（Jurassic Park），这些影史名作几乎全部是通过奥维茨开创的“项目打包”（Packaging）模式撮合诞生的。深入了解这些划时代作品从构思到落地的幕后商业博弈，让人大呼过瘾。 ### [02:26:33](https://www.youtube.com/watch?v=tuB3_fJyC5Q&t=8793s) 播客的口碑传播与节目收尾在正片正式结束前，主持人向听众发出社群邀请，欢迎大家加入官方 Slack 社群（`acquired.fm/slack`），与社群中“才华横溢、聪明伶俐且颜值超高”的听友们互动交流。随后，话题转向了播客这门媒介独特的传播机制。主持人感慨道，播客是一个非常奇特的媒介形态——它本身完全没有内置的“病毒式传播回路”（Viral Loop），不像其他社交产品那样能够自动裂变。尽管在社交媒体上广而告之非常受欢迎，但播客真正赖以生存和壮大的机制极其纯粹，那就是“口耳相传”（Word of Mouth）。听众听完一期节目后觉得很有收获、很有趣味，然后主动分享给身边的朋友，推荐对方也来听听看。主持人诚挚呼吁，如果大家喜欢这期台积电（TSMC）的深度长节目，或者有任何赞同与分歧的观点，都请大胆分享给身边的朋友，或者直接联系节目组表达看法。在热烈的致意声中，本期重制版台积电史诗特辑在标志性的片尾曲“Who got the truth... is it you?”中圆满落幕。 ## Notable quotes > [07:40](https://www.youtube.com/watch?v=tuB3_fJyC5Q&t=460s) 在 50 年代初的美国，华人的出路只有洗衣工、中餐馆老板、工程师以及大学教授。对华人而言，只有这几个算得上受人尊重的职业。这里没有华人律师，没有华人会计师，更不可能有华人政客。 — 张忠谋 > [14:17](https://www.youtube.com/watch?v=tuB3_fJyC5Q&t=857s) 在希凡尼亚，凡是我们能造出来的东西，我们都卖不掉；凡是我们能卖出去的东西，我们都造不出来。 — 希凡尼亚高管 > [18:41](https://www.youtube.com/watch?v=tuB3_fJyC5Q&t=1121s) 当时我们在发明它时并未完全意识到，集成电路将把电子功能的成本降低到原本的一百万分之一。 — 杰克·基尔比 > [31:20](https://www.youtube.com/watch?v=tuB3_fJyC5Q&t=1880s) 那是在 60 年代末，当时的波士顿咨询公司（BCG）还只是一家很小的机构。我们结合了海量数据、理论模型并付出了巨大心血。最终的成果就是所谓的‘学习曲线定价法’：起步就定低价，随后每个季度主动且连续地降价，哪怕市场根本没有提出降价要求。这项举措非常成功，尽管在当时引发了巨大争议——很多人觉得我们疯了：‘市场又没逼你降价，你为什么平白无故主动降价？’但我们坚持做下去，因为我们笃信这个底层逻辑。结果，我们的市场份额持续狂飙。 — 张忠谋 > [36:17](https://www.youtube.com/watch?v=tuB3_fJyC5Q&t=2177s) 当时的德州仪器董事长兼 CEO 马克·谢泼德（Mark Shepherd）深信当时盛行的一套管理信条，即‘一个出色的管理者能管好任何业务’。但在这次实践中，我认为他完全错了。我发现消费品业务截然不同：客户群体完全不同，市场逻辑完全不同，要想脱颖而出所需要的核心能力也完全不同。在半导体业务中，本质就是技术与成本；而在消费市场，技术虽有帮助，但更核心的是迎合消费者的喜好——那是一种极其玄妙、难以捉摸的东西。 — 张忠谋 > [45:13](https://www.youtube.com/watch?v=tuB3_fJyC5Q&t=2713s) 我们在研发上毫无优势，在电路设计与产品设计上毫无优势，在市场营销上毫无优势，在知识产权上更是几乎一无所有。台湾唯一可能具备的优势——而且还只是一种潜在优势——就是半导体制造、晶圆制造。那么，什么样的公司能够匹配这一优势并避开所有劣势？答案就是纯晶圆代工厂（Pure-play Foundry）。 — 张忠谋 > [56:09](https://www.youtube.com/watch?v=tuB3_fJyC5Q&t=3369s) 真男人才拥有晶圆厂 — 杰里·桑德斯 > [59:02](https://www.youtube.com/watch?v=tuB3_fJyC5Q&t=3542s) 在德州仪器和通用仪器时，我看到很多集成电路设计师想出来自立门户，但阻碍他们的最大障碍，就是他们筹集不到足够的钱来创办公司……因为在当时，大家都认为每家半导体公司都必须拥有晶圆制造能力，而制造是资本最密集的部分。我看到那些想离开的人，被无法筹集巨资建造晶圆厂的现实所阻拦。 — 张忠谋 > [01:00:57](https://www.youtube.com/watch?v=tuB3_fJyC5Q&t=3657s) Morris，我们很尊敬你，但这事行不通。 — 德州仪器/英特尔高管 > [01:19:57](https://www.youtube.com/watch?v=tuB3_fJyC5Q&t=4797s) 半导体行业就像一台不断加速的跑步机，如果你跟不上，就会掉下去。 — 张忠谋 > [01:38:21](https://www.youtube.com/watch?v=tuB3_fJyC5Q&t=5901s) 如果我们重注台积电，就没有任何备用方案 — 杰夫·威廉姆斯 > [02:06:45](https://www.youtube.com/watch?v=tuB3_fJyC5Q&t=7605s) 永远不要仅仅基于价格或估值（pricing）去否决重大的战略决策。 — David Rosenthal ## Mentioned 台积电 (TSMC), 张忠谋 (Morris Chang), 英伟达 (Nvidia), 苹果 (Apple), 英特尔 (Intel), 德州仪器 (Texas Instruments), 阿斯麦 (ASML), 三星 (Samsung), AMD, 高通 (Qualcomm), 博通 (Broadcom), 联发科 (MediaTek), 联华电子 (UMC), 工业技术研究院 (ITRI), 李国鼎, 杰克·基尔比 (Jack Kilby), 罗伯特·诺伊斯 (Robert Noyce), 唐·瓦伦丁 (Don Valentine), 杰里·桑德斯 (Jerry Sanders), 飞利浦 (Philips), ARM, P.A. Semi, 蔡力行 (Rick Tsai), 保罗·欧德宁 (Paul Otellini), 帕特·基辛格 (Pat Gelsinger), 迈克尔·奥维茨 (Michael Ovitz), 《谁是迈克尔·奥维茨？》 (Who is Michael Ovitz?), 《足球教练》 (Ted Lasso), 《希腊同盟》 (Greek), Vanta, StatSig, J.P. Morgan ## Video description We dive into the unbelievable and unlikely history behind the quietest technology giant of them all: the Taiwan Semiconductor Manufacturing Company. Founded in 1987 by the then-56 year old Morris Chang, already a legend in the semiconductor industry by virtue of his meteoric rise and fall at Texas Instruments, TSMC today manufactures nearly all the leading-edge chips for Nvidia, Apple, Broadcom, Qualcomm, AMD, and yes — even Intel. Tune in for an incredible story of innovation, perseverance and lasers. Lots and lots of lasers! Note: this is a remastered version of our original 2021 episode. We don’t often re-release old episodes, but in this case we have a very timely reason for doing so. Stay tuned! :) *Links:* - Episode Sources: https://www.acquired.fm/episodes/tsmc#sources *Carve Outs:* - Ted Lasso (Season 1): https://tv.apple.com/us/show/ted-lasso/umc.cmc.vtoh0mn0xn7t3c643xqonfzy - Greek: https://www.imdb.com/title/tt0976014/ - Who is Michael Ovitz?: https://www.amazon.com/dp/B07B2HS77M/ *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=tuB3_fJyC5Q)

Source quotes

[7:40] 张忠谋: 在 50 年代初的美国，华人的出路只有洗衣工、中餐馆老板、工程师以及大学教授。对华人而言，只有这几个算得上受人尊重的职业。这里没有华人律师，没有华人会计师，更不可能有华人政客。 [14:17] 希凡尼亚高管: 在希凡尼亚，凡是我们能造出来的东西，我们都卖不掉；凡是我们能卖出去的东西，我们都造不出来。 [18:41] 杰克·基尔比: 当时我们在发明它时并未完全意识到，集成电路将把电子功能的成本降低到原本的一百万分之一。 [31:20] 张忠谋: 那是在 60 年代末，当时的波士顿咨询公司（BCG）还只是一家很小的机构。我们结合了海量数据、理论模型并付出了巨大心血。最终的成果就是所谓的‘学习曲线定价法’：起步就定低价，随后每个季度主动且连续地降价，哪怕市场根本没有提出降价要求。这项举措非常成功，尽管在当时引发了巨大争议——很多人觉得我们疯了：‘市场又没逼你降价，你为什么平白无故主动降价？’但我们坚持做下去，因为我们笃信这个底层逻辑。结果，我们的市场份额持续狂飙。 [36:17] 张忠谋: 当时的德州仪器董事长兼 CEO 马克·谢泼德（Mark Shepherd）深信当时盛行的一套管理信条，即‘一个出色的管理者能管好任何业务’。但在这次实践中，我认为他完全错了。我发现消费品业务截然不同：客户群体完全不同，市场逻辑完全不同，要想脱颖而出所需要的核心能力也完全不同。在半导体业务中，本质就是技术与成本；而在消费市场，技术虽有帮助，但更核心的是迎合消费者的喜好——那是一种极其玄妙、难以捉摸的东西。 [45:13] 张忠谋: 我们在研发上毫无优势，在电路设计与产品设计上毫无优势，在市场营销上毫无优势，在知识产权上更是几乎一无所有。台湾唯一可能具备的优势——而且还只是一种潜在优势——就是半导体制造、晶圆制造。那么，什么样的公司能够匹配这一优势并避开所有劣势？答案就是纯晶圆代工厂（Pure-play Foundry）。 [56:09] 杰里·桑德斯: 真男人才拥有晶圆厂 [59:02] 张忠谋: 在德州仪器和通用仪器时，我看到很多集成电路设计师想出来自立门户，但阻碍他们的最大障碍，就是他们筹集不到足够的钱来创办公司……因为在当时，大家都认为每家半导体公司都必须拥有晶圆制造能力，而制造是资本最密集的部分。我看到那些想离开的人，被无法筹集巨资建造晶圆厂的现实所阻拦。 [60:57] 德州仪器/英特尔高管: Morris，我们很尊敬你，但这事行不通。 [79:57] 张忠谋: 半导体行业就像一台不断加速的跑步机，如果你跟不上，就会掉下去。 [98:21] 杰夫·威廉姆斯: 如果我们重注台积电，就没有任何备用方案 [126:45] David Rosenthal: 永远不要仅仅基于价格或估值（pricing）去否决重大的战略决策。

Source quotes

[0:00] David Rosenthal: essentially the only manufacturer of the Leading Edge chips [7:40] Morris Chang: Chinese laundrymen, Chinese restaurant owners, Chinese engineers, and Chinese professors [17:44] We at Sylvania cannot make what we can sell, and we cannot sell what we can make [17:44] Jack Kilby: reduce the cost of electronic functions by a factor of a million to one [36:17] Morris Chang: put out to pasture [54:07] Morris Chang: it was like in the movie The Godfather, it was an offer I couldn't refuse [56:10] Jerry Sanders: real men have Fabs [73:55] Morris Chang: The semiconductor business is like a treadmill that speeds up all the time. If you can't keep up, you fall off. [98:28] Jeff Williams: no backup plan [126:11] biggest strategic error of all time [126:11] never make strategic decisions based on pricing rather than long-term economics [131:54] doesn't make your beer taste better [134:30] When Mark Andreessen says software is eating the world, it's only eating the world because TSMC has made it so freaking cheap to manufacture silicon.

Original transcript

[0:00] hello acquired listeners we regularly [0:03] get feedback that this episode on tsmc [0:06] the Taiwan semiconductor manufacturing [0:08] company is one of the best acquired [0:11] episodes ever and interestingly it [0:13] predates our Nvidia episodes we did it [0:16] way back in 2021 when the acquired [0:19] audience was about 12% the size of what [0:22] it is today which means that the vast [0:24] majority of you have never heard it so [0:26] uh we definitely wanted to fix that [0:28] since then semiconductors have become so [0:30] much more important in our world and [0:32] tsmc has essentially become the only [0:35] manufacturer of the Leading Edge chips [0:38] they make the primary chip inside every [0:41] MacBook and iPhone shipped today they're [0:43] powering the AI wave manufacturing all [0:45] of nvidia's chips they make the chips [0:47] for a whole bunch of other fabulous [0:49] companies like Qualcomm AMD broadcom and [0:53] hyperscalers like AWS and it turns out [0:56] they even manufacture a lot of chips for [0:59] Intel too yes little known fact yeah [1:02] tsmc rode the smartphone era to Crazy [1:04] Heights as we all know and here now in [1:07] the next AI era here in 2025 it turns [1:10] out that they are the manufacturing [1:12] superpower behind all of that too yep [1:15] well listeners without doing too much [1:17] foreshadowing now is a very good time [1:20] for anyone to listen or relisten to the [1:22] tsmc episode so we decided we should go [1:25] all the way back to the raw audio tracks [1:27] and remaster this whole thing from [1:29] scratch [1:30] for your listening pleasure Ben in fact [1:33] I looked it up since we were going back [1:34] to 2021 when we initially recorded this [1:37] tsmc's market cap has doubled since then [1:40] from 550 billion to over a trillion [1:44] dollar and in fact you were the one that [1:46] tipped me off to this as we were re- [1:48] researching here they and Saudia ramco [1:51] are the only trillion dollar companies [1:53] in the world that are not located on the [1:56] west coast of the United States wild [1:58] this is such a crazy stat it's crazy [2:00] that the rest are located on the west [2:02] coast of the United States but it really [2:03] underscores what a extreme outlier tsmc [2:06] is so without further Ado the story is [2:09] truly unbelievable and we hope you enjoy [2:11] this presentation of tsmc remastered who [2:14] got the [2:16] truth is it you is it you is it you who [2:19] got the truth [2:21] now is it you is it you is it you sit it [2:25] down say it straight another story on [2:28] the way got the truth welcome to season [2:32] 9 episode 3 of acquired the podcast [2:35] about great technology companies and the [2:37] stories and playbooks behind them I am [2:39] Ben Gilbert and I'm David Rosenthal and [2:42] we are your hosts today's episode is on [2:45] tsmc or the Taiwan Semiconductor Company [2:49] is your classic most people have never [2:51] heard of it but it's the ninth largest [2:53] company in the world episode this is a [2:55] wild Morris Chang founded tsmc at age 56 [2:59] retired at 74 then came back at age 78 [3:04] ink to the deal to make all of Apple's [3:06] chips and yeah we're going to tell the [3:08] whole story here it's wild it's nuts [3:11] they make literally every chip in every [3:13] iPhone sold today and soon to be in [3:15] every Mac sold if you're excited at all [3:18] about Nvidia AMD qualcom or even any of [3:21] the chips that Amazon Microsoft Facebook [3:23] Apple are making all of those chips or [3:26] nearly all of them are actually made by [3:28] tsmc along with all the in your cars and [3:30] your smart home devices and fighter jets [3:33] and everything unbelievably this company [3:36] that the entire world relies on is on an [3:39] island that some countries feel is a [3:41] Sovereign Nation and the People's [3:43] Republic of China feels is actually [3:46] theirs so today's episode has it all [3:48] ascending from startup to Tech [3:50] superpower an underdog founder and of [3:53] course a good dose of [3:55] geopolitics indeed all right well [3:58] listeners it finally felt like the right [4:01] time to do this episode amidst This [4:03] Global chip shortage that we've got [4:05] going on that David I think I've heard [4:07] even Ford has paused the production of [4:09] F-150s because of this so it is like a [4:12] massive impact on the world and we've [4:13] had tsmc on the agenda to do for like [4:16] two and a half years now on our little [4:17] Google doc totally well I feel like we [4:20] haven't called it a miniseries but let's [4:21] call it a miniseries on semiconductors [4:24] and like Silicon the arm episode yep [4:26] Sequoia part one PA semi yep okay [4:30] listeners it is time to jump into the [4:32] history and facts and David's going to [4:34] lead us in that but as usual even though [4:37] we're going to be probably very excited [4:38] about some companies less excited about [4:40] other companies this show is not [4:41] investment advice we may have [4:44] investments in the companies we discuss [4:45] it's for entertainment and informational [4:47] purposes only and you should do all of [4:49] your own research okay speaking of we [4:53] start in Ningbo China in July [4:58] 1931 just about one year after Warren [5:01] Edward Buffett was born in Omaha [5:04] Nebraska and there going to be quite a [5:05] few parallels here as we go through this [5:07] episode but in July 1931 in dingbo China [5:11] our protagonist Dr Morris Chang order of [5:15] the propitious clouds with special Grand [5:18] Cordon which is the highest civilian [5:21] honor that anyone in Taiwan can hold [5:24] sweet so he's like a knight of Taiwan [5:26] it's the order of propitious clouds and [5:28] I think there's like nine ranks of it [5:30] and the highest is special Grand cordon [5:33] and he's special Grand he's special yes [5:35] he's very special so he was born then [5:38] for those who are unfamiliar with [5:40] Chinese geography Ningbo is a small City [5:43] just a bit south of Shanghai you know [5:45] small it's about 8 million people just [5:48] casual no big deal China skill is [5:49] ridiculous but certainly wasn't 8 [5:51] million people when Morris was born in [5:52] 1931 no but I bet it was still probably [5:56] pretty big but yeah today 8 million [5:58] people crazy so Morris's father was a [6:01] county official and later became a bank [6:04] manager so the family moved around a [6:06] good bit within China as his father was [6:08] transferring for work this is pre [6:10] People's Republic of China this is pre [6:12] World War II this is a very different [6:14] place right the leadership is not [6:16] communist no no no so his early [6:19] childhood years were like middle class [6:22] not wealthy but pretty well to do [6:24] relative to your average Chinese citizen [6:27] then when he was six the second Sino [6:30] Japanese war breaks out and Morris and [6:33] his mom flee the main part of China to [6:37] Hong Kong and they go to live in Hong [6:39] Kong for a few years to escape the Air [6:41] Raids and the fighting and then on [6:44] December 8th [6:47] 1941 3 hours after Pearl Harbor the [6:52] Japanese attack and invade Hong Kong [6:55] Morris talks about this he's like yeah [6:57] everybody knows Pearl Harbor December [6:58] 7th 19 41 what people don't often talk [7:01] about is the same thing happened in Hong [7:03] Kong 3 hours later on the next day so [7:06] they're in Hong Kong so they flee again [7:09] back to China they end up in Shanghai [7:12] this time and they stay there for a few [7:13] years until [7:15] 1948 after World War II is over but [7:18] that's when the Chinese Civil War breaks [7:21] out that would lead to the Chinese [7:23] Communist Revolution and so they flee [7:25] again back to Hong Kong so this is crazy [7:28] Morris before before he turns 18 he has [7:30] lived through three major Wars the [7:33] second signon Japanese War World War II [7:36] and the chines Civil War so the next [7:40] year in [7:42] 1949 which is the same year as the [7:44] establishment of the PRC the People's [7:46] Republic of China Morris turns 18 and [7:49] with the help of an uncle that he has in [7:52] Boston his life completely changes he [7:54] gets accepted to Harvard so he goes to [7:59] the US he goes to college at Harvard wow [8:01] talk about a change of Fate talk about a [8:03] change of Fate a change of scene [8:06] everything Morris says much later my [8:09] reaction entering Harvard was sheer [8:12] ecstasy almost disbelief what a country [8:16] the United States was at its peak in its [8:18] moral leadership and its political [8:20] leadership in terms of democracy and it [8:23] was the richest country in the world not [8:25] to mention stable I mean you could say [8:27] what you want you could count on the [8:29] fact that it's likely that 10 years from [8:31] now whatever economic structure [8:33] political structures exist will continue [8:35] to exist if what you want to do and what [8:38] he ended up doing with his whole life is [8:40] innovate having that stability around [8:42] you and all those structures enable you [8:45] to do that yeah like we just take this [8:47] for granted but it's a good reminder at [8:49] the very least he's probably not going [8:50] to have to flee Boston to continue his [8:53] studies but he does end up fleeing [8:55] Harvard as we'll get into so Morris [8:57] loved it it was like that quote we read [9:00] he was so Overjoyed to be there but he [9:03] realizes he has kind of a new problem in [9:06] America and at Harvard his parents [9:08] aren't coming over he's on his own he's [9:10] got to support himself and make his own [9:14] way and at that time his race is [9:19] probably going to limit his [9:21] opportunities so as he says quote in the [9:24] early 50s in the United States there [9:26] were Chinese laundrymen Chinese [9:28] restaurant Terrors Chinese engineers and [9:30] Chinese professors those were the only [9:32] respectable professions for Chinese no [9:34] lawyers no accountants no politicians [9:37] and what does Harvard turn out lawyers [9:40] sort of accountants maybe politicians [9:42] yes yeah not a lot of certainly Finance [9:44] professionals certainly Finance [9:46] professionals as we will see as we will [9:48] go along Morris is much more than a [9:49] finance professional but Harvard [9:52] actually didn't have an undergrad [9:54] engineering program at the time huh [9:57] that's crazy to think about if you're [9:59] really really focused you're probably [10:02] going to go down the street in Cambridge [10:04] from Harvard to MIT to MIT which Morris [10:08] does so he only Spends His freshman year [10:09] there and then for his sophomore year he [10:12] transfers to MIT so that he can study [10:15] mechanical engineering so Morris our man [10:19] has learned the ways of the world in the [10:20] US he's focused he starts mechanical [10:23] engineering a year behind at MIT he [10:25] finishes both undergrad and his Masters [10:28] in the Romania 3 years and what year is [10:31] this this would have been [10:34] 1951 when he transferred fall of 1951 [10:38] okay so to contextualize what's going on [10:40] in the tech world right now with quotes [10:43] around it because it's not so much a [10:44] world as a very small continent I mean [10:47] you have all of the postor War II [10:50] defense spending that went in [10:52] particularly on the west coast with the [10:53] Innovations from Stanford but has [10:55] Fairchild semiconductor been started yet [10:57] Noe no no so maybe shockly semiconductor [11:00] shly semiconductor was probably just [11:02] getting going but we're probably still [11:03] in vacuum tube like Bell labs to give [11:06] you a sense silicon is years away [11:09] transistors are probably just getting [11:11] going we're not in the integrated [11:12] circuit yet and it's all being done in [11:14] geranium not silicon wow so it's like [11:18] this is OG yep so after he gets his [11:21] Masters in the three years Morris wants [11:23] to stay and do a PhD fully complete his [11:26] Technical Training but he ends up [11:28] failing his qualifying exams twice they [11:31] give you two chances to take and he [11:32] fails twice by the way this a good time [11:34] to say so David and I watched and [11:36] listened to every footage that Morris [11:38] has ever spoken that has been released [11:40] publicly to prepare for this he is very [11:42] funny oh he's G the way he talks about [11:44] this he says that unfortunately the [11:47] biggest impediment to him going forward [11:49] was that he failed the qualifying exam [11:51] but fortunately for him they were kind [11:53] enough to let him take it a second time [11:54] which he also failed yeah and he has [11:56] this really dry clever sense of humor so [11:59] one of the interviews he talks about one [12:00] of the Stamford ones he gets a question [12:02] from the audience about how did he kick [12:05] his smoking habit oh yeah that like the [12:08] questioner is like I know you used to [12:10] smoke how did you finally stop and he's [12:12] like I never stopped I still smoke he's [12:14] like 94 years old and he goes on to make [12:17] the case for why he's a pipe smoker and [12:20] actually even though smoking is hurtful [12:23] to his lungs it's actually beneficial [12:25] for his mental life so he's pretty sure [12:27] it's prolonged his life well he said [12:29] he's delve into the data and pipe [12:31] smokers live longer than nonsmokers [12:34] which I'm sure you can find data support [12:35] that I'm also sure you can find plenty [12:37] of data to refute that but yes this [12:39] gives you a sense of who Morris is okay [12:42] so he's failed his qualifying exams he's [12:43] got to go out and get a job and you know [12:46] not as a PhD he's got to go get a job as [12:49] a super entry level as an engineer I [12:50] mean he has a master's degree but still [12:53] okay so Legend has it he has a couple [12:56] job offers the one he really wants [12:58] remember he's a mechanical engineer and [13:01] this is like super early days of [13:04] Technology it's not really a thing yet [13:06] there was electrical engineering at this [13:07] time right right right yes he could but [13:09] he didn't study electrical engineering [13:11] in terms of where you would want to work [13:12] like it's not really on anybody's radar [13:14] screen especially morus is that you're G [13:16] to go enter the tech industry right so [13:19] he gets his dream job offer from the [13:21] Ford Motor Company oh no way I didn't [13:25] hear that and I'm sure this is [13:26] apocryphal but let's repeat the [13:28] apocryphal story [13:30] hundreds of thousands of people here [13:32] totally so the legend has it that Ford [13:36] offers him a salary of [13:39] $479 a month to go take an entry level [13:42] job and then he has a competing offer [13:44] from sylvania's new semiconductor [13:47] Division and Sylvania I know of this [13:50] company only because my vacuum growing [13:51] up was made by Sylvania yeah oh well [13:53] we're going to talk much more about [13:54] Sylvania in one second this is the [13:56] competing job offer he's considering [13:58] they offer him a salary of [14:00] $480 a month $1 more and Legend has it [14:03] that Morris asked Ford to beat [14:06] sylvania's offer they didn't and so he [14:09] took the Sylvania job offer I'm sure [14:11] that is a 100% apoc yeah but you know [14:15] Morris he's great so speaking of [14:17] Sylvania do you remember or some portion [14:20] of our audience remembers but do you [14:22] been remember who else started their [14:25] career in sylvania's semiconductor [14:27] division right around this exact same [14:31] time we've talked a lot about this in [14:33] this person on the show no Donald T [14:38] Valentine no way yep that's right so he [14:42] started at Sylvania after forom or maybe [14:45] it's was after the military he ended up [14:47] sh well no then he was at Ron and then [14:49] he joined Fairchild Fairchild okay right [14:51] after the trader a left shockly and [14:53] started Fair child you're better at [14:55] remembering these deep details of older [14:57] episodes than I am well I do a lot of [14:59] resch for this [15:00] show and sometimes research includes [15:03] past acquired episodes there you go so [15:06] they didn't overlap Don Valentine and [15:09] they were never in the same place they [15:10] were in different locations and [15:12] different job functions very different [15:14] job functions but they were both I [15:16] believe both at Sylvania amazing at the [15:18] same time crazy so don is out chilling [15:22] in California like we were talking about [15:24] and falling in love with California he's [15:26] playing water polo he's like oh my gosh [15:28] I'm never going to leave this Place [15:29] Morris he's on the grind he gets posted [15:33] as a junior engineer at sylvania's IP [15:36] switch Massachusetts plant not quite the [15:39] same glamour as Don out in Southern [15:42] California so remember Morris is a [15:44] mechanical engineering he doesn't know [15:45] anything about electrical engineering [15:47] but he's working in this new [15:48] semiconductor division so after work [15:51] he's living in a hotel by the way he [15:53] doesn't even get an apartment it's like [15:54] some company sponsored Hotel he goes [15:57] home back to the hotel from work and he [16:00] studies the best textbook that he can [16:02] find about electrical engineering which [16:05] is entitled electrons and holes in [16:07] semiconductors with applications to [16:10] transistor Electronics written just [16:13] recently a couple years before in 1950 [16:15] by William Shockley oh wow yeah Shockley [16:19] and two other guys basically invented [16:22] the I'm not sure it was the first [16:24] transistor but the first transistor of [16:26] the type that everything else would then [16:28] be built upon when they were at Bell [16:31] Labs not too long before this yeah not [16:33] too long at all I mean anyc was vacuum [16:35] tubes and then shle invented the [16:38] transistor and then in a sec we're going [16:40] to talk about the integrated circuit [16:42] that Bob noise and Jack Kilby who are [16:44] going to talk about invented co-invented [16:47] but anyway okay back to this moment in [16:49] time so Morris is just studying this [16:52] shockly textbook in his hotel room but [16:54] like he's not a college he doesn't have [16:55] any teachers he just has the book wow [16:58] but he's very resourceful [16:59] so he figures out that one of the senior [17:01] engineers at the plant is kind of an [17:04] alcoholic and hits up the hotel bar [17:06] almost every night so what Morris does [17:10] is he comes home from work in the early [17:12] evening he studies in his room for a [17:14] couple hours and then later at night [17:16] when the older colleague shows up at the [17:18] bar Morris goes down to the bar not to [17:20] drink but he brings the textbook and he [17:22] asks the guy Quest he's like I don't [17:24] understand this I don't understand that [17:25] like Grill me and he's just like buying [17:27] drinks for his buddy so great incredible [17:30] here's the quote he says later he being [17:32] the older colleague didn't solve all my [17:35] problems but he solved enough so that I [17:37] could move ahead he was my main teacher [17:39] about electrical engineering so great [17:42] wow so this goes on for three years [17:44] morus is like rolling hard he's burning [17:46] the candle at both ends working and at [17:50] the bar but not drinking uh learning but [17:54] as he like is learning the industry [17:55] coming up to speed it becomes pretty [17:57] clear to him that if he really wants to [17:59] go places in this new emerging industry [18:04] svenia not really the right bus to be on [18:07] so to speak and obviously Don Valentine [18:10] figures out the same thing and jumps to [18:12] Ron and then to Fairchild Morris says [18:14] the moment when this crystallized for [18:16] him was there was a talk that a senior [18:18] manager at Sylvania gave at the plant [18:21] and the quote that the senior manager [18:22] said that stuck with Morris for the rest [18:24] of his life was we at syania cannot make [18:28] what we can [18:29] and we cannot sell what we can [18:31] make real great position to be in so [18:34] Morris is like damn I got to get hell [18:36] out of here that's a signal to move on [18:38] if ever heard one totally so like down [18:41] Morris leaves Sania for greener pastures [18:45] however not to California halfway in [18:48] between or to Silicon Valley yep halfway [18:51] in between so we talk a lot about you [18:53] know Fairchild and the trador eight [18:55] Silicon Valley blah blah blah you know [18:58] the place to be here's the secret [19:01] Silicon Valley all marketing the biggest [19:03] semiconductor company in all types [19:05] digital analog everything at that time [19:07] was not in California it was in Dallas [19:10] Texas it was Texas Instruments which of [19:12] course me you many people in our [19:15] generation know of as the people that [19:16] made our graphing calculators in high [19:18] school and college but of course at this [19:20] time I don't even think they had a [19:21] consumer division yet they no no no and [19:23] that that's going to come up later no TI [19:25] was the jugun like now Silicon Valley is [19:28] silicon Valley but then it was yeah okay [19:30] California I don't know West Coast [19:31] whatever like TI was the big incumbent [19:34] they were the Juggernaut TI actually got [19:36] it start I had no idea before doing the [19:38] research here in the 30s you're like how [19:41] did a technology company and a [19:42] Semiconductor Company end up in Dallas [19:44] Texas they started making instruments [19:46] Texas Instruments for measuring seismic [19:51] activity for oil exploration so all the [19:54] oil comp Texas they were like the tsmc [19:57] like the technology provider to oil [19:59] companies and that's what led them into [20:02] Computing and into digital to power that [20:04] business they were huge not just huge in [20:07] terms of the company but they were the [20:09] technology leader so Bob noise like I [20:13] was saying a minute ago is credited you [20:15] know when he was a Fairchild inventing [20:17] the integrated circuit and all that [20:19] while he was the coinventor [20:21] simultaneously it was co-invented by [20:23] Jack Kilby who was at TI and Jack was [20:26] actually the one who got the Nobel Prize [20:29] for inventing the integrated circuit [20:31] Gordon Moore who was also you know a [20:33] Fairchild and then founder of Intel [20:35] along with noise it would coin Mo's law [20:37] but Jack has a great quote too about the [20:40] implications of the integrated circuit [20:42] and semiconductors he says what we [20:45] didn't realize then this was a little [20:47] later when they were inventing it was [20:49] that the integrated circuit would reduce [20:51] the cost of electronic functions by a [20:54] factor of a million to one nothing had [20:57] ever done that for anything before wow [21:00] it's such a great way to frame it too [21:01] like this had never happened in human [21:03] history there was this thing that used [21:05] to be X expensive in terms of resources [21:09] and then magically one day it's a [21:10] million times cheaper yeah that's crazy [21:13] I didn't realize it was on that scale [21:15] this is probably a good time to talk [21:17] about some definitions because there are [21:18] some things that we've thrown around [21:20] already I think everyone has a general [21:22] understanding of what these things are [21:23] but it's worth understanding more [21:24] precisely before we move on the first of [21:27] which is a transistor [21:29] the best way to think about a transistor [21:30] is not the tiny little transistor that's [21:33] on a silicon die today but think about [21:36] it as a little encased piece of [21:38] circuitry with three prongs coming out [21:41] of it and those three prongs will save [21:43] the technical names basically have an [21:45] input an output and something that [21:46] controls the input and the output it's a [21:49] switch it has two purposes the first of [21:51] which is being a switch where you can [21:53] decide that either stuff is going to go [21:54] through it stuff being voltage current [21:57] or none or it rounds to none and so that [22:00] way you can decide hey this binary piece [22:02] of equipment is either off zero or on [22:05] one okay so that's a transistor now a [22:09] transistor can be made out of lots of [22:12] different things it can take any [22:14] implementation why is everybody talking [22:16] about silicon well silicon as an element [22:20] is a semiconductor it is a metaloid it [22:24] has some properties that make it like a [22:27] metal like a conductor it has some [22:29] properties that make it non-conductive [22:32] imagine trying to move electrical signal [22:34] through a piece of wood it's not going [22:36] to work but imagine moving it through [22:37] copper it's going to work really well [22:39] and you're never going to be able to [22:40] interrupt it well geez wouldn't it be [22:41] great if we had some material a [22:43] semiconductor where we could modify [22:46] whether current was flowing through it [22:47] or not make a switch really easily [22:49] exactly exactly well and lots of things [22:51] are semiconductors germanium was the [22:54] main material for a while but germanium [22:56] is expensive and rare Sil [22:58] sand I think it's like the second most [23:00] plentiful minable element on Earth yeah [23:04] I mean it's sand right yeah there's one [23:06] other major thing though so we've been [23:07] talking about transistors the IC yeah [23:09] the IC the integrated circuit the [23:12] integrated circuit a transistor it's a [23:13] switch before the IC people are making [23:17] switches like you make one switch at a [23:18] time you wire it to another switch you [23:21] know if you've seen photos of VC and [23:22] vacuum tubes literally they're plugging [23:24] one tube into another you're still doing [23:26] that with transistors yep when noise and [23:28] Kilby invent the IC now you can put a [23:31] lot of switches on one thing and you [23:35] know fast forward today the latest [23:37] processor you know the 5 nanometer [23:38] processors that tsmc and basically [23:40] nobody else is churning out I don't know [23:43] billions trillions of switches are in [23:44] like a tiny little integrated circuit [23:47] without the integrated circuit that [23:48] never would have happened so this [23:51] invention this miraculous invention of [23:53] the integrated circuit it happened in [23:56] 1958 [23:58] when did Morris Chang join Texas [24:01] Instruments 1958 o fascinating [24:05] coincidence yes totally a coin [24:07] absolutely Co absolutely a coincidence [24:08] and again to pegas and history here [24:10] we're still I think 10 years before the [24:12] founding of Intel yes exactly 10 years [24:15] yeah moris obviously wasn't working [24:17] directly with Jack on inventing the IC [24:21] but this gives you a sense TI this is [24:23] the place this is like Google Plus [24:25] Facebook without the world paying [24:26] attention to them yes and in Texas so [24:29] Morris gets assigned as his first [24:31] project to a sort of Problem Child [24:34] within TI they have entered into a deal [24:38] with IBM IBM is working on their first [24:41] Mainframe computer major project that's [24:44] going to use transistor logic instead of [24:46] vacuum tubes the IBM [24:49] 7090 and they anticipate so much demand [24:52] for this product usually IBM [24:55] manufactures everything for all their [24:57] products themselves [24:59] but they're like we need more chips than [25:01] we're going to be able to make ourselves [25:03] so we need a second source for our chips [25:05] they turn to TI and they're like hey [25:08] we're going to give you all the designs [25:09] for you know how to do this this chip [25:11] that we want for our product we want you [25:14] to additionally manufacture some of [25:16] these in addition to our own line you [25:19] might even say almost like a contract [25:21] manufacturer of chips so like a Foundry [25:23] business almost you know hm interesting [25:27] but it's not going too well well so [25:30] IBM's own plant is turning out [25:33] transistors with about a 10% yield which [25:36] means that of every hundred chips that [25:39] they turn out of the plant 90% of them [25:41] fail and only 10% of them work Yik [25:43] that's the first party line the ti line [25:47] has about a 0% yield they're lucky if [25:49] they're getting any that way almost [25:51] everything coming off the line fails at [25:54] TI when Morris shows up so Morris would [25:57] say about this later quote the [25:59] supervisor was concerned The Operators [26:01] were concerned everybody was [26:05] concerned so Morris remember he's a [26:07] mechanical engineer by training right so [26:09] he starts tinkering he's like well I [26:10] know this is a mechanical process [26:13] chemical and mechanical process creating [26:14] this stuff I'm just going to use my [26:16] training and optimize it like a good [26:18] mechanical engineer so he starts doing [26:21] some stuff after about 4 months he gets [26:23] the yields at the ti plant up to 20% so [26:27] Twice As Good As the First party line at [26:29] IBM and there's a great profile that was [26:31] one of the main sources for this episode [26:33] in i e Spectrum oh yeah great industry [26:36] magazine that we'll quote from here they [26:38] write suddenly even TI president Pat [26:41] Hagerty knew Morris's name IBM thought [26:44] Chang had just gotten lucky but when the [26:46] company IBM sent Engineers down to talk [26:49] to him Morris described the theories [26:51] he'd been testing and explained why his [26:53] experimental process worked this [26:55] achievement propelled him into his first [26:57] managerial job job creating a germanium [26:59] transistor Development Department with [27:01] 20 plus Engineers reporting to him so [27:04] this is his first big win here in The [27:06] Foundry business so on the back of all [27:08] this TI is like all right we got a [27:10] rising star here they offer to sponsor [27:13] him to go finally get his PhD they even [27:16] offer to continue paying his full salary [27:17] while he's getting his PhD which they're [27:19] paying for all right so they think like [27:21] very highly of him very very highly of [27:24] Morris I mean this one probably made the [27:26] millions doing this in [27:29] 1958 it's funny I don't know anything [27:31] about the commercial success of that [27:32] particular IBM Mainframe but if it's the [27:35] first one that's transistor based [27:37] instead of being vacuum tube based I [27:38] have to imagine that it was far more [27:40] efficient for customers customers are [27:41] probably lining up for it I bet there's [27:43] a lot of demand and what's Morris making [27:45] a year like $20,000 maybe you know how [27:48] much does it cost to go to Stanford then [27:49] not much so they're like sure so Morris [27:52] goes to Stanford but he's now like a pig [27:53] in mud he has found his calling he can't [27:55] wait to get back to Texas back to ti so [27:58] finishes his PhD in 2 and 1/2 years wild [28:02] one of the Stanford interviews is with [28:04] John Hennessy the president of Stanford [28:05] at the time John's like Morris tell the [28:07] students how did you finish your PhD in [28:09] two and a half years Morris is like I [28:11] know I was focused I didn't do much else [28:13] so by 1964 he's done he's back at TI and [28:17] this is right as people have discovered [28:20] that silicon is way more cost effective [28:22] and scales up way better and if I [28:24] remember right the initial attempts at [28:26] using silicon [28:28] were that people didn't know how to work [28:30] with it yet and so even though it was [28:31] more abundant and cheaper there's some [28:33] particular manufacturing process that [28:35] you have to do to Silicon in order to [28:38] make it as viable as it became yes how [28:41] to dope silicon to make it function and [28:44] produce it at scale as a semiconductor [28:47] and listeners this is where you should [28:49] start to get the idea that especially [28:52] today manufacturing these products [28:55] involves the most advanced process in [28:59] human history consisting of layers of [29:01] innovation in chemistry physics [29:03] mathematics it's breakthrough after [29:05] breakthrough after breakthrough all [29:07] building on top of each other which need [29:09] to all happen in the manufacturing [29:12] process so even here in what [29:15] 1964 we're starting to get into the [29:19] level of complexity where it's some of [29:21] the most advanced science ever done [29:23] being applied in an engineering and [29:25] Manufacturing fashion to get even [29:27] marginal results you know at 20% yields [29:29] off the manufacturing line and little [29:32] preview to fast forward today tsmc [29:35] they're contract manufacturer for [29:37] silicon that is what they are tsmc has [29:40] 40% operating margins as a contract [29:43] manufacturer it's not like this is just [29:45] there's no technology or R&D like they [29:47] are one of the most advanced technology [29:49] organizations in the whole world there [29:51] is so much IP just in the manufacturing [29:54] take out the design take out the [29:55] functions just making this stuff is so [29:58] hard I mean now it involves like lasers [30:00] like it's we'll get to it later it's [30:03] going to blow your mind how this stuff [30:04] is done but anyway so Morris he's coming [30:06] up he's learning like literally as this [30:08] whole industry is getting developed he's [30:10] right there so a couple years after he [30:13] gets back from Stamford he's still [30:15] Rising through the ranks in [30:17] 1967 TI makes him a general manager of [30:21] one of the divisions within the [30:22] semiconductor business and that's where [30:25] he has his next big breakthrough and [30:28] this is on the business side so Morris [30:33] notices what they're doing setting up [30:36] these new plants for all these [30:38] successive new methodologies and [30:40] processes of manufacturing you know at [30:42] this point integrated circuits and [30:44] silicon and pumping out these chips it's [30:46] super expensive to do this super cost [30:49] Capital intensive so what TI and [30:52] everybody else in the industry did when [30:54] they would start a new product line that [30:56] would use a new Fab for chips they [30:58] charged a lot of money for it cuz like [31:01] man they put a lot of money into these [31:02] things so right off the gate you want [31:04] the latest new hotness in the end [31:06] products that TI is selling they're [31:08] going to charge a lot of money for them [31:09] yep Morris realizes he's like that's not [31:12] actually optimal to do that because as [31:15] evidenced by his first big win at TI [31:18] with the IBM line there's a learning [31:20] curve to getting the yields right and [31:23] learning how to manufacture a new [31:25] process and in the beginning you're [31:27] going to have really low yield and so [31:30] what you want ideally from a fabrication [31:34] perspective is you want to have a ton of [31:36] volume from the get-go like as soon as [31:38] the plan is online you want to be [31:39] running at Max Capacity so that you can [31:42] a learn as fast as possible get yields [31:45] up to the profitable levels and then you [31:48] want to still be running at Max Capacity [31:49] as long as possible because you already [31:51] spent the fixed cost right to make the [31:53] plan basically you always want Max [31:54] Capacity so when you start it out by [31:57] pricing so high you kept demand low and [32:00] you weren't able to get up to capacity [32:02] fast enough it's almost like they didn't [32:04] realize the benefit of the potential [32:06] operating leverage that they had because [32:07] they were just passing their exact [32:09] economics onto their customers and [32:11] saying you basically have to pay us for [32:13] us to do all these fixed costs and then [32:15] you'll get all the benefits of how cheap [32:18] it is to stamp it off the press every [32:19] time whereas what they really should [32:21] have been doing is saying we will make [32:23] an investment you know we'll eat the [32:25] cost of having to spin all this up but [32:26] boy are we going to be super profitable [32:28] on every chip that comes off the line [32:30] yep totally so Morris is thinking about [32:32] this he hires BCG and they come up with [32:36] the idea of actually pricing low to [32:38] start to drive this volume and speed up [32:40] the yield curve and then also the side [32:42] benefit of that is if they're pricing [32:44] low and everybody else is pricing High [32:45] they're going to grab a ton of market [32:46] share and probably keep that paying [32:48] Consultants I know well so here's Mars's [32:50] quote about this he says this was in the [32:51] late 60s and Boston Consulting Group was [32:54] a very small outfit when we did this and [32:57] we used load of data a lot of theory and [32:59] a lot of effort the result was so-called [33:01] learning curve pricing so start low and [33:05] then continually automatically reduced [33:07] the price every quarter even when the [33:10] market did not demand it and this was a [33:12] very successful effort even though it [33:14] was somewhat controversial a lot of [33:16] people thought we were being foolish why [33:18] would you reduce the price when you [33:19] didn't have to but we did it because we [33:22] believed in it and indeed our market [33:23] share just kept expanding that combined [33:26] with other strategy [33:28] made the ti integrated circuits business [33:30] the biggest IC business in the world and [33:33] also the most profitable this is right [33:36] when Intel's getting founded so screw [33:37] Fairchild screw National screw and like [33:40] TI is kicking all of their butts and [33:43] it's thanks to Morris and interestingly [33:45] enough the reason I always thought that [33:47] Fairchild was sort of so successful in [33:50] those days was out of all the defense [33:53] spending and research that was being [33:55] done at Stanford the Govern as a [33:58] customer but is Texas Instruments [34:01] playing in that ecosystem at all good [34:03] question probably I mean I think this is [34:05] a case of the rising tide is floating [34:07] all boats yeah fairchild's killing it [34:10] Intel's killing it National's killing it [34:12] TI is just killing it bigger than [34:14] anybody else all right listeners this is [34:17] a great time to thank our longtime [34:19] friend of the show vanta the leading [34:21] trust management platform vanta [34:23] automates your Security reviews and [34:25] compliance efforts so Frameworks like [34:27] sock 2 ISO [34:29] 271 gdpr and hipa compliance and [34:32] monitoring vanta automates all of these [34:34] for your company and makes them fast and [34:36] simple plus vant is always connecting [34:38] and monitoring continuously not just a [34:41] static point in time it's funny when we [34:43] first started working with vanta four [34:45] years ago we thought oh this is one of [34:47] those great acquired Universe products [34:49] that lets you focus on just making your [34:51] beer taste better in other words [34:53] customers can focus on what [34:54] differentiates their product and [34:56] Outsource things that don't to free up [34:58] time and resources but over the last few [35:01] years their product has advanced so much [35:03] that it's not just vanta does that for [35:05] you it's now actually vanta does that [35:07] better than you could on your own [35:09] without a continuous real-time [35:11] monitoring system there's no way you [35:13] could do this that you could give your [35:14] vendors and customers this level of [35:16] confidence and trust there's even data [35:18] now on this too between the ability to [35:20] enable new revenue and The Time Savings [35:23] a new IDC white paper found that vanta [35:25] customers achieve 530 $5,000 per year in [35:29] benefits and the platform pays for [35:31] itself in just three months vanta is the [35:33] one single software pain of glass that [35:36] connects to all of your services via [35:38] apis and eliminates countless hours of [35:40] manual work for your organization they [35:42] now have ai capabilities to make this [35:44] even more powerful and they integrate [35:46] with hundreds of external tools plus [35:48] they let customers build private [35:49] Integrations with all of their internal [35:51] systems so whether you're a startup or a [35:53] large Enterprise if your company is [35:55] ready to go back to making your beer [35:56] taste better and and automate compliance [35:58] and streamline Security reviews like V's [36:01] 9,000 customers around the globe [36:03] including at lass and Kora and Factory [36:05] head on over to v.com Acquired and just [36:08] tell them that Ben and David sent you to [36:10] get $1,000 of free credit that's v.com [36:13] acquired thank you vanta so on the back [36:17] of this Morris gets promoted to VP at TI [36:22] one level below the CEO running the [36:24] entire semiconductor business thaten s [36:27] in 1972 and he becomes the obvious [36:31] leading candidate to be the next CEO of [36:33] TI which he's like yeah I want to do [36:36] that I'm focused this is what I love [36:38] this is my this is why I've been going [36:39] to the bar for three years reading a [36:40] textbook exactly but might be fair to [36:44] say history turns on a knife Point [36:46] things don't entirely go as planned [36:48] three different viewpoints as far as I [36:51] could identify on what happens next to [36:54] Morris at TI he does not become the next [36:56] CEO obviously Viewpoint number one is [37:01] simply and probably fair that he was [37:02] just discriminated against because he [37:04] was ethnically Chinese although at this [37:07] point I'm pretty sure already he was an [37:08] American citizen but anyway and he got [37:11] passed over for I have no evidence for [37:13] it but not be at all surprised that that [37:15] was part of what was going on yep so [37:17] that's one two second point which Morris [37:20] totally acknowledges TI was a really big [37:23] company the semiconductor division he [37:25] had made it probably the most successful [37:27] and the most fastest rising division [37:29] within the company but you mentioned [37:31] calculators they were starting to launch [37:34] the consumer products division at this [37:36] time and so in 1978 so six years he's [37:40] running the semiconductor division as VP [37:43] they move him over to VP of consumer [37:45] products in [37:46] 1978 because this was a big new [37:48] strategic initiative and it wasn't going [37:50] super well and they're like oh Morris is [37:52] a great manager he can fix this and turn [37:55] it around different set of competencies [37:56] though I mean like marketing yeah here's [37:59] Morris's quote on this Mark Shepard then [38:02] chairman and CEO of TI agreed with the [38:04] prevailing wisdom at the time that a [38:06] good manager could manage anything in [38:08] this case I think he was wrong I found [38:10] the consumer business to be very [38:12] different like you were saying the [38:14] customer set completely different the [38:16] market completely different and what you [38:18] need to get ahead in that business is [38:20] different too in the semiconductor [38:21] business it's just technology and costs [38:24] in consumer technology helps but it's [38:26] also the appeal to Consumers which is a [38:29] nebulous thing not Morris's Dr suit or [38:32] at least not anything he's trained in [38:34] yeah that makes total sense so in 1983 5 [38:38] years after he gets moved over to take [38:40] over the consumer business he hasn't [38:42] turned it around it's still struggling [38:43] he gets demoted to quote head of quality [38:47] and people Effectiveness which is pretty [38:49] much a slap in the face like this dude [38:51] built your semiconductor business is [38:53] this when he says he was put out to [38:54] pasture exactly so that's number two [38:57] here's number three I found some [38:59] evidence on this it's unclear to me how [39:02] much of this is Morris's fault versus [39:04] his successor but while Morris was [39:07] definitely responsible for making TI [39:10] semiconductor a [39:12] Powerhouse at some point towards either [39:15] at the end of his tenure running it or [39:17] under his successor they totally dropped [39:19] the ball and this is when Silicon Valley [39:22] in California takes over so in the mid [39:26] 70s the semiconductor industry [39:28] transitioned over to the metal oxide [39:31] process the Moz you ever hear about MOS [39:33] Moz semiconductors the precursor to seos [39:36] exactly so that happened in the 70s and [39:40] TI again was well they had the best [39:42] Engineers they were well position to [39:44] lead this transition they didn't and [39:46] actually most of the talent within TI [39:50] that were the ones that led the industry [39:52] transition to M left including probably [39:55] most prominently a guy named LJ [39:58] s who left and founded a company called [40:01] mzc and then he later became a [40:03] semiconductor venture capitalist and [40:04] founded seven Rowen Ventures which was [40:07] one of the early VC firms so he was a TI [40:09] guy and he left and you know the culture [40:12] at TI as shown by Morris's experience [40:15] was not like Trader a Silicon Valley you [40:17] leave it was like you're a company man [40:19] you know you stay at the company right [40:21] so Motorola poached a whole bunch of Moz [40:23] Engineers from TI and it all kind of [40:26] fell apart [40:27] culminating in the biggest huge loss [40:31] that this is really history turning on a [40:32] knife point in [40:34] 1980 so Morris has already transitioned [40:36] to Consumer products IBM puts out a [40:39] secret RFP bid proposal for a secret [40:41] project that they're working on this is [40:44] 1980 by a new group based out of boka [40:47] ratan Florida do you know what I'm [40:49] talking about Ben I have no idea so some [40:51] listeners might know what I'm talking [40:52] about this is the secret project this is [40:54] the RFP to be the micro processor cessor [40:58] the CPU for this secret project the IBM [41:02] PC ah okay that was that a Boke of rone [41:05] yeah it was a secret project like a [41:06] Skunk Works division of IBM oh wow to [41:08] build the PC which was a big you know [41:10] IBM was the Mainframe company we're [41:11] going to build a personal computer so [41:14] skunkworks project and TI you know a [41:17] couple years earlier under Morris would [41:18] have been an obvious candidate remember [41:20] he had the relationship with IBM going [41:22] all the way back TI probably should have [41:25] been the processor chosen right instead [41:28] of course it was Intel I think it was [41:31] the 8088 that was chosen for that first [41:33] one wow boy did that set things in [41:35] motion well then the architecture [41:37] standardizes on x86 and boom there goes [41:40] the whole next generation of computing [41:43] away from TI over to Intel the sort of [41:45] family of IBM with Intel processors and [41:48] eventually running Microsoft operating [41:50] systems and then all the IBM clones all [41:52] running Intel processors okay so this is [41:55] really where sort of major loss in the [41:58] highway of History TI accidentally took [42:00] the offramp there they did now okay is [42:02] that Morris's fault is that not Morris I [42:04] don't know certainly the culture at TI [42:07] was we rotate you around you're going to [42:09] fix consumer he didn't fix consumer but [42:11] couldn't and then this like the [42:14] semiconductor Powerhouse took an [42:16] off-ramp as you say all that like his [42:19] career TI is basically over so he was [42:21] the rising star everybody thought he was [42:23] going to be the next CEO and at age 52 [42:27] in 1983 after he stays a couple years [42:29] being the head of whatever yeah he was [42:32] something [42:33] staff yeah he just resigns uh and he's [42:36] like wow well I guess this is it my [42:39] career at TI 30 years done he's still [42:43] regarded super highly in the industry [42:45] though in the semiconductor industry so [42:46] people start calling him with [42:48] opportunities and he wants to be a CEO I [42:50] mean that's what that's the he wanted to [42:52] be CEO of TI that didn't happen like he [42:53] wants to be CEO but he whittel it down [42:55] to two opportunities he's going to [42:57] consider one is to go to a competitor [43:01] called General instrument which people [43:03] may have heard of another one of these [43:05] old chip companies was based in New York [43:07] and Manhattan in New York City actually [43:09] to go be their coo the number two there [43:12] with a understanding they like hey if [43:14] things go well in a couple of years [43:15] you'll replace the CEO become the CEO [43:17] there or to become a venture capitalist [43:20] really yeah so he was weighing the two I [43:22] don't know where or how I couldn't find [43:24] that out but he was weighing these two [43:27] opportunities the VC idea is going to [43:29] come back up in a big way in a second [43:32] but obviously he goes with General [43:33] instruments GI a his dream is to be CEO [43:37] B he's got this chip on his shoulder [43:39] from the way TI ended so yep great so he [43:43] goes off to New York you know he leaves [43:44] Texas lives in Manhattan things are [43:46] going to work out at GI the thing though [43:49] is GI had a very different culture than [43:52] TI you know TI was this research build [43:54] develop technology push the ball forward [43:58] GI is this new york-based they were [44:00] almost like at the time like a Proto [44:02] Tech private Equity Firm their strategy [44:05] was they just acquired lots of oh [44:08] different semiconductor businesses [44:09] either independent companies or [44:10] divisions from other companies try and [44:12] integrate them no they would acquire [44:14] them they would get these business units [44:15] into good shape and then they'd sell [44:16] them again oh really yeah they were like [44:18] literally they were like a financial [44:19] engineering firm basically definitely [44:22] not Morris's cup of tea so he only stays [44:25] there a year it's clear that that's not [44:28] a good fit so he resigns again so within [44:31] less than 18 months he's had two major [44:36] setbacks basically you know his dream is [44:38] over like here's the quote from him he [44:40] says after these two setbacks at TI and [44:42] GI I did not think that my aspiration to [44:45] be the CEO of a major US company was in [44:47] the cards Well turns out he was right he [44:50] was not going to be the CEO of a major [44:52] US company so how do we go from this [44:57] dude in his [44:58] mid-50s former Rising Star now washed up [45:02] from that to he's in Taiwan he's co of [45:05] tsmc I don't think you could ever script [45:07] this out I think this is probably the [45:09] most unique every founding story is [45:10] unique but I think this might be the [45:11] most unique founding story we've had on [45:13] acquired so far so back when Morris was [45:18] at TI when he was running the [45:20] semiconductor business there he went [45:22] over to Taiwan a couple times to talk [45:25] about building a Manufacturing plant [45:27] there TI would own and build a [45:29] manufacturing plant but Outsource to [45:30] Taiwan not like a tsmc style business [45:33] like is a ti plant there anyway he had [45:35] no connection to Taiwan remember he's [45:36] Chinese he's not from Taiwan people are [45:38] like oh Morris you went back to Taiwan [45:40] he didn't go back to Taiwan yeah he [45:41] talks about how Taiwan was a Strange [45:43] Land to him when he first got there that [45:45] it's not going back so it's not the land [45:47] is a strange place to him but if he is [45:49] going to call some place home and return [45:52] there is it the People's Republic of [45:54] China well I think he would say at this [45:56] point it's America he's been in America [45:57] for he's a US citizen well I don't know [46:00] what he would say it's complicated so [46:02] anyway so he had met a bunch of [46:04] government officials in Taiwan when he [46:06] was talking about building this plant [46:08] over there and that was back in the 70s [46:11] now we're in the 80s mid 80s taiwwan at [46:15] this point it's a manufacturing Nation [46:18] you know they have no IP they have no [46:20] technology okay the quote's great all [46:21] right so this is Morris we had no [46:24] strength in research and development or [46:26] very little at anyway we had no strength [46:28] in circuit design I see product design [46:30] we had little strength in sales and [46:32] marketing and this is of course [46:33] referring to Taiwan as a nation and we [46:35] had almost no strength in intellectual [46:37] property the only possible strength in [46:39] Taiwan that we had and even that was [46:41] just a potential one not an obvious one [46:43] was semiconductor manufacturing wafer [46:45] manufacturing and so what kind of [46:47] company would you create to fit that [46:49] strength and avoid all the other [46:50] weaknesses the answer was a pure play [46:53] Foundry yeah I mean that was Taiwan at [46:54] the time so to give you a sense the [46:56] average margin of a Taiwanese company at [46:58] this point in time in the mid 80s is 4 [47:00] to 5% 04 to 05% gross gross before you [47:05] even have overhead operating costs yeah [47:08] I mean it was like um you know if you [47:10] were a grew up around when Ben and I did [47:12] you know sort of born in the 80s in the [47:14] US you see made in Taiwan on everything [47:17] like you know Barbie dolls toys clothes [47:19] everything was made in Taiwan you know [47:21] now it's made in China or made in [47:22] Vietnam or elsewhere but made in Taiwan [47:24] was super low-end physical manuf in [47:27] stuff yeah and to way pull forward the [47:30] seven power section as Hamilton Helmer [47:32] would sort of explain if your margins [47:34] particularly your gross margins are only [47:36] four or 5% you're in an industry or a [47:40] business where all the profits are [47:42] arbitraged away and everyone's just [47:43] raised to the bottom on prices and no [47:45] one's able to build any real Enterprise [47:47] Value because everyone's just out [47:49] competing each other for Pure commodity [47:51] I mean four to 5% gross margins people [47:54] used to hammer on Amazon I guess for [47:56] being like a low gross margin business [47:58] in the like 40% like you know anyways I [48:02] can't even imagine running a company [48:04] with that level of gross margins so the [48:07] Taiwanese government though they wanted [48:09] to come up in the world they're like [48:10] this is where we are now this is not [48:12] where we want to be yep so they knew [48:14] that technology was the way and so they [48:16] had decided back in the 70s that they [48:18] would establish an initiative called the [48:21] industrial technology Research Institute [48:24] or it and the goal was for it to be like [48:27] the Bell Labs of Taiwan to do some tech [48:29] transfers from the US and elsewhere and [48:31] hom grow some real technology businesses [48:35] in Taiwan so that you know maybe they [48:37] can lift uh businesses out of poverty [48:40] there at least and so Morris wasn't [48:41] going to Taiwan to start tsmc he was [48:44] being recruited to itri so one of the [48:46] ministers he had met a guy named KT Lee [48:49] who because of this he would also become [48:52] venerated in Taiwanese history he's [48:53] known as the father of taiwan's economic [48:55] Miracle literally because this wow he [48:58] recruits Morris to come over and run rri [49:02] be like the head of bell Labs Taiwan [49:04] essentially and this is like a [49:05] ridiculous thing for Morris to do he had [49:07] been you know captain of American [49:10] Semiconductor industry he was put out [49:12] the pasture at TI but at least he was [49:13] still a TI and then he was coo at [49:14] General instrument he's going to go over [49:16] to Taiwan and take and run like a [49:19] research park there like what and like [49:21] every time someone starts something like [49:23] this it doesn't go well a government top [49:26] down down Innovation mandate from a [49:28] country that's not a world power tends [49:31] not to turn into a gigantic economic [49:33] success this is like all the countries [49:35] and cities and the like that are like oh [49:37] we're going to build the next Silicon [49:38] Valley in totally XYZ and we're going to [49:41] recruit some former Silicon Valley [49:42] person to come do that and it's going to [49:43] work probably not going to work yeah so [49:46] everybody tells him not to do this all [49:48] his former colleagues his wife at the [49:51] time tell him not to do this his [49:52] marriage was actually falling apart [49:54] maybe in part because of this and you [49:56] know he's had all these experience he's [49:58] like you know what I just I need to [50:00] change the scene I gotta get out of here [50:03] so he takes the job and he figures you [50:05] know it's going to be cushy this is like [50:07] a he think about this is like the pseudo [50:09] retirement going into so here's his [50:11] quote by then I was financially pretty [50:13] secure I was not rich but you also have [50:16] to realize that the standards of wealth [50:18] were much lower back in 1985 and he's [50:20] going to live in Taiwan where you know [50:22] corporate magnets have 5% gross margins [50:25] but says but still in absolute standards [50:27] I was financially secure which meant [50:28] that I could live according to the way I [50:30] desire which was actually pretty modest [50:32] for the rest of my life without having [50:33] to earn a living or a salary this is [50:36] retirement he also makes a joke I [50:38] remember after that about how by the way [50:40] interest rates were higher back then so [50:41] that was much more achievable princip [50:44] totally so 1985 he goes over he takes [50:47] over as president of rri it's kind of a [50:50] culture [50:51] Clash so this is retirement for Morris [50:55] but he's still coming from from this [50:56] hard charging industry all of the [50:59] employees of rri are people in [51:01] government jobs in Taiwan and government [51:05] jobs not even you know in like a [51:07] democracy because Taiwan is like gets [51:10] under martial law or I think that it [51:11] just ended you know this is not the same [51:14] you know these are like jobs for life [51:16] you're a government official in a [51:17] non-democracy type organization Morris [51:21] says back then they considered me a [51:22] foreigner who suddenly became their boss [51:24] they were scared of me and they were [51:26] right to be scared of him so there was [51:28] one thing though that the government had [51:30] done right before Morris showed up which [51:33] was they had successfully negotiated one [51:36] technology transfer license in the [51:39] semiconductor industry from did you find [51:41] out what company this was this is [51:43] probably what they were trying to [51:44] negotiate with ti for I do it's a [51:48] three-letter acronym oh y yep yep yep we [51:51] haven't talked much about it on this [51:53] show but this is another yeah talk about [51:55] captains of American indust [51:57] RCA RCA that's right yeah so RCA had a [52:01] semiconductor line and the government in [52:04] the 70s the Taiwanese government had [52:07] negotiated a tech transfer but this is [52:09] like 10-year-old semiconductor [52:11] technology right this is not like the [52:12] latest generation no TI and Intel and [52:16] everybody like you know in Fairchild [52:17] they're you National they're leading the [52:19] way they're at the bleeding edge of [52:20] semiconductor manufacturing process RCA [52:23] was already at least a generation behind [52:27] by the time it actually gets on the [52:29] ground in Taiwan they're two and a half [52:31] Generations behind the leading producers [52:34] so it's like the only thing that you [52:36] could do with that is super low-end [52:39] stuff right there are some category of [52:40] goods that don't need a fast or the [52:43] latest processor and totally even today [52:46] when tsmc or Intel or Samsung or whoever [52:49] builts a Fab the Leading Edge Fabs they [52:52] produce the Leading Edge stuff for a [52:54] while and then the new generations come [52:55] on they don't shut down the old ones [52:58] it's just chips that don't need the same [53:00] bleeding edge performance they keep [53:02] getting made on the old ones and often [53:04] that's Automotive or now what we think [53:05] of as iot but the stuff in your [53:07] smartphone obviously is the the Leading [53:09] Edge yeah so the government rri does [53:13] actually spin out a company using this [53:16] old RCA technology that would be called [53:19] UMC United micro Electronics Corporation [53:23] not a technology leader it actually does [53:27] okay in the long run they would later [53:29] spin out their own chip design business [53:31] so UMC was doing both fabrication for [53:34] third party clients and designing some [53:36] of their own chips with the Fab that [53:38] they created they spin out their chip [53:40] design business later that becomes [53:42] mediate tech oh no way yeah which is a [53:44] $50 billion doll company today so like [53:47] you know the government did pretty good [53:49] like this is pretty good totally they [53:50] were doing and you know when Morris [53:52] arrives because of this he's not [53:53] starting from a standing start like it's [53:55] not good but there's acquired IP they've [53:57] created a company there's a paved path [54:00] so he gets to work at rri he's working [54:01] on all this he's transforming the [54:03] organization into a high performing [54:05] organization and then all of a sudden [54:07] out of nowhere KT Lee comes back to him [54:10] and he like hey great you're running you [54:12] know our Bell Labs you're run an e Tre [54:14] now I want you to start a [54:16] company Morris is like uh M's like yeah [54:19] yeah yeah I don't want you to have [54:21] somebody else in need do it I want you [54:23] Morris Chang to start a new [54:25] Semiconductor Company here in Taiwan and [54:27] I want you to make it into a global [54:29] leader Morris is like [54:32] um okay he doesn't say this directly [54:36] well he's got a great quote I'm gonna [54:37] say in a minute but again remember this [54:38] is not a democracy in Taiwan at this [54:40] time Morris is also on his third job in [54:43] three years yeah he doesn't need a [54:44] salary to survive but this is kind of [54:46] the end of the Rope for him if he gets [54:48] fired here at e tree he's legit done [54:51] done done done so he kind of doesn't [54:54] have a choice here quote this is so [54:57] Morris so great he says it was like in [54:59] the movie The Godfather it was an offer [55:01] I couldn't [55:03] refuse and I do think the implication [55:06] was go start an Intel or go start an IBM [55:08] it wasn't go start the very first Pure [55:11] Play Foundry Lee had no he was the [55:13] government he was a minister right he [55:14] was like go start a Semiconductor [55:16] Company make it a world leader right [55:18] those semiconductor companies they do [55:19] really well so go do that and that's of [55:22] course when Morris says okay I'm being [55:25] told I should do this I have some [55:28] latitude I can take and some Liberties I [55:30] can take on how I do it and the quote [55:32] that I read earlier about evaluating [55:34] exactly what type of Semiconductor [55:36] Company should I start that's how he [55:38] sort of informs the business plan so Lee [55:41] is like all right good we're capiche [55:43] we're clear come back to me in a week [55:45] with a business plan tell me what you [55:47] need and we're going to make this happen [55:49] right like okay a week all right and [55:53] then like a day later Lee supposedly is [55:55] like I'm need you come in on Friday so [55:57] like you got like 3 [56:00] days you know they say necessity is the [56:02] mother of invention and yet these three [56:04] days are what creates the you know now [56:07] ninth most valuable company in the world [56:09] Morris comes up with this brilliant idea [56:12] to create a pure play Foundry Company to [56:16] be a contract manufacturer sounds genius [56:18] today and hindsight as Steve Jobs would [56:21] say you know it's easy to connect the [56:22] dots looking backwards but at that time [56:25] was a good idea David [56:29] well [56:30] no the answer is no you know like we've [56:34] said all along all the chip companies [56:37] all the you know American and European [56:40] and you know Japanese all the leading [56:42] semiconductor companies they made their [56:43] own stuff you know and there was some [56:46] sharing of production and some companies [56:49] were emerging that were borrowing [56:52] production from the big guys there's a [56:55] great quote right around this time from [56:56] Jerry Sanders who is the co-founder and [56:58] CEO of AMD and he famously said in the [57:01] mid 1980s that quote real men have Fabs [57:05] that's right oh what a quote so ironic [57:08] because in the 2000s AMD would spin out [57:11] its Fabs and go fabulous Global [57:13] foundaries yeah into Global foundaries [57:15] but yeah like this was not an obvious [57:17] idea like if you wanted to be a real [57:18] Semiconductor Company you made your own [57:20] chips and the idea was like this isn't [57:22] like manufacturing Barbie dolls here [57:24] like this is real technology like you [57:26] need to control it Soup To Nuts and [57:29] already at this point in history I mean [57:30] this is an important point to make [57:32] because I didn't realize this coming in [57:34] where I thought wow Apple really [57:36] outsources their manufacturing they [57:38] outsourced some of it to tsmc and some [57:40] of it to foxcon and maybe some of those [57:43] people will start to do each other's [57:44] work no this is a completely different [57:46] thing assembling an iPhone is completely [57:50] completely different than taking a brand [57:53] new design for a Next Generation ship [57:56] and Manufacturing that chip one is [57:59] manufacturing and one is alchemy the [58:01] Alchemy can only be done by Alchemists I [58:03] think even here in the late 80s we're [58:05] already at the point where it's [58:07] manufacturing broad but yeah it's not [58:10] like well I got a [58:12] factory no no no no no no far the [58:15] opposite of that you we said a minute [58:17] ago this is a bad idea so Morris says [58:19] now however there was one problem with [58:22] the Pure Play Foundry model and it was a [58:24] fatal problem [58:26] it could be a fatal problem which was [58:29] where's the market he sounds like Don [58:30] Valentine here where's the market show [58:32] me the market this whole idea it was [58:35] really a solution looking for a problem [58:37] and of course the solution being that [58:39] all we have is manufacturing capability [58:41] here so let's start a company that just [58:43] manufactures and it's like you're [58:45] looking around like okay who manf like [58:48] real men have Fabs yeah you know there [58:50] no real startup I mean there are [58:52] startups but all these startups are [58:53] building their own Fabs like nobody [58:55] wants to do this so nonetheless you know [58:57] he has to start a company he's literally [59:00] got a gun to his head but he does have [59:02] the core Insight here it's interesting [59:05] these companies don't exist yet but [59:06] Morris has reason to believe that people [59:08] will want to start fabulous chip [59:10] companies and that they will need a [59:12] Foundry to Fab those chips and so he [59:15] says when I was at TI in general [59:17] instrument I saw a lot of integrated [59:19] circuit designers wanting to leave and [59:21] set up their own business but the one [59:23] thing or the biggest thing that stopped [59:24] them from leaving those companies was [59:26] they couldn't raise enough money to form [59:27] their own company because at the time as [59:29] we were just saying real men it was [59:32] thought that every company needed [59:34] manufacturing needed their way for [59:35] manufacturing and the most Capital [59:38] intensive part of a Semiconductor [59:39] Company of an IC company does the [59:42] manufacturing and so I saw those people [59:44] wanting to leave but being stopped by [59:46] the lack of ability to raise a lot of [59:48] money and build a for Fab totally right [59:52] but those companies this like if you [59:55] build it they will come they haven't [59:56] started yet they haven't come yet they [59:58] haven't come yet so Mor he knows what [60:00] the long-term Market is going to be but [60:02] he's got to find the short-term Market [60:06] he needs some like real politic here [60:08] like so what's that going to be so he [60:10] says well maybe I can go around to the [60:13] big guys they've been doing you know [60:15] just like my first thing back at TI [60:17] they've been doing some line sharing you [60:19] know for either new products that they [60:20] need excess capacity for or for older [60:23] products that they need to transition [60:24] some Fabs but they still need to make [60:26] components maybe I can take some of that [60:29] off their hands and so he goes around he [60:31] talks to Intel he talks to TI he talks [60:34] to everybody in the industry and they're [60:35] like yeah he talks to Motorola like sure [60:39] fine and the government had told him we [60:42] know it's goingon to take a lot of money [60:43] to set up a Fab we're good for half of [60:45] it but you got to go raise the other [60:46] half of it and we want you to raise it [60:48] from like an Intel or a TI you know [60:50] somebody who's going to be your first [60:51] customer and that they're going to be [60:52] bought in so he does the rounds he goes [60:55] and talks to everybody [60:57] he gets meetings with Intel he gets [60:58] meetings with ti they're both like you [61:01] know Morris we like you but [61:04] no so he's said the last ditch effort [61:07] and he has a meeting with Phillips the [61:09] Dutch company they have a semiconductor [61:11] business this so Morris he has a great [61:14] quote about this he says he would [61:16] describe Phillips as the first rung of [61:18] the second Raiders in in semiconductors [61:22] but they were the only interested option [61:24] so they put up 28% of the capital [61:27] government puts up 50% it ends up being [61:30] 220 million in total 110 is probably a [61:33] lot more than what the Taiwanese [61:34] government thought they were here and [61:36] then literally the premier of Taiwan [61:39] like the head of the government has to [61:41] then go around to all the other Business [61:42] Leaders in Taiwan and like Strongarm [61:44] them into investing the rest of it the [61:46] other what is that [61:48] 22% I guess yeah we also should say [61:52] remember that Phillips was a Dutch [61:54] company because that's going to come [61:55] into play later oh I don't know how [61:58] that's going to come into play yeah [61:59] putting a pin in Dutch okay we got a [62:00] surprise coming I'm going to be [62:02] surprised here we're doing it real time [62:03] doing it live this may be the craziest [62:05] part about the whole tsmc founding story [62:08] I'm 99.9% sure Ben you do not know this [62:11] o do you know what the pre-money [62:13] valuation was on tsmc no I couldn't find [62:16] that anywhere it was zero dollar Morris [62:20] Chang got no equity in the company zero [62:24] so 100% % of the company was owned by [62:28] the investors 50% by the government the [62:31] other 50% were owned by the investors [62:34] Morris got nothing and just got to keep [62:37] his salary he was a government [62:39] employee wow there by the grace of the [62:42] government oh my God isn't that [62:45] unbelievable like this is so the [62:47] opposite of Silicon Valley in worth [62:49] three billion dollars today well what he [62:51] did as tsmc started to work he basically [62:54] put all of his money into buying he [62:55] bought his own shares what in the [62:58] company I don't know if it was privately [63:00] before they went public on the Taiwan [63:02] stock exchange in 1994 and then the New [63:05] York Stock Exchange in 1997 but yeah he [63:07] put basically all of his excess cash [63:09] flow into buying tsmc shares oh my God [63:13] isn't that wild so the government owned [63:15] 50% of the whole business and you can [63:18] see their perspective too they're like [63:19] hey we hired you to do this and then we [63:20] told you to do this you are our foot [63:22] soldier you know we are the mafia [63:26] wow yeah things had really not gone well [63:29] in his career that he was willing to [63:30] take that deal yeah crazy right and okay [63:35] before we go on in the tsmc story we [63:38] need to have two real quick sidebars [63:40] yeah so was 1987 yep when tsmc gets [63:44] officially stood up they raise the money [63:46] at a Zer pre money valuation do you know [63:49] what other company other big thing [63:51] happened in 1987 we have covered it on [63:53] this show in the chip world [63:56] is this the founding of arm yes it is [63:58] yes yes arm JV between apple Acorn and [64:03] vsi logic which was the sort of [64:05] manufacturing partner they were an A6 [64:07] company that's a whole another sidebar [64:09] we're not going to get into but yeah [64:10] 1987 what a year brand new [64:13] unconventional instruction set [64:15] architecture that's totally different [64:16] than the x86 stuff that the whole [64:19] industry and world seems to have [64:20] standardized on at this point the Anis [64:23] mirabellis for the semiconductor [64:25] industry and useless right it's in 1987 [64:28] it's hamstr strong it's very few [64:30] instructions PCS are always plugged in [64:32] so what do we need a low power chip for [64:34] this thing is pathetic real men have [64:35] Fabs and real men use power okay so [64:38] that's sidebar number one arm get [64:39] started okay now I was wondering I don't [64:42] actually know the relationship because [64:43] obviously today a huge amount of [64:47] volume of tsmc's manufacturing is making [64:50] chips for iPhones which since the outset [64:53] has used arm chips that are using all [64:56] mobile devices iPhones and Android all [64:58] of which are arm and lots of servers [65:01] presumably there's some relationship [65:02] coming between tsmc and arm um well [65:06] they're really close Partners I mean the [65:07] way now this stuff is so integrated yeah [65:10] all the companies like arm the design [65:13] the Eda companies like synopsis these [65:15] guys are all deeply like the engineering [65:18] is all deeply ined with one another okay [65:20] so you mentioned Eda I'm going to take [65:22] your sidebar and I'm going to you're [65:24] going to raise me yeah I'm going to [65:25] raise you one more sidebar so listeners [65:26] we're two clicks out here so this is a [65:29] pretty good point to talk about how the [65:31] value chain went from one company that [65:36] created transistors and then they [65:39] designed the chip they manufactured the [65:41] chip they marketed the chip here's how [65:43] the value chain looks today and I think [65:45] you've already alluded to I think in the [65:46] 80s it already started to look like this [65:49] first there's Eda there's electronic [65:51] design automation this is the software [65:54] that professional ship designers use to [65:57] do their work so synopsis I think [65:59] Cadence yeah Cadence they're the yeah [66:01] they're the two leaders so that's like I [66:03] don't know Excel or like figma for chip [66:06] designers that's what they use [66:07] productivity tools yeah so that's [66:09] category one of four and of course as [66:11] you can imagine the software to design [66:13] the chips probably has to be very aware [66:15] of the manufacturing capability of who's [66:17] going to be manufacturing the chips but [66:19] let's put a pods in that for a second so [66:21] then of course there's the fabulous chip [66:23] design companies so today I think apple [66:25] and qualcom FRC eventually AMD after [66:29] they stopped being real men apparently [66:31] tons of innovative new startups now like [66:32] Cerberus PMI before Apple acquired it [66:36] yeah oh PMI is coming in a sec okay okay [66:38] so you've got the Eda companies that are [66:40] making the software The Fabulous [66:42] companies that are designing the chips [66:44] using the software then third there's [66:46] one company that we have not talked [66:48] about yet one component of the value [66:50] chain and these are the people that [66:52] manufacture the machines oh yeah that go [66:56] into the factories that the foundaries [66:58] operate there's actually one between [67:02] well no actually I think above Eda there [67:03] is one more part of the value chain [67:05] there's a fifth which is IP so that's [67:07] all like arm oh right yeah like [67:09] architecture IP there's actually a ton [67:10] of companies now that do just straight [67:12] up IP and I thought before this episode [67:14] these were like oh just shell companies [67:16] that Sue one another about IP it's not [67:18] that systems are on a chip now so like [67:20] everything is on one chip basically you [67:23] need a USB functionality in your chip [67:26] you don't need to design that you just [67:28] buy some IP off the shelf so there [67:29] companies that do that yep okay so [67:32] that's a good point so that's our fifth [67:34] sort of like IP they own the instruction [67:36] set architecture they kind of create the [67:38] general rules that you're playing by [67:39] when you're designing a chip such that [67:41] whoever is writing the compilers knows [67:43] what Assembly Language they're targeting [67:45] that can then operate on the chip that's [67:47] going to be designed so we covered the [67:50] Eda we covered the IP we covered the [67:52] fabulous companies there's somebody [67:54] before we get to foundaries which is the [67:57] equipment manufacturers that sell to [67:59] tsmc so more historically you've got [68:02] lamb research you've got Applied [68:04] Materials in the US you've got Tokyo [68:06] electron in [68:07] Japan but today I just want to give [68:10] everyone a taste of this and then we'll [68:11] get more to it later there's a company [68:14] that is also dutch-based [68:16] called there it is there it is asml [68:19] which was originally ASM lithography and [68:22] lithography is marginally in scope for [68:25] this episode there's a whole thing we [68:26] could do on the magical process that is [68:28] lithography me back to my high school [68:30] photo lab right yeah and the L you know [68:32] is lithography so the company was [68:33] originally called ASM lithography they [68:36] make the most advanced chip [68:38] manufacturing machines in the world [68:41] they're the only company that makes them [68:44] they're located still in the Netherlands [68:46] their biggest customer is [68:48] tsmc they and this is where I want to [68:51] bring it all the way back around and we [68:53] of course will talk about the magic that [68:55] is these machines later it was founded [68:58] in 1984 as a joint venture between [69:01] Advanced semiconductor materials [69:02] International ASM lithography and [69:06] Phillips oh wow I did not know that so [69:10] that that's crazy it's the beginning of [69:12] the relationship between tsmc and their [69:17] equipment provider W and what a [69:18] strategic point I mean because well it's [69:21] tsmc's insane Capital operating cash [69:24] flow production that enables them spend [69:26] capex above anybody else that allows [69:27] them to buy more asml equipment than [69:29] anyone else but that relationship wow I [69:32] mean these machines we'll get into it [69:34] later yeah it's gonna blow your mind [69:35] what this stuff does all right listeners [69:38] this is a great time to thank our big [69:40] partner here at acquired JP Morgan [69:42] payments when we told the team over at [69:44] JPM that we were remastering our tsmc [69:47] episode they smiled and said oh we have [69:49] the perfect thing for this episode that [69:51] most listeners won't know about but is [69:53] super important to how International [69:55] compan accomplish global trade which of [69:57] course that is like catnip for us so [69:59] we're like well tell us more yes JP [70:02] Morgan has a whole trade and working [70:04] capital group that helps companies [70:05] manage risk and access liquidity for [70:08] global trade in addition to of course [70:10] facilitating the actual payments to [70:12] drive growth yep imagine this scenario [70:15] you're a global auto parts manufacturer [70:18] and you build up inventory of a key car [70:21] engine component because it requires a [70:23] super complicated production process [70:25] but then you have to wait for the final [70:27] demand forecast from the actual car [70:29] company your customer before you can [70:31] ship units to them but in the meantime [70:34] you have to secure credit lines and [70:35] manage your own supply chain sounds [70:37] complicated enter JP Morgan's trade and [70:39] work in capital group not only is JP [70:42] Morgan payments the leader in all the [70:43] payments products we've shared over the [70:45] last year that drive growth for your [70:46] business they also have 150 years of [70:49] global expertise to help businesses [70:51] navigate issues like this this Legacy [70:54] dates back to 1799 so they literally [70:57] have centuries of leading the industry [70:59] through times of both economic growth [71:01] but also instability and as you all know [71:04] you can't time the market but if you [71:06] intend to be in business for decades you [71:08] know you're going to exist through all [71:10] sorts of different geopolitical [71:11] macroeconomic and supply chain climates [71:14] just like tsmc has so make sure you've [71:16] got a great payments partner that's also [71:19] prepared for anything as your business [71:21] grows JP Morgan really is a behind [71:23] the-scenes Powerhouse that helps all all [71:25] types of businesses and industries [71:26] facilitate payments mitigate risks [71:29] access liquidity to accelerate cash flow [71:31] and so much more our huge thanks to JP [71:34] Morgan payments you can learn more at JP [71:36] morgan.com acquired or click the link in [71:38] the show notes and just tell them that [71:40] Ben and David sent you okay back to my [71:43] second sidebar also going to be worth it [71:46] P semi we did an episode wave is like [71:48] episode 20 something this is like when [71:50] acquired was a very different show when [71:51] it was actually about small Acquisitions [71:53] that totally so I don't know that we [71:55] actually covered this but I uncovered in [71:57] the research for this episode do you [72:00] know the origins of PMI no I don't okay [72:03] so arm my sidebar number one you know [72:06] 1987 also created right they're just an [72:09] IP design company like I was saying in [72:11] your sidebar um it's like Inception over [72:15] here so they just license out the arm [72:18] architecture to other companies that [72:19] then design using the arm architecture [72:22] one of their original licenses was deck [72:25] digital Equipment Corporation like OG [72:28] way back in the day yep so they took the [72:30] arm architecture and they tuned it for [72:33] performance and they called what they [72:36] did at deck their version of arm that [72:38] they created strong arm and that product [72:42] line within deck would later be acquired [72:46] by Intel of all places okay crazy what [72:50] Intel acquired an arm architecture right [72:53] they're the x86 [72:55] marketed it as X scale I think they [72:57] ended up shutting it down huh so a bunch [72:59] of the core Engineers on the team like [73:01] the deck team that have been working [73:02] with arm from back in the day and [73:04] they're like we just got acquired by [73:05] Intel what the hell like you know screw [73:08] this we don't go work for Intel there's [73:10] no interesting flourishing alternative [73:12] architectures at Intel yeah we're arm [73:14] Engineers we're g go start our own [73:16] company that's PMI that's PMI ah so of [73:21] course the underpinnings of all of [73:24] Apple's chips today totally so the [73:26] lineage of all of Apple silicon probably [73:29] the most valuable defensible part of [73:31] Apple today in terms of Technology was [73:35] Armen deck to Intel to BMI to Apple [73:41] whoa that's wild I don't think I ever [73:44] knew that so you can trace Apple silicon [73:46] all the way back to Apple yeah because [73:48] arm was a JV with apple crazy with Intel [73:51] and deck in the middle wow okay so back [73:55] to tsmc the short-term Market is Morris [73:59] basically begs all of his old colleagues [74:02] in the US and European and Japanese [74:05] semiconductor Industries to just give [74:07] the drags to tsmc and it really was the [74:10] drgs so here's Morris on what this was [74:14] the idms would let us manufacture their [74:17] Wafers only when they didn't have [74:18] capacity or when they didn't want to [74:20] manufacture the stuff themselves anymore [74:23] now when they didn't have the capacity [74:24] and as us to do the manufacturing then [74:26] as soon as they got the capacity they [74:28] would stop giving us orders so it wasn't [74:31] a stable Market when they so it wasn't [74:33] actually a thing they wanted to [74:34] Outsource they were just they didn't [74:36] have the capacity so they needed some [74:38] extra excess space but then when they [74:40] got the capacity online they took it [74:41] away right and then the chps that they [74:43] gave us that they didn't want to make [74:44] anymore well the reason they didn't want [74:46] to make it was it was losing money so [74:48] they basically were just transferring [74:50] their losses on producing these chips to [74:52] TSS so how do they get out of this so [74:54] Mar continues the conventional [74:55] conclusion at the time was that there [74:57] was no Market that's why the Pure Play [74:59] Foundry idea was so poorly thought of [75:01] what very few people saw and I can't [75:03] tell you that I saw was the rise of the [75:05] fabulous industry I only hoped for it [75:08] and then as you said but I had better [75:09] reasons for hoping for it than the [75:10] people had Intel at TI in Motorola [75:12] because I was now standing outside when [75:14] I was at TI in general instrument I saw [75:16] a lot of these IC designers wanting to [75:18] leave start their own businesses and the [75:20] constraint was setting up their own Fabs [75:23] so like yes he saw that but remember he [75:26] had been considering becoming a VC [75:28] instead of going over to rri yep so this [75:32] is the ultimate end around he becomes [75:34] essentially like the world's best [75:37] semiconductor VC he takes an index out [75:39] on the whole future Innovation and [75:42] Entrepreneurship Market in [75:44] semiconductors by becoming the platform [75:47] that they're going to build on instead [75:49] of like going and investing in them like [75:51] he enables all of it he's like the Y [75:52] combinator of semiconductors right or in [75:55] many ways the tensent 10cent of course [75:57] also does direct investing but the idea [75:59] that you could get distribution through [76:01] WeChat it's kind of like it's not [76:02] distribution but it is manufacturing [76:05] there is a thing that you have to raise [76:07] 10 to 20% of the capital that you [76:09] otherwise would have needed to raise if [76:11] tsmc exists yep and just like Don [76:15] Valentine you know kind of when he left [76:17] to go join VC a generation earlier again [76:19] it's not VC it's tsmc it's building the [76:22] platform but Morris is a hero so all [76:25] these Engineers they all look up to him [76:27] and he knows a lot of them personally [76:29] the ones he doesn't know like who's not [76:30] going to take a meeting with Morris [76:32] Chang right he almost ran TI totally [76:34] like he did all this amazing stuff it's [76:37] interesting because it's like with the [76:38] incumbent of course because they had it [76:41] in their DNA to be a manufacturer of [76:44] course they wanted to take the most [76:46] profitable things and manufacture them [76:47] in house but if you actually are betting [76:49] on all these startups that will never [76:51] develop DNA to be their own manufacturer [76:54] they never want to take that back yeah [76:56] and so Morris is now going out and [76:57] evangelizing and he's like all these [76:59] great designers like we're an option for [77:02] you now like you want to leave you want [77:03] to start your company you don't need a [77:04] Fab we'll be your Fab you know it takes [77:06] a couple years for a couple years tsmc [77:08] has to survive on the drags from the [77:10] idms the big guys but after a couple [77:14] years these startups get going you know [77:15] little companies like Qualcomm broadcom [77:19] Marvel Nvidia these are all started with [77:22] tsmc Nvidia was started in [77:25] 1993 only Ever Raised $20 million $20 [77:28] million and never opened their own Fab I [77:31] believe 100% with tsmc wow well maybe [77:34] they have other foundaries too but the [77:36] vast majority of their business Jensen [77:37] talks about this Jensen H you know it [77:39] took him actually a little while to get [77:40] on Morris's radar but once he did the [77:44] vast majority of nvidia's chips tsmc [77:47] makes them Nvidia is what like a$ 350 [77:50] $400 billion market cap company now it's [77:53] wild and only raised 20 m million it's [77:56] like the AWS for chip companies never [77:59] would have been possible before nope so [78:01] this is what's super cool I don't think [78:03] Morris saw this like this even exceeds [78:05] his Wildest Dreams he was hoping for [78:07] this fabulous Market to take off but [78:11] this creates this insane flywheel for [78:14] tsmc so the fabulous Market starts to [78:17] grow which there're like seeding and [78:19] enabling it as that happens tsmc's [78:22] Revenue grows and because they have 50% [78:25] gross margins and 40% operating margins [78:28] they can take that profit and buy more [78:32] capex build more Fabs Advance the level [78:34] of their technology remember they were [78:36] starting from behind on technology [78:38] within about 10 years they catch up and [78:40] then they start to exceed everybody else [78:43] so as they push manufacturing process [78:45] technology forward they're manufacturing [78:47] better chips with smaller wave you know [78:49] process lengths they're enabling their [78:52] customers which are the fabulous [78:53] companies to get better and better [78:55] performance as they get better [78:56] performance The Fabulous companies can [78:58] address more of the market and more use [79:01] cases so their existing customers get [79:04] bigger and new Fus customers start which [79:07] gives them more Revenue which repeats [79:10] the whole cycle and you know it goes [79:12] slowly like any flywheel it takes a lot [79:15] of effort and a lot of time to start [79:16] turning it but fast forward to now so in [79:20] the early 2000s when tsmc finally caught [79:23] up to the Leading Edge level of [79:25] Technology with other semiconductor [79:27] companies there were 22 companies that [79:29] were at the Leading Edge I think it was [79:32] like I don't know let's call it 150 [79:34] nanometer process or something like that [79:35] at that point in time 22 and tsmc [79:38] finally broke in to the pack they were [79:39] one of the 22 by the late 2000s it gone [79:43] from 22 down to 14 that were at the [79:45] Leading Edge by the mid 2010s there are [79:48] six it's basically Samsung and tsmc [79:50] right today there are two at 5 nanometer [79:52] process is the current Leading Edge [79:55] only tsmc and Samsung Intel has been [79:57] trying to get there but they haven't [79:59] been able to they've fallen behind and [80:03] the next process is going to be three [80:05] nanometers tsmc is going to launch that [80:07] next year by the way just slip six [80:10] months ah interesting well Samsung has [80:12] already slipped to 2024 whoa so very [80:15] likely in the next process it's just [80:17] going to be tsmc which means that you [80:19] will see that on an Apple slide [80:21] somewhere announcing the next iPhone [80:23] talking about how it's a [80:25] they'll take all theit it and tsmc is [80:29] totally fine with that because their job [80:30] is not to Market it's to EMP their [80:32] customers this flywheel gets just unreal [80:37] what happens here they run the table on [80:39] the whole industry it is interesting the [80:41] industry went from vertical to [80:43] horizontally integrated where the very [80:45] best products in the market became [80:46] horizontally integrated and it's [80:48] interesting how I'm trying to figure out [80:50] what drove that because at some point I [80:53] guess there's a couple components to it [80:55] one is the speed at which More's law [80:59] happens makes it such that you can't be [81:02] good at everything you can't be good at [81:03] everything from Eda to making the [81:06] manufacturing equipment to running the [81:07] manufacturing process to designing the [81:09] chips you're not going to write your own [81:11] instruction set architecture people did [81:13] need to break into best of class Morris [81:14] has got this great quote about this that [81:16] I have in here so he says the [81:17] semiconductor business is like a [81:19] treadmill that speeds up all the time if [81:21] you can't keep up you fall off and [81:23] that's Mo's law from 22 down to two down [81:26] to one even when their competitors are [81:28] only doing the one thing the tsmc has [81:30] done like if you fall behind by a step [81:32] you're toast right and it's because [81:35] there's this big part of it that you're [81:37] talking about that hasn't come up on [81:39] other episodes because we tend not to [81:41] talk about companies that require a lot [81:43] of manufacturing prowess but in order to [81:46] stay on that treadmill the number of [81:48] tens of billions of dollars that you [81:50] need to be spending into capex is going [81:53] up so you need to be enormously [81:56] profitable so you can build the [81:59] factories for the Next Generation yeah I [82:01] mean well there's two things so yes that [82:04] is 100% true and the scale of this now I [82:06] mean tsmc just announced they're going [82:07] to spend a hundred billion dollar in [82:09] capex over the next three years $30 [82:11] billion this year 60 over the next two [82:14] and I bet that keeps going up so that's [82:16] a lot of billions you might even say [82:18] like this is so strategically important [82:19] and people are talking about this [82:21] certainly China's talking about this the [82:22] US government's now talking about this [82:24] governments might need to come in with a [82:26] bazooka of money and create other [82:30] options because almost all their [82:32] manufacturing is in Taiwan it's in this [82:33] strategically geopolitically challenged [82:35] location we need to re onshore some of [82:38] this in the US China of course wants [82:39] their own you can't just spend the money [82:41] and do this US government could come in [82:43] and say we're going to spend a trillion [82:44] dollars this year to do this they can't [82:45] do it because we're going to get to [82:48] Powers later but there's this marriage [82:50] of scale economies and process power [82:52] that tsmc like in this industry [82:55] there is no amount of money you could [82:56] spend to catch up next year you can't [82:59] because the engineering is so hard and [83:01] the learning curve takes decades to get [83:04] to this point I was listening to a [83:06] podcast at Bloomberg oddlot's podcast [83:07] about this where they were talking about [83:09] this and their reporter who covers tsmc [83:11] is Great China you know they asking the [83:13] question like the well China could just [83:15] spend a billion dollars and do this [83:16] create their own Fabs and they're doing [83:17] this what's the company called smic SM [83:20] smin y cuz basically tsmc seems to have [83:24] picked a di in the US and so with a [83:26] little bit ofing I'm sure from various [83:29] presidential administrations over the [83:30] last 5 years yeah the guy who covers [83:32] tsmc was like they can do that and they [83:34] are doing that but they wouldn't know [83:36] what to do with it and it's not because [83:37] they're dumb it's the hardest thing in [83:39] the world the hardest thing in the world [83:40] yes to do this stuff to make the [83:42] equipment that asml does and to [83:44] manufacture the way that tsmc does it is [83:46] the hardest thing to do in the world [83:48] yeah anybody else could get all the same [83:50] equipment from asml actually that's not [83:52] true oh no I'm saying even if you could [83:54] you wouldn't know what to do with it [83:55] right and it's not cuz you're dumb it's [83:57] there only like a small number of people [83:59] in the world that can operate this stuff [84:00] all right I'm jumping out of my seat [84:01] here so I'm going to do the asml thing [84:03] now so the reason that some people can't [84:06] get their hands on the asml equipment is [84:07] because the Netherlands did not renew [84:11] their trade agreement with China also [84:14] likely it has been reported that [84:16] probably that is because of us prodding [84:19] to say hey these pieces of equipment [84:21] you're making seem pretty specialized [84:22] you're the only people in the world who [84:23] can do it and makes the most Cutting [84:25] Edge semiconductor manufacturing [84:28] technology uh maybe let's not sell that [84:33] to smic in China and so they're not [84:36] doing that now you might say oh come on [84:38] how hard can this stuff be well these [84:41] machines tell us then what these [84:43] machines do well first of all they cost [84:44] $200 million for a machine that makes [84:48] the chips and that's going to go up to [84:51] like $300 million and by the way on a [84:53] lot of this we have a lot of thank yous [84:55] for John Bathgate and Britain John's [84:58] from the episode of the knowledge [84:59] project that they went on to talk about [85:01] a lot of this stuff it takes four [85:04] 747s to ship one of these machines so [85:08] you buy one your tsmc you buy one and it [85:10] arrives in of course the 747s then [85:13] there's a crew of asml employees on site [85:16] not only to assemble it but then to help [85:17] you run it so like you mentioned these [85:19] companies are deeply integrated with [85:21] each other to pull this off okay so what [85:24] is running it me what do these machines [85:25] do okay so it becomes exponentially [85:29] harder to manufacture chips the more [85:32] dense they are so David you mentioned [85:34] that 150 nanometers or so from several [85:37] years back and we know now that the m1's [85:39] are made on this 5 nmet process well the [85:43] wavelength of white light of regular [85:46] light is 193 nanet oh that seems like a [85:49] problem well it's certainly wide but you [85:52] know we're humans we come up with clever [85:53] Solutions we can solve this and so you [85:55] shoot it through a lens and maybe you [85:57] shoot it through some water like a laser [86:00] well not yet but even that really only [86:05] gets us to like 11 nanometers so how the [86:08] heck are we supposed to make these chips [86:09] where the transistors are ostensibly [86:12] only a five nanometers apart when what [86:15] we've done to date shooting through [86:17] lenses and shooting through water gets [86:19] us to 11 nomers well okay so this is [86:23] crazy you have to create a plasma so [86:27] what they do and this is called Extreme [86:29] ultraviolet lighter euv this is a [86:32] process that is just wild on one side of [86:35] the machine you drop molten tin on the [86:38] other side of the machine you then hit [86:39] it with a highly specialized laser you [86:42] perfectly pulse them it explodes into a [86:45] plasma which creates extreme ultraviolet [86:47] light now of course this is hard enough [86:49] to do as you can imagine how that might [86:52] work but you actually have to do that [86:54] 50,000 times per second yeah and what I [86:57] read is that the accuracy with which [86:59] that laser needs to hit the drop of [87:01] molten tin is more precise than the [87:04] calculations to send the Apollo missions [87:06] to the mo and you got to do that 50,000 [87:08] times a second [87:10] unbelievable now of course think a [87:12] little bit more about this well wait a [87:14] minute that wavelength is so small we're [87:17] going you know shy of 11 nanometers here [87:19] we're going to 5 nanm 3 NM that actually [87:22] it is absorbed by all known mirrors [87:25] which were used to reflecting light but [87:27] they don't reflect this light cuz the [87:28] wavelength is so small so part of this [87:31] process involves reflecting it a bunch [87:33] of times like 20 or something before [87:36] etching the [87:37] Silicon so what do we do well asml [87:40] actually needed to invent a new type of [87:42] mirror to do this and they also needed a [87:45] contract with a German company to make [87:47] this special type of laser which is the [87:48] only known company in the world capable [87:51] of making it this is crazy hard stuff [87:54] they only make 50 of these machines per [87:57] year or so they used to have competitors [88:00] like Nikon used to compete with asml on [88:02] this but it's too hard they gave up [88:04] that's how hard extreme ultraviolet [88:07] lithography is and of course we haven't [88:09] talked a lot about this and I think it's [88:11] outside the scope of the show but just [88:12] to overly simplify lithography is kind [88:14] of the process of taking that silicon [88:17] wafer and etching a design on it and if [88:19] we want to do that in smaller and [88:20] smaller ways we got to do with more and [88:22] more specialized equipment and at the [88:24] end of the day if you want to make the [88:26] M2 the M3 you know the a 18x bionic [88:31] whatever it's going to be called bionic [88:33] there is no other way to make it than [88:35] this extreme Cutting Edge Alchemy it [88:39] truly is alchemy so you know Ben so like [88:42] you're a government you want to throw [88:43] 100 billion you know like acquired is [88:45] doing really well like we're we're we're [88:46] on a tear here you know we've got power [88:48] we got brand power we got Network you [88:50] know economies we got our community like [88:52] we're doing well we should invest in [88:53] this [88:54] opportunity we should compete with tsmc [88:56] screw the governments we'll do it we've [88:57] got a couple hundred million dollars [88:59] we'll buy this stuff you know you have a [89:00] CS degree you know you're the more [89:02] technical one of you can run this stuff [89:03] right when we get the shipments from [89:05] asml you can make this happen would know [89:07] the first thing to do even if we could [89:09] invest the cash even if we could build [89:10] the facility even if we could buy the [89:12] machines which by the way that's going [89:14] to be hard because there's 50 some on [89:16] back order so like I can't even get has [89:19] Ed out all of them for years it takes [89:22] people who have done the most Advanced [89:24] Manufacturing in the world ever in [89:27] history in order to know how to do the [89:29] next version of it and this is why tsmc [89:32] has 40% operating margins it's [89:34] crazy totally crazy I'm just like in awe [89:38] of this completely okay so a little [89:40] while back before we get totally geeked [89:43] out on that which was awesome you said [89:46] something like how do we get this [89:48] flywheel effect you know it's great but [89:49] like how do we really get from tsmc [89:52] started taking the drgs from the idea [89:54] then the fabulous companies came along [89:55] how do we get from there to like now [89:58] there's another really important chapter [90:00] here and you're going to flash us [90:01] forward from like [90:03] 9395 to like 2010ish is that what's [90:06] about Happ 2008 well first we we'll stop [90:08] in 2005 so 2005 you know things are [90:10] going well better than Morris ever [90:12] imagined these fabulous companies are [90:14] getting started nvidia's killing I mean [90:16] I was making gaming PCs at the time I [90:18] wanted those Nvidia gpus but Nvidia [90:20] wasn't a top 20 stock in the world no I [90:23] mean was like Nvidia come on right real [90:26] men have Fabs like okay maybe we were [90:29] beyond that part but they were making [90:31] gpus like Nvidia stock tracked whether [90:35] they won the next Sony contract for the [90:37] next PlayStation or the Xbox like that [90:39] was the market for the market right you [90:41] know great Market but it's not what [90:42] we're talking about it's not about [90:43] machine learning it's not about crypto [90:46] it's like is the next PlayStation going [90:47] to include your chip or not totally but [90:50] still great for tsmc it's awesome 2005 [90:53] Morris is 7 4 years old he's like all [90:56] right I did it I've been buying tsmc [90:59] stock with my own money it's done well [91:01] enough I don't really need to work [91:02] anyway I'm gonna call it ready to retire [91:06] ready to right off into the sunset he [91:08] hands the Reigns of tsmc over to his [91:10] longtime Lieutenant Rick Sai and he [91:13] retires he spends a couple years he's [91:15] just chilling I don't know what he's [91:16] doing he's he loves literature he's like [91:17] reading all sorts of stuff he's on his [91:19] second marriage which he credits his [91:21] second wife for really reinvigorating [91:23] him and firing him then its summer of [91:26] 2009 by the way that's right around the [91:28] time that people were starting to [91:29] speculate that euv might work all this [91:32] had been kind of an idea to this point [91:34] they been science projects before oh [91:36] cool I didn't realize that y oh well [91:38] this is going to make what happens even [91:40] more sense the financial crisis had [91:43] happened in [91:44] 2008 and you know chaos everywhere we've [91:47] talked about it lots on this show [91:49] surprise press conference tsmc summer [91:52] 2009 they announced [91:54] that Morris is returning to lead tsmc's [91:58] CEO Rick is out Morris is coming back [92:01] for the third Act of his career I don't [92:03] even know what number he's wearing he's [92:04] not 45 because that was you know the [92:06] second act you like Jordan he's Beyond [92:08] Jordan at this point coming back he's [92:10] going to be CEO again at age 78 Rick [92:13] would actually have a second act himself [92:15] do you know what Rick is doing now no [92:17] Rick is CEO of Medi Tech which SP out [92:19] ofc so like he's doing fine Rick's doing [92:21] great but Morris comes back why does [92:23] Morris back but this is heralded is like [92:25] kind of a botched transition right yeah [92:28] well there's a lot of stuff going on [92:29] like from Morris to Rick people kind of [92:31] viewed it as like you didn't really do a [92:33] great job bringing in the next CEO of [92:35] the company maybe I don't know enough to [92:38] say okay I think maybe but also like [92:40] there's a lot going on at this moment in [92:41] time so the financial crisis y That's [92:43] like a crisis that's affecting everybody [92:46] so that's one thing but the other thing [92:48] so in the press release there's a quote [92:50] from Morris and he says one this move [92:53] will not affect tsmc's fighting spirit [92:56] and is likely to Spur greater intensity [92:58] but two he says that he sees quote [93:01] golden opportunities ahead what are [93:04] these golden opportunities that he's [93:05] referring to it's 2009 mobile right the [93:09] smartphone 2007 in July the iPhone comes [93:12] out 2008 the iPhone 3G comes out with [93:17] the app store for the first time the SDK [93:19] all these developers building for it but [93:21] of course an Android comes out in 2008 [93:24] yep Apple had to this point while [93:27] building you know this operating system [93:29] this scaled down version of os 10 it's [93:31] Unix but they weren't designing their [93:34] own chips they just used an [93:35] off-the-shelf Samsung chip they got it [93:37] right with saying like hey we got to use [93:38] arm in these things because we need a [93:40] really low power device so they've done [93:43] actual Gods work and Magic to be able to [93:46] bring a PC x86 operating system create a [93:49] sub operating system comp totally that [93:51] runs on arm yep Mir [93:54] but of course it's an off the-shelf [93:55] Samsung processor totally well even [93:57] that's great for tsmc like you know [93:59] Intel's not making that okay so that's [94:00] one we're going talk more about that in [94:01] a sec but we should say and Samsung also [94:04] fabbed it because Samsung is both a ship [94:07] designer and [94:08] aurer on mobile the previous whole [94:12] Paradigm of computing and silicon [94:14] everything was PC it was like stuff [94:15] plugged into a wall yeah it was Intel it [94:17] was x86 and like yeah tsmc could now [94:21] access some of that cuz AMD went [94:22] fabulous but come on but now all of the [94:26] leading companies that are going to make [94:27] silicon for design armed companies you [94:30] know Qualcomm broadcom mediatech Apple [94:32] who all are fabulous all are fabulous [94:35] okay so that's a big opportunity and [94:37] guess who knows all of those people [94:39] Morris and we should say too 2009 was an [94:41] interesting Tipping Point because if [94:42] you'll remember back to the 2007 [94:45] introduction of the iPhone Steve Jobs [94:47] has a slide where he says their hope [94:50] their goal is to get 1% of the existing [94:53] smartphone market so Apple had no notion [94:56] I mean the Google had no notion of how [94:58] big smartphones were about to become in [95:00] 2009 we're starting to see I think the [95:02] iPhone 4 came out we're starting to see [95:05] a ton of different oems making Android [95:07] phones you're moving into this era where [95:09] everyone's looking at each other going [95:11] oh this might actually be the next [95:12] Computing Paradigm yeah yeah yeah well [95:14] it was half of the next Computing [95:15] Paradigm remember back this is when I [95:17] started in BC there were two waves that [95:20] everybody was talking M and Social [95:22] Mobile and on the consumer side [95:24] everything shifting to mobile that was [95:26] what happened bring your own device well [95:28] sort of you're on the right track what [95:29] happened in the Enterprise the cloud the [95:32] cloud the cloud right so you got mobile [95:33] and you got cloud and it's so simplistic [95:36] but those are the two things that drove [95:39] trillions of dollars a market cap over [95:41] the next decade well what's the cloud so [95:43] at first the cloud is good for Intel [95:45] right x86 you're putting CPUs in the [95:47] cloud Amazon's buying lots of dude the [95:50] Cloud's the best thing that ever [95:50] happened for Intel those are incredible [95:53] server architecture it was the best [95:55] thing that ever happened to Intel but as [95:58] the cloud [96:00] progressed and Computing workloads [96:02] progressed the CPU became a lot less [96:06] important like AI started becoming a [96:08] thing CPU like yeah May mean maybe you [96:11] need some of that maybe you'll use Intel [96:13] maybe you'll use Arm like whatever but [96:15] what really matters well the majority of [96:18] cloud workloads are still on CPUs okay [96:20] fine fine but you're right future [96:21] looking why is NVIDIA now a 400$ 3400 [96:24] billion market cap company it's not [96:26] because of the PlayStation and it's [96:27] bigger than Intel right and nvidia's 2x [96:30] Intel's market cap something like that [96:33] yeah the notion of chips that are really [96:36] good at parallelized processing which is [96:39] gpus and matrix multiplication [96:42] effectively Vector Math versus the CPU [96:44] which are sort of these general purpose [96:46] workhorses built for the operating [96:48] system that runs on your computer super [96:50] good for seral of course there's like [96:53] multi-or you know there's 64 cores on a [96:55] CPU now so they're good at [96:56] parallelization too but all this stuff [96:59] especially machine learning is gpus it's [97:02] gpus and specialized like the Tesla Dojo [97:04] stuff that's not x86 oh yeah we're in [97:07] this I mean the other thing that [97:09] foundries enabled the fabulous era [97:11] enabled is the custom chip like [97:13] everybody's building custom chips for [97:15] all sorts of things yep so you got these [97:18] two big golden opportunities that are [97:21] coming online and Morris is like like I [97:24] got this and we should say we should [97:26] clarify too I think [97:28] Tesla uses Samsung oh interesting I [97:31] didn't not tsmc or at least for part of [97:33] it and I think they actually even Fab [97:35] their chips in Austin in the US really [97:38] yeah so this can't imagine that's going [97:39] to last this is like the beginning of [97:41] the like what everyone's sort of hoping [97:43] for in the US is this like return to [97:45] American manufacturing of chips they're [97:48] going to have to go to tsmc though in [97:49] the Next Generation CU you want three [97:52] nanometer it depends depends I mean it [97:54] depends what the workloads are and if [97:55] you I guess so what depends what you [97:57] need well anyway point is Intel's [98:00] dominance is over and the index on all [98:04] that's going to take over is tsmc yep [98:07] and Morris riding back in he comes in he [98:12] gets these deals done so like the Apple [98:14] deal 2012 Morris Chang 78 80 years old [98:19] and I think the Apple rep on that was [98:21] Jeff Williams the classic Cooks Tim Cook [98:25] that's right I think there was something [98:26] where it was even like a one went over [98:28] to the other's house for dinner or [98:30] something and it was like a living room [98:31] conversation to Ink the deal for uh hey [98:35] we bought this company PA semi we've [98:37] been designing our own chip architecture [98:39] in house we're going to launch I think [98:41] it was the A4 yeah I think that was the [98:43] first one first one and it was Apple [98:46] basically saying we think a lot of [98:47] people are going to buy a lot of iPhones [98:49] in the future and we are competing [98:52] head-to-head with Samsung because [98:54] they're a company that is not clear on [98:56] strategy they have a consumer angle here [98:59] with the Galaxy phones they think [99:01] they're also kind of a Foundry and jobs [99:04] hated Samsung famously right what did he [99:06] call them he called them some derogatory [99:07] term uh well there's been a few [99:10] interesting things there was Steve Jobs [99:12] saying he was gonna wage thermonuclear [99:15] war that was on Google right I think [99:17] that was Google but he had some like oh [99:19] Samsung they're just like copy like [99:21] something that like really put them in [99:22] it was about the lawsuits it was like [99:24] when they kept stealing Apple's designs [99:27] yeah and then there was something else [99:29] where someone this is later but Tim Cook [99:32] read the quote on stage about it being a [99:33] toxic hell [99:36] stew well doesn't get any better than [99:39] that no but Bloomberg reported that it [99:42] was a really big risk for both companies [99:44] both apple and tsmc apple was relying on [99:47] a company that was then seen as an also [99:50] ran and the quote is I think this [99:53] actually Jeff Williams if we were to bet [99:55] heavily on tsmc there would be no backup [99:58] plan and for tsmc it meant an initial [100:01] investment of $9 billion Fabs are [100:04] expensive to build and devoting 6,000 [100:07] employees to building a dedicated plant [100:09] for Apple in just 11 months it took [100:12] several years before it even began [100:13] producing the chips so that was in 2010 [100:16] and then I think 2012 was the launch of [100:18] the A4 designed by Apple built on the PA [100:21] semi acquisition and of course fabbed by [100:24] tsmc and I think it wasn't until the [100:26] iPhone 6 which was what 201 14 2013 [100:30] something like that that they were [100:31] solely tsmc I think so and that was like [100:33] the huge hit product cuz remember the [100:36] six was when they first increased the [100:37] screen size and those things flew off [100:40] the shelves I'm pretty sure some iPhones [100:42] had Samsung fabed yeah a4s and fives in [100:46] them and some had dsmc fabbed ones but I [100:48] think by the six all iPhones were huge [100:50] winners but I think the six was like [100:53] mega mega winner and I think that was [100:55] all tsmc huh 9 billion of manufacturing [100:58] capacity just for a deal with one [100:59] company paid off that was a bet the farm [101:02] deal and kind of like something only [101:04] Morris could do totally I mean it really [101:06] speaks to founder gravitas yeah even if [101:09] he had no equity as a Founder no equity [101:12] that he didn't buy so after getting that [101:15] deal done in [101:17] 2013 Morris steps down as CEO again but [101:20] he stays on as chairman and then finally [101:24] once it all plays out and tsmc is on top [101:26] in June of 2018 Morris retires [101:29] presumably for real he even stepped down [101:31] from the chairman role this time fully [101:33] retires from chairman at age [101:35] 86 oh my God crazy wow so that was 2018 [101:40] so I mean let's talk about now so 2020 [101:44] tsmc we've alluded to this operating [101:47] profit of $20 billion on $48 billion of [101:51] Revenue they took 17 of the $20 billion [101:54] in operating profit and plowed it all [101:55] back into capex last year in 2020 [101:59] beginning of this year January [102:01] 2021 they give guidance that they will [102:04] raise capex from 17 last year in 2020 to [102:06] 25 to 28 billion in [102:09] 20121 in April of this year 2021 they [102:12] rais it again to 30 billion forecast for [102:15] the year and 100 billion over the next 3 [102:19] years that's like the real shot across [102:21] the bow that everybody wakes up the [102:23] financial markets wake up and they're [102:24] like holy crap tsmc has cornered the [102:27] market even Samsung's not going to be [102:28] able to keep up with this it's wild all [102:31] right listeners it is time to talk about [102:34] one of our favorite companies stat Sig [102:37] it's funny David stat Sig has gone from [102:38] this little startup when we first [102:40] started working with them a couple years [102:41] ago to this total Powerhouse now I know [102:44] it's wild I was looking it up and they [102:46] have added all these customers since we [102:48] started working together open Ai figma [102:51] atlassian versel No and tons more at [102:54] this point if there's a growth stage [102:56] tech company out there there's a pretty [102:57] good chance they're using stats IG yep [102:59] so listeners if you are unfamiliar with [103:01] stat Zig they basically took what was [103:03] the standard product infrastructure at [103:06] every big tech company and they built it [103:08] as a standalone company this includes [103:10] Advanced experimentation tools AB [103:13] testing feature Flags product analytics [103:15] session replays and more so if you're [103:17] building the next great software company [103:19] this sort of infrastructure is essential [103:21] because it allows your product and [103:23] engineering teams to release things [103:25] quickly measure the impact of them and [103:27] track progress over time totally so I [103:30] mean as we've talked about on the show [103:31] Forever at companies like Facebook or [103:33] Netflix data was just a part of how [103:35] everything was built which contributed [103:37] to all the crazy Bottoms Up organic [103:39] growth that they had now with stats Sig [103:41] you can get that from day one at your [103:42] startup and today they're not only [103:44] trusted by startups but also by born [103:47] mature Enterprises like Bloomberg and [103:49] Microsoft and Electronic Arts turns out [103:51] that a single system for datadriven [103:53] decisions is useful at any scale yeah [103:56] and by the way the scale they're [103:57] operating at is completely insane they [103:59] process over 2 trillion events per day [104:02] now by the way David this is updated the [104:03] last I checked it was 1 trillion and [104:05] then this morning I pulled it up two [104:07] trillion and they handle releases to [104:09] billions of end users if you're [104:11] listening to this podcast and you've [104:12] used software in the last few years [104:14] there is a very good chance you've been [104:16] a part of many experiments orchestrated [104:18] by stat Sig yeah it's just awesome and [104:21] as they've gone up Market they've also [104:22] started to offer some interesting [104:24] deployment models like being able to run [104:26] the whole thing natively inside your [104:27] existing data warehouse or just using [104:29] stat Sig's fully hosted solution if you [104:31] want to leverage stat Sig to grow your [104:34] business there are a bunch of great ways [104:35] to get started stat Sig has a very [104:37] generous free tier for small companies a [104:39] startup program with a billion free [104:42] events that's $50,000 in value and [104:45] significant discounts for Enterprise [104:46] customers to get started go to stat [104:48] sig.com Acquired and just tell them that [104:51] Ben and David sent you thank you stat [104:53] sake so more on today David tsmc today [104:58] well okay so speaking of data I think [105:01] this is the data point that really kind [105:02] of says everything so since the first [105:05] IPO in Taiwan in 1994 [105:08] tsmc has had compound annual revenue [105:11] growth of [105:13] 17.4% for 27 years Revenue growth [105:18] 17.4% compounded for 27 years now the [105:23] irr the equivalent on valuation on [105:25] market cap so it was 4 billion market [105:27] cap at the Taiwan IPO in 1994 today it [105:30] is 550 billion so that is a [105:35] 19.9% irr starting from a $4 billion [105:39] base over the last 27 years so 20% irr [105:42] over 27 years incredible by any means [105:46] starting from a $4 billion base it is [105:50] now currently as we record the ninth [105:51] most valuable company in the world and I [105:54] think other than Saudi aramco it is the [105:57] only company in the top 10 that we [106:00] haven't done unacquired oh interesting [106:03] yeah the US oil companies are no longer [106:05] corre in the top 10 but Burk that might [106:07] be foreshadowing some future episodes [106:09] this season oh I mean they're in [106:12] hallowed grounds at this point the other [106:14] thing that just talking about financials [106:15] today so crazy that they grew 31% in [106:19] revenue from 2019 to 2020 they double [106:23] their ker from 2019 to 2020 yeah it's [106:27] nuts talk about accelerating growth okay [106:30] so in 2020 their adjusted net income was [106:33] $17 billion how are they going to go [106:35] spend a hundred billion over three years [106:38] is that going to be out of profits of [106:42] each of those years or do you think [106:43] they're doing some kind of financing I [106:45] don't know I don't actually don't know [106:46] if they've done any financing I'm quite [106:49] confident they'll make enough profit to [106:51] fund it organically [106:53] because big news just in the past week [106:56] they started this a little early in the [106:57] year but now they're really doing [106:58] they're getting away from Morris's [107:00] second big innovation of reducing prices [107:02] in fact I think they're going to raise [107:04] prices this year by 20% so the first [107:07] announcement a couple months ago was [107:08] they're not going to cut [107:10] prices and then they just announced [107:12] they're going to raise prices nobody's [107:13] ever done this since the proris days [107:15] pricing power in action totally I mean [107:18] what a clearer picture of how they have [107:20] taken a commodity business and turned it [107:24] into I mean this has got to be one of [107:26] the biggest Moes of all time totally I [107:29] mean they've got 28 billion of cash and [107:31] cash equivalents on the balance sheet [107:33] and they're going to use that and all [107:35] the cash that they generate from their [107:36] operations to plow directly back in to [107:39] making sure that everybody else is five [107:41] plus years behind unbelievable the other [107:44] thing is that they already are the [107:46] largest they have over 50% of the market [107:50] for foundaries for all contra track [107:53] manufacturing of chips and like 95 plus% [107:57] of the profit correct I thought where [107:59] you were going with that it is also true [108:01] that they have 90% market share on the [108:04] current generation like the Leading Edge [108:05] chips yeah in the 5 nanometer Samsung [108:08] has like 5 10% market share and tsmc has [108:12] 90 plus percent in many ways going to [108:15] 100 they're the apple of semiconductors [108:19] they don't have all the market share but [108:20] they have all the most profitable market [108:21] share yeah exactly they are the iPhone [108:24] of semiconductors like you can still buy [108:27] previous generation you know worse [108:29] technology from and for plenty of you [108:31] know on the odd Lots podcast this was [108:33] actually they talked about sort of like [108:34] the be case going forward for tsmc and [108:37] one potential one is that oh well the [108:39] processing power is so good that you're [108:41] not going to need the Leading Edge [108:43] anymore I find that a really weak [108:45] argument like you always need the [108:47] Leading Edge you think Tesla doesn't [108:48] want the Leading Edge totally you think [108:50] Apple doesn't want the Leading Edge [108:51] software will always match the [108:52] complexity on the most advanced Hardware [108:54] it can run on totally which is why like [108:56] I love when people are like apple [108:57] slowing down my computer I'm like yes [108:59] I'm sure that's what's happening they [109:00] wrote special code that they're putting [109:02] on there to make the consumer no it's [109:04] because every piece of software just [109:06] always assumes that it has the most [109:08] advanced processor on Earth and it [109:10] always gets to developers sure they test [109:13] on Two And three-year-old equipment but [109:15] no one's making sure that the six and [109:16] seven-year-old laptops are as performant [109:19] software designed for the current [109:20] generation of Hardware think they Google [109:21] and Amazon are going to be like now [109:23] we're good right hell no it actually is [109:27] worth touching on there's one other [109:28] interesting bit about this five [109:30] nanometer process which first of all is [109:33] a marketing name at this point what it [109:35] used to originally referred to was the [109:37] length on the gate on the transistor at [109:40] this point it's not exactly 5 nanometers [109:43] and the additional performance is not [109:46] going to come from you know making [109:48] smaller Gates here's the interesting [109:51] thing though you actually can't put [109:53] these transistors much closer to each [109:54] other so if you think about silicon [109:56] atoms that are between the transistors [109:59] you can only fit five of them in a [110:00] nanometer so in a three nanometer [110:02] process sure [110:03] it's right like at some point you cannot [110:07] subdivide silicon anymore so either we [110:09] need to change the substrate or the [110:12] Innovations are sort of going to come [110:13] from elsewhere which has always been the [110:16] case like Mor's law was technically the [110:19] doubling of the number of transistors on [110:20] an integrated circuit now it comes from [110:22] multicore it comes from all the other [110:25] advancements of figuring out how to make [110:27] chips do more stuff faster yep that I [110:31] think is going to keep going and I think [110:34] it's going to keep being expensive and [110:35] getting more expensive and I think tsmc [110:38] is the only company that's going to be [110:39] able to keep up at the Leading Edge yeah [110:43] do you know David about more Second Law [110:45] oh no I don't so everyone knows about [110:47] Mor's law but there's this second one [110:50] which is also known as rocks law after [110:52] arur rock yeah OG it states that the [110:56] cost of a semiconductor chip fabrication [110:58] plant doubles every four [111:01] years so with Fabs today costing 1520 [111:05] billion doar I don't know that that's [111:07] proven exactly true but it's certainly [111:09] well shoot if we just look at tsmc's [111:11] capex forecasting they're going from 17 [111:14] to 30 yeah to 60 over two years so [111:17] that's way faster than four years so the [111:19] interesting thing is when you combine [111:21] these two things the Mo's law and Mo's [111:23] Second Law it implies that the leading [111:27] company the most profitable company will [111:30] become a monopoly win or take all there [111:32] you go and it's fascinating that both of [111:35] these things these laws aren't actually [111:37] in conflict because Mo's law is about [111:39] effectively when you really look at it [111:41] from a financial perspective operating [111:43] expenses when producing at scale and [111:45] Rock's law is about The Upfront Capital [111:47] expenditures to enable all that [111:49] production so it's everything we talk [111:50] about on the show it's being able to P [111:52] investment into fixed cost as much as [111:55] possible at huge scale in order to [111:57] realize the benefits of making as many [112:01] of the thing as humanly possible at [112:03] global scale and tsmc interestingly is [112:07] the most perfect example of this and I [112:10] say interestingly because we almost [112:11] always talk about operating leverage and [112:14] scale in the context of software on the [112:17] internet this is how Venture Capital [112:20] started because actually manufacturing [112:23] chips the operating leverage that comes [112:25] from huge amount of fixed costs into [112:27] foundaries to make chips and then [112:30] hopefully be very profitable 50% gross [112:33] margin on those chips Venture Capital [112:35] financing was built for that for [112:39] semiconductors and it just so happened [112:42] to work just as well or even better with [112:44] software on the internet even better in [112:46] the notion that gross margins of [112:48] software could be 80 to 90% not 50% but [112:51] I would back that down because it [112:53] doesn't have the sort of Mo [112:54] defensibility characteristics that being [112:56] able to plow your copex into [112:58] manufacturing capability does yeah [113:01] should we do power now absolutely let's [113:03] do it let's do it so for folks new to [113:06] the show this is one of the discussion [113:09] topics we do for every episode is we go [113:10] through Hamilton helmer's excellent [113:13] seven Powers the best business Theory [113:15] Book totally we've had Hamilton on the [113:17] show he's amazing go read the book If [113:19] you haven't he identifies seven powers [113:22] essentially sources of Defense ability [113:25] which he defines as long-term [113:27] differential profit margins versus your [113:29] competitors as we've been talking about [113:31] on the whole show yep the seven that he [113:33] identifies are counter positioning scale [113:36] economies switching costs Network [113:38] economies process power branding and [113:40] cornered resources and we almost always [113:43] talk about Network economies we talk [113:44] about counter positioning on this [113:47] show we talk about branding sometimes we [113:49] talk about branding we're talking about [113:50] none of those this time yeah we [113:52] sometimes talk about scale economies [113:53] which we're going to definitely talk [113:54] about here but I think we're going to [113:55] have our first process power if uh I'm [113:58] going to forecast but let's start let's [114:00] go down the list counter [114:02] positioning I mean like maybe you could [114:04] when they were starting and in [114:07] particular would the incumbents have [114:10] started with the exact business model no [114:13] because their profit Center was the [114:15] integration that all the margin you get [114:17] of integrating design and Manufacturing [114:19] and by saying nope we're going to be a [114:21] pure playay manufacturer [114:23] tsmc theoretically was saying no we're [114:25] going to take less gross margin and [114:27] we're just going to make it up in volume [114:28] I'm actually not sure it played out that [114:30] way I think they have more do you know [114:31] what Intel's gross margins are I [114:33] actually don't know I would suspect [114:35] they're higher but I don't know right [114:38] yeah there was counter positioning here [114:39] I don't think I said this when we were [114:41] going through it but before tsmc and the [114:44] Pure Play Foundry model if you were [114:47] either a fabulous company one of the [114:49] very very few or you were another IDM [114:52] and you were trying to get some excess [114:54] capacity rent it from another IDM most [114:56] of the idms be like okay you strong AR [114:59] them you got a great strategic [115:00] relationship they'll give you some [115:01] capacity but they also demanded the [115:04] right to Market your products under [115:06] their brand too so like which obviously [115:09] tsmc wasn't GNA so yeah there was [115:11] counter positioning like the idms no way [115:13] they were going to do what tsmc was [115:14] going to do right huh okay scale [115:17] economies absolutely that is the biggest [115:20] if it's one of top two with process [115:21] power yep in my opinion switching costs [115:25] well it's funny now there are huge [115:26] switching you can't switch off tsmc no [115:29] unless you're going to stop being on the [115:30] Leading Edge if you're going to change [115:32] from being a phone company to an [115:34] automotive company you can switch off of [115:35] them well I think it's even deeper than [115:37] that again we haven't gone listeners [115:39] probably think we've gone deep [115:40] technically on this episode we haven't [115:41] even scratched the surface totally but [115:43] yes if you want the Leading Edge now you [115:45] got to be [115:46] tsmc but you got to be so integrated [115:50] with tsmc to do this say you want to [115:52] switch to Global foundaries or you know [115:54] one of the other competitors out there [115:56] of which there are a few you can't just [115:58] call up Global foundaries and be like [116:00] hey I'm porting over expect my business [116:03] on Monday it takes years because you're [116:05] so deeply integrated with the process [116:09] yeah so yeah big switching costs Network [116:12] economies it's not really worth talking [116:14] about not in the traditional sense you [116:17] know this is not Facebook here and [116:19] certainly none of tsmc's customer [116:23] really benefit from other customers [116:25] being on it no I do think there actually [116:27] is I don't think Hamilton captures this [116:30] in his seven powers and I don't know if [116:32] he would consider this one but there is [116:34] like an ecosystem aspect here because [116:36] the EDM companies and the IP companies [116:39] are so deeply integrated with [116:41] tsmc if you want to be using arm for [116:44] instance or you know they're kind of the [116:46] best integrated with now I don't think [116:48] that's Network economies that is kind of [116:50] like this ecosystem thing TSM actually [116:52] has a name for this they call it like [116:53] the open Innovation something or other [116:56] you know it's some corporate name but it [116:58] means this I do wonder if it's actually [117:00] worse for a lot of people that apple is [117:01] a tsmc customer because who else has [117:04] access to the 5 nanometer process right [117:06] now they're going to take as much as [117:08] they yeah good point process power yes I [117:13] think Pixar this is the first time we've [117:16] really although we weren't doing seven [117:18] powers during yeah to me this is the [117:20] clearest example I could ever imagine of [117:22] process power it takes all 40 Years of [117:26] tsmc's History to have arrived at where [117:29] they are today and even if 10 people [117:31] left and tried to start the next tsmc to [117:34] be able to create what they've created [117:37] at this point from scratch virtually [117:39] impossible all of their IP all of their [117:42] you know people all of their knowhow all [117:44] of their relationships with asml and the [117:47] like yeah no amount of money can [117:49] replicate it I think the only thing that [117:52] will unseat tsmc is a complete Paradigm [117:56] Shift yeah something where like what [117:59] mobile did to desktop if there's [118:01] something where the computer required in [118:03] the future is unable to be provided by [118:06] anything that tsmc is good at today if [118:08] all the crazy laser molten tin asml [118:11] stuff we were talking about if all of a [118:14] sudden there's discovered a new either [118:16] different or way cheaper Quantum [118:17] Computing yeah way to do this then that [118:20] kind of could reset the playing field [118:22] yeah yeah yeah totally but even little [118:24] shifts I bet they'd be fine if everyone [118:25] figured out that like hey silicon's not [118:28] the best substrate and we can figure out [118:29] a better substraight if there were like [118:31] an AWS moment oh which is funny because [118:33] tsmc is the AWS equivalent but where uh [118:36] something happened that just made it way [118:37] cheaper than it used to be you could now [118:40] get access to the technology and the [118:43] knowhow orders of magnitude cheaper than [118:45] it is now that would take away a big [118:47] part of yeah their power but I don't see [118:50] that happening no [118:52] absent a paradigm shift this is tsmc's [118:56] lose they're pretty much they're in the [118:58] groove so I think we should skip [119:01] branding in corner resource for now it's [119:03] not really worth talking about literally [119:04] they're antithetical to branding like [119:05] it's Apple's brand it's not TSC so this [119:08] is I think a good time to enter our [119:11] geopolitics discussion because I was [119:14] thinking about the other way that tsmc [119:16] could fail would be that China decides [119:19] the moment is right to go and assert our [119:21] force and and take over Taiwan depending [119:24] on how you see it either Annex Taiwan or [119:26] assert it's as always claimed [119:29] sovereignty in Taiwan Yes actually start [119:31] enforcing what has been right the whole [119:33] time I think as they would sort of say [119:34] yeah if they speaking in my casual tone [119:37] in English from America then doing all [119:40] this business with the West I have to [119:43] imagine that assuming that it didn't [119:45] start a full War like an actual World [119:47] War which it may then of course they [119:50] would start using all the tsmc [119:51] manufacturing capacity for all the [119:54] Chinese and and huawei's been a TS [119:57] custom cut off yeah so how do you [120:01] capture that in power what is the power [120:04] or maybe like let that's a risk that's [120:05] like a bare case right let's not get too [120:07] specific on this but maybe in a general [120:09] sense how do you capture the power that [120:13] a company has that comes from regulatory [120:17] environment where would that get [120:18] classified under that like they have a [120:21] lot of room to be operating safely um [120:25] maybe cornered resource I guess so [120:27] you're saying this is like an anti-power [120:28] for them this is a weakness exactly [120:31] exactly yeah I suppose all that matters [120:34] is things that you have that your direct [120:36] competitors don't so I in this strange [120:38] straw man that I'm putting together it [120:40] would really be about what if you were [120:43] located in a country that none of your [120:45] competitors were also domiciled in and [120:47] being there gave you some special [120:49] ability to be more profitable than [120:51] others which they had in the beginning [120:53] with the government of Taiwan basically [120:55] the mafia boss was like this is [120:57] happening we are going to strong arm all [120:59] the Business Leaders in the country to [121:00] investing in this we're gonna make sure [121:02] that this happens yeah okay well let's [121:05] put a pin in that because you're right [121:06] it turns out that it's actually not a [121:08] perfect power discussion but the [121:09] geopolitics thing is interesting well I [121:12] think it's the bare case right that to [121:14] me absent an enormous Computing paradigm [121:17] shift is the way that tsmc has an [121:20] enormous risk in the business yeah [121:22] totally which does make it kind of [121:24] surprising that you know they haven't [121:26] Diversified their geographical [121:28] operations very much so this is [121:30] interesting so they're facing a lot of [121:31] pressure for this they are spending I [121:35] think $12 [121:37] billion this year to start a plant in [121:41] Arizona which will not be the 3 [121:43] nanometer I don't even think it'll be [121:45] the 5 nanometer it's not their most [121:47] advanced manufacturing I think the US is [121:49] subsidizing in a big way I think it's [121:51] part of the Biden administration's most [121:52] recent Bill to try and bring some [121:54] semiconductor manufacturing here but [121:57] they're also starting a Fab in Japan [121:59] that came out on their last earnings [122:01] call so they're doing some and they have [122:03] operations in China I believe too Yep [122:05] they're doing some [122:06] diversification but I don't think it's [122:08] for this reason I think it's because [122:11] they're basically getting free money to [122:13] open Fabs and other places and Morris [122:16] has even made comments like I don't [122:19] think it makes any business sense for us [122:20] to have the Leading Edge in those [122:22] countries even though those countries [122:25] want us to have them there I think it [122:26] makes sense based on the ecosystem that [122:28] we've created in Taiwan to keep [122:30] operating it here so the question is if [122:33] it directly helps let's take the US for [122:35] example the US's prowess as a [122:39] semiconductor manufacturing force in the [122:42] world to have tsmc's Arizona plant or if [122:46] it's really just indirect and the idea [122:48] is like let's try this as a first stb [122:50] we'll get more people in the US familiar [122:54] with doing this again in case we need to [122:57] restore this yes yeah I mean this is a [123:01] scary scary future to contemplate and [123:03] hope to God it doesn't happen but really [123:06] the thought exercise here is what would [123:07] happen if China annexes yes Taiwan [123:10] tomorrow which is scary for a number of [123:12] reasons the smallest of which is this [123:14] corporate takeover it's scary for a lot [123:16] of people their lives yeah scary but I [123:19] wouldn't say it's the small like [123:20] everything imagine if we didn't have [123:22] access to Leading Ed semiconductors yeah [123:25] that's everything what part of our lives [123:27] do not run on right semic can't make [123:30] F150s right now basically all of our [123:32] technological progress would stop yeah [123:35] you're right so I think the question is [123:37] I don't know enough to answer this what [123:39] would happen would it be possible to [123:41] airlift the process power that tsmc has [123:44] physically out of Taiwan to somewhere [123:46] else you get all the people asml now [123:49] sends the stuff somewhere else you air [123:51] everybody out there's an [123:53] evacuation does the process power come [123:55] with it or not I don't know that's a [123:57] good question I mean if the Toyota [124:00] production system is an example where [124:02] Toyota tried to there was that factory [124:04] that joint venture with was it GM yeah [124:07] the numi plant that's now the Tesla plan [124:08] Inon with Toyota trying to replicate [124:12] their process somewhere else didn't work [124:15] now it wasn't under threat of War right [124:17] this one would need to it's actually a [124:19] good question if you think about the [124:21] US's [124:22] strategic defensive [124:24] weaknesses what's more important having [124:26] onshore semiconductor capability to [124:29] continue to advance technology in the [124:31] nation or Boeing our ability to build [124:33] which we've always held up as this [124:35] example of the US needs that to stay us [124:39] owned to stay operating to stay [124:41] profitable to stay prosperous because it [124:43] is a matter of the US way of life that [124:46] we're able to protect Boeing needs [124:48] semiconductors that's a great point so [124:51] we now outside our depth but is it [124:53] actually more important to have Cutting [124:56] Edge semiconductor capability here than [124:58] airplanes or any of the other sort of [125:00] Defense supply chain and you know maybe [125:02] the answer here is like Korea same [125:04] situation exists in Korea with samung [125:06] right like North Korea is right there I [125:07] been there I've been to North Korea like [125:09] I went to the DMZ like yeah it's so [125:11] weird it's like an amusement park weird [125:13] it's super super weird and bizarre but [125:16] yeah North Korea's right there you know [125:17] maybe it's the same like China's right [125:19] there but this isn't actually going to [125:20] happen but I don't know it feels in the [125:22] last year like the risk of it actually [125:24] happening has ratcheted up quite a bit I [125:27] think so I mean it's like globalization [125:29] as a whole it is in the best interest of [125:31] everyone to continue to share resources [125:34] to continue to entangle everything until [125:37] somebody decides that it's not and then [125:38] we have a big problem and hopefully for [125:41] lots of reasons it just continues to be [125:43] okay that tsmc is located on an island [125:45] that is of disputed claim yeah maybe the [125:48] best thing that could happen is my car [125:51] Val a while back was the book by the [125:53] Harvard chair of the Harvard [125:54] astrophysics Department about umu amua [125:57] that he postulates was an alien [125:59] spaceship maybe if we discover that [126:01] aliens are real that's going to be the [126:02] uniting force you know like all these [126:05] conflicts seem pretty Petty yeah I [126:07] wouldn't use that as an investment [126:09] thesis though no okay before we get into [126:13] Playbook and just hit some things that I [126:15] think we missed during the narrative or [126:17] at least didn't put a fine enough point [126:18] on in the narrative I have a what would [126:20] have happened otherwise that I want to [126:22] hit we haven't done this in a while no [126:23] we haven't and I'll just read this as a [126:25] direct quote from Bloomberg and there [126:27] were some awesome sources for this [126:28] episode all of which are linked in the [126:30] show notes in the mid 2000s as Apple Inc [126:33] Was preparing for the release of its new [126:34] smartphone Steve Jobs approached then [126:37] CEO of Intel otellini about providing [126:40] the chips for the iPhone Intel already [126:43] sold iPhone the processors that ran on [126:45] its Mac but add video to acquired so [126:48] that everybody can see the look on my [126:49] face right now I'm just like literally [126:51] got Fists in the air I'm so happy and [126:53] remember otellini was the guy that jobs [126:55] brought out on stage during the Intel [126:57] transition they were burying the power [126:59] PC to say this is the future this is the [127:01] partnership so okay but Jobs made what [127:06] otellini considered a lowall offer and [127:08] apple awarded the contract to Samsung it [127:12] later began designing the chips itself [127:14] eventually outsourc production at tsmc a [127:16] contract manufacturer in Taiwan that [127:18] been found blah blah blah blah blah so [127:20] what could have been went to int and [127:22] said do you want this contract CU they [127:24] were Partners on the Mac totally and [127:26] apparently it was less about the fact [127:29] that I'm sorry you want to use Arm what [127:32] no we're the company and it was more [127:34] about we felt it was a lowball offer [127:36] biggest strategic error of all time all [127:38] right I'm going to postulate A playbook [127:40] theme and I'm put forward as a post more [127:41] than a Playbook theme that what's the [127:43] you know like in Geometry there's like [127:45] laws right that are proved but then [127:46] there's like postulates that like you [127:47] can't prove them but our fundamental [127:49] understanding of the universe doesn't [127:50] work if they don't work [127:52] whatever that is AXS I don't know [127:53] whatever it is I'm going to put one of [127:54] those out there please [127:56] never make strategic decisions based on [128:01] economics this is prime example like the [128:03] number of we talk about this all the [128:05] time on this show VC's passing over [128:08] valuation on something you know uh [128:10] andreon getting cold feet about a $300 [128:12] million valuation on Uber y right this [128:15] Intel move passing on being partnering [128:18] with apple and maybe more specifically [128:20] than economics [128:22] cuz like you could imagine that you [128:23] would want to pass on this if intel [128:26] didn't get to the upside from the deal [128:28] assuming that the structure is right [128:30] than passing because a number is too low [128:32] in the structure or Ford Motor Company [128:35] not hiring Morris Chang over1 you know [128:39] like whatever like it's just H humans [128:40] are so prone to cutting off their noses [128:43] despite their faces so we already have [128:45] the Rosenthal doctrine of never bet [128:46] against the internet but now we have the [128:48] Rosenthal postulate which has never made [128:50] strategic decision ision bicing Bing [128:53] based on not economics but [128:55] pricing I like it I need to add a new [128:58] section to the acquired website all [129:01] right all right next on [129:04] playbook is another one on Intel fading [129:07] so it takes a very long time to become [129:09] irrelevant so despite Intel's stock [129:11] price being I think tsmc is like 2 and [129:14] a2x Intel stock price as a matter of [129:17] fact asml is actually larger than Intel [129:20] by market cap now they are the sole [129:22] source provider of one thing in the [129:24] value chain to mostly one company and [129:27] they're bigger than Intel now i' be [129:28] fascinated okay so public markets [129:31] investors who are listening shoot us a [129:33] DM in slack or post in general acquired [129:35] fm@ gmail.com or acquired FM whatever [129:38] Channel works for you or Twitter [129:39] whatever be super curious if you are [129:41] along this thesis that we're sort of [129:43] laying out on the show how are you [129:45] playing it between tsmc and asml which [129:49] is now Europe's Most Valuable company [129:51] right [129:51] I mean probably you just invest in both [129:53] but how do you think about that right [129:54] and what's the upand comer that's kind [129:56] of speculative at this point but could [129:58] be another puzzle piece here are you [130:00] also shorting Intel through all like [130:02] what do you what do you do here all [130:04] right so my point on Intel is it takes a [130:06] long time to become irrelevant they [130:08] still control 80% of the computer [130:11] processor market and they have an even [130:13] bigger share in servers so despite [130:16] everything we're saying it's still huge [130:18] workloads running on CPUs that are in [130:20] computers and on the the cloud pretty [130:22] big business yeah the majority of [130:24] workloads that are happening in the [130:25] cloud is not Tesla Dojo it's I don't [130:30] know some company that's not a tech [130:32] company somewhere in the world running [130:34] their Outlook server on Office 365 yep [130:37] Absolut doesn't need 5 nanometer process [130:40] two other Intel things one is that [130:43] indecision has been very tough on the [130:45] company Bob Swan who was the former CEO [130:48] started to prepare to Outsource Manu [130:51] facturing of Intel design chips to tsmc [130:55] like that I think even like two years [130:56] ago this was the plan they finally [130:58] decided thrown in the towel you know [131:01] greatest ship Manufacturing Company in [131:03] the world but real men are sensitive and [131:05] they talk about their feelings Bob Swan [131:08] is no longer the CEO of Intel and now in [131:11] a complete reversal their new CEO Pat [131:13] galinger wants to turn Intel into a [131:16] Foundry themselves by which other [131:20] fabulous companies can contract with [131:22] Intel to build maybe that's right but if [131:25] so they got to figure out and I think [131:27] they're thinking about this the right [131:28] way because they said it's going to be a [131:29] fully separate autonomous division they [131:31] got to run that like a completely [131:33] separate independent company of the rest [131:35] of Intel and if so I don't actually know [131:38] why Intel owns it yeah I mean well a [131:42] let's look at AMD here right they did [131:44] this they spun out their manufacturing [131:47] into Global foundaries which has been [131:49] good for Global foundaries and am like [131:51] foundaries is getting ready to IPO yep [131:54] yeah there probably the right strategic [131:55] decision but it's not going so well I [131:57] mean like it's going fine it's not tsmc [131:59] right it's going better probably than if [132:01] they had not done that but they're not a [132:02] winner here like tsmc is the winner yeah [132:06] I guess the Playbook theme there is [132:08] indecision is paralyzing I mean this [132:09] company has spun its Wheels One [132:11] Direction or the other and all it's done [132:12] is make itself deeper in the mud H I [132:14] just looked up I was trying to remember [132:16] this Ginger was the VMware CEO he [132:21] started his career at Intel and then [132:24] EMC and then EMC owned the majority of [132:27] VMware and became the CEO of VMware yeah [132:30] and he was the outside candidate to [132:33] replace bomber as Microsoft CE yeah y [132:37] yep huh you know I hear he's really [132:39] revered in the organization that people [132:41] think he's really gonna gonna make some [132:43] good change there we'll see the last [132:46] thing on Intel and it's funny this is [132:47] not the Intel episode but there's a [132:50] thing that happened here that is very [132:51] similar to the fact that Kodak [132:54] developed the digital camera first in [132:58] their lab they knew it they knew this [133:00] was the future and they didn't [133:01] commercialize it because it's impossible [133:04] to counterposition yourself because of [133:05] the innovator's Dilemma Intel actually [133:08] saw extreme UV lithography euv first so [133:14] intel was the biggest early investor in [133:18] euv committing more than 4 billion to it [133:20] in 20 2012 who it was slower than its [133:23] main Rivals and this is from The Wall [133:25] Street Journal in adopting the [133:26] technology and skeptical about whether [133:28] it would work eventually Intel [133:30] calculated that it was a sure bet to try [133:32] and improve existing ways of handling [133:35] lithography and of course where we are [133:38] today euv completely enabled the next [133:41] generation of chips to be built that [133:43] couldn't great argument an example for [133:45] why you need startups right right like [133:48] totally like yeah intel was there they [133:50] invested in it they saw it and they were [133:52] like they put 4 billion in and I think [133:55] even to this day there is not a shipping [133:57] Intel chip that was manufactured by [134:00] Intel using UV wow that's crazy you're [134:04] right it is the most perfect pure [134:06] example of the innovator's Dilemma in [134:08] action it's why you need startups yep [134:12] all right my next one is that if you're [134:14] only looking at the outcomes that [134:16] happened you cannot reverse engineer [134:18] what the probability that it would [134:20] happen is [134:21] and this is a very abstract way of me [134:25] saying the strategy of if you build it [134:27] they will come that Morris [134:30] implemented is a bad strategy and it [134:33] also worked right like if something [134:36] Valentin hated they would never invest [134:38] in developing a market that was like [134:40] rule number one we invest when the [134:42] market already exists not when we need [134:45] to develop it and this is like the [134:46] classic problem this is the knock up [134:48] here in Seattle there's a lot of people [134:49] spinning out of Microsoft starting [134:50] companies classically people coming out [134:52] of Microsoft would always want to build [134:54] platforms because Microsoft was the [134:56] platform company and they would always [134:58] have a small too small of an [135:00] understanding of the market of people [135:02] that wanted that platform today and they [135:03] assumed if you build it they would come [135:05] Morris was that exact problem and yet if [135:08] something is going to be true 10% of the [135:11] time and fail 90% of the time one out of [135:14] 10 times it's going to work and it may [135:16] have been the case I guess what I'm [135:17] saying here is if you're starting a [135:19] startup it's impossible to know if this [135:22] was actually a good strategy or if it [135:24] was a bad strategy that [135:25] probabilistically just happened to work [135:27] I mean that's the thing about startups [135:28] right is there are all these rules but [135:30] they can all be broken there is no [135:32] formula yep totally all right other [135:35] Playbook themes I just have one more [135:36] that again we talked about a bunch in [135:38] the episode but I want to highlight and [135:40] actually add one spin on here you know [135:42] the Jeff Bezos quote about AWS as a [135:44] startup anything that doesn't make your [135:46] beer taste better the analogy back to [135:48] German beer factories and Outsourcing [135:50] electricity generation Outsource things [135:52] that aren't your core competency right [135:54] focus on what makes the beer your beer [135:56] whatever that is proverbially taste [135:58] better and everything that is not that [136:00] like Finance and Accounting Outsource [136:02] Etc double underscore that but you know [136:05] like this is obvious so obvious but [136:08] obviously Bezos didn't say it directly [136:11] and thus I think we don't highlight it [136:13] enough the Counterpoint to that is [136:17] anytime you see something that lots of [136:19] people lots of companies are doing that [136:22] is not making their beer taste better [136:23] that is a massive opportunity to go [136:25] build a platform [136:26] company that is how you build a platform [136:29] company grading all right so we were [136:31] thinking for grading look we could grade [136:34] like I don't know taiwan's decision to [136:37] do this to own 50% of the company at the [136:40] whatever like you know uh A+ you know [136:42] not interesting so we had thought [136:45] experiment we'll try this for um this [136:48] episode rather than like letter grading [136:50] this we'll ask a question where does [136:54] tsmc belong in the pantheon of great [136:58] technology companies of all time is it [137:01] Fang level is it like top five is it top [137:04] 10 is it like top 20 like where is this [137:06] what is the right context in which we [137:10] should be placing tsmc this whole story [137:13] the company the power all of it so [137:15] interesting because it really does raise [137:17] this question of value chain I mean we [137:19] talked about the five part value chain [137:21] that exists today for making chips and [137:25] so it's interesting because you could [137:28] say Well it belongs wherever Intel [137:29] belonged Circa [137:31] 2,000 or you could say well the set of [137:36] products that tsmc manufacturers have [137:39] 100x the scale that Intel in 2000 had [137:43] like if you think about it all this [137:45] stuff that everyone's all excited about [137:46] every time someone talks about the next [137:47] wave of computing and they're like [137:48] machine learning or they're like crypto [137:51] or they're like 5G and anything they [137:53] tell you is something tsmc makes that [137:55] enables it all when Mark andreon says [137:57] software is eating the world it's only [137:59] eating the world because tsmc has made [138:00] it so freaking cheap to manufacture [138:03] silicon and then you can run whatever [138:05] you want on that silicon and it's the [138:08] cost of compute ASM totically approaches [138:11] zero because tsmc tsmc tsmc so how much [138:15] do we ascribe to them versus asml how [138:18] much do we ascribe to them versus the [138:21] entire landscape of talented chip [138:24] designers out there including the like [138:26] 600 chip designers at Apple working on [138:28] the Apple silicon is hard to [138:31] disambiguate that so where does it [138:35] belong I mean it's probably the most [138:38] successful and important [138:42] B2B hardware company of all time I think [138:45] we can safely [138:46] say at this point it surpasses Intel I [138:51] mean gosh right like that's a big [138:52] statement to say right like Intel [138:54] Silicon Valley the Traders say like all [138:55] of it More's law but in compounding all [138:59] the value shows up at the end so it is [139:01] true that the value that tsmc will [139:03] create in the world over the next year [139:05] two years 3 years is probably more than [139:08] the entire silicon industry leading up [139:10] to this point combined I mean hell they [139:12] grew 30% last year at like an already [139:15] unimaginable scale like Intel's not [139:17] doing that right okay I think we can say [139:20] it's above Intel I probably wouldn't say [139:23] it is above Facebook Amazon Apple [139:25] Microsoft Google in terms of pure value [139:30] Creation in the world I mean Devil's [139:33] Advocate You could argue that none of [139:35] the Innovative things those companies [139:36] are doing now happens without [139:39] tsmc yeah unless The Foundry model and [139:42] the fabulous model was inevitable yeah [139:45] maybe somebody else would have done it [139:47] yeah maybe but they didn't they didn't [139:51] um did I mean guys the thing that has [139:53] really just beaten me over the head in [139:55] this episode we've probably beaten all [139:58] of you over the head with or at least I [139:59] have is you know look there's the [140:02] geopolitical risk with being in [140:04] Taiwan other than that I don't know that [140:07] there is a stronger moat that any [140:10] company has in the entire world than [140:12] tsmc compare it to all the Fang [140:14] companies in Microsoft those are very [140:16] very strong modes but we've seen all of [140:18] those you know they've CH their new [140:20] companies emerged they've you know [140:22] Microsoft fell and then now it came back [140:24] with new strategy and Facebook's not [140:26] that old and Google's not that old tsmc [140:28] is impenetrable yeah their business [140:30] model and the costs required to compete [140:32] are such that they have it's like [140:35] bulletproof it's everything but [140:37] bulletproof yeah totally sadly sadly [140:40] yeah so you know I don't know maybe [140:42] we're exagger because we're like so deep [140:43] in it like we always go native on these [140:45] episodes the only way it could be more [140:47] valuable is if the company had an army [140:50] it's like people talk about the US [140:52] dollar is backed up by the full faith of [140:54] the US government which implies guns [140:57] yeah and so it's only because [140:59] everybody's currently playing by the [141:01] rules that any business gets to stay in [141:03] business and so this one just happens to [141:06] be a little bit more at risk than other [141:08] ones all right so I think we can safely [141:10] say top 10 I think the question is is it [141:12] top five well defensible this [141:14] interesting question so in 30 years will [141:18] tsmc be a huge company well they've got [141:21] this Dynamic going right now with this [141:23] flywheel where like structurally nobody [141:26] can catch them something unforseen has [141:29] to change but something unforeseen will [141:31] change because it always changes right [141:33] right right yes yes true who's had the [141:36] most similar dynamic in the past [141:39] Standard Oil either been successful or [141:41] unsuccessful Standard Oil is a good one [141:42] I mean that's very different style but [141:45] same sort of dynamic with Standard Oil [141:47] right was they crowded out structurally [141:50] how they were set up nobody else could [141:52] compete and the rich kept getting richer [141:55] and they kind of still exist that's the [141:56] best that is the best part they kind of [141:58] still exist yeah all right I'm with you [142:01] I'll go top 10 but probably not top five [142:04] what I'm wrestling with is how much of [142:05] it is just marketing and I don't mean [142:07] Marketing in a bad way but intentionally [142:09] tsmc rides under the radar they [142:12] intentionally have no brand right the [142:13] brand is the customers they want the [142:15] customers to succeed so we don't hear [142:17] all the time about them like we do the [142:19] Fang companies yeah we will start to I [142:22] think anybody who tunes into this [142:23] episode probably saw the name of the [142:25] episode and then thought hm I should [142:27] tune into that because I've seen more [142:28] about this thing recently prly didn't [142:30] know about kind of like we did when we [142:32] were like we should do this episode it's [142:34] finally time yeah yeah all right well [142:37] that's where I want to leave it all [142:38] right I'll put a s in the ground I'm [142:40] gonna say I think I'm with you top 10 [142:42] not top five yet but you know maybe we [142:44] need to revisit this I will definitely [142:46] say it's the most successful B2B [142:47] hardware company ever and the question [142:49] is is that the most successful Babb [142:52] company ever I'd say it's probably just [142:55] competing with Microsoft there yeah I [142:57] mean and again maybe even like across [142:58] all Industries right like look at I mean [143:00] shoot semiconductors run everything yeah [143:03] and they run [143:04] semiconductors semiconductors are the [143:06] new oil David uh okay okay enough enough [143:09] enough we got to bring this one home [143:11] carve outs carv outs let's do it I've [143:14] got two Jenny and I were just down in [143:16] Santa Barbara for a couple weeks rented [143:18] an Airbnb down there it was so great we [143:20] did that last last year hopefully this [143:21] becomes an annual thing in the Summer [143:23] Escape the freezing San Francisco [143:25] Summers um the uh so while we were down [143:29] we don't we don't watch a lot of TV [143:31] usually but you know it's like change a [143:33] scene summertime in a new place we're [143:35] like all right we'll watch some TV [143:36] together at night so this is like for [143:39] the percentage of you out there who are [143:40] living under a rock like me with TV [143:43] we've watched now most of Ted lasso [143:45] season one CU we heard season two was [143:47] terrible but that made me think well it [143:48] was terrible it is terrible but season [143:50] one's great made me think oh if people [143:51] are this upset about season 2 that means [143:53] season one was really good it's so good [143:55] if you haven't watched it we're on [143:56] episode 8 [143:58] now so good love it and then the other [144:01] TV show we watched this was Jenny's [144:03] suggestion old school throwback uh show [144:06] called Greek which aired in the mid [144:08] 2000s and is about oh yeah Greek like [144:11] sority and fraternity life a fol [144:13] university and like it's just so it's [144:15] like one of those heartwarming period [144:17] pieces but it was right from like when [144:19] we were in college so yeah it's fun nice [144:22] all right David watching TV who knows [144:24] what could change in the world right [144:26] maybe tsmc's mode isn't as deep as we [144:29] thought all right well mine is a book [144:31] that has been recommended to me for two [144:32] or three years now and I finally got [144:34] around to reading and it was awesome [144:36] it's called who is Michael Ovitz and if [144:40] you've read shoe dog and you've read the [144:43] ride of the lifetime and you've read [144:45] what's the Ford one an American icon [144:47] these iconic clastic yeah what's uh [144:49] there's [144:52] America this one needs to be on your [144:54] list especially if you've enjoyed any [144:57] movies or TV shows that were put [144:59] together in the last well let's be [145:01] specific or our two-part injuries in [145:03] horwood series totally from like 1975 to [145:07] 2000 Michael Ovitz put everything [145:10] together and it is this wonderfully [145:13] written book about an unbelievable [145:15] business Story the strategy behind it [145:18] the way that with creative artist agency [145:20] they just completely upended the entire [145:23] industry in Hollywood and did it really [145:25] without ever talking to the press and [145:27] were very tight lipped about it for some [145:29] Hollywood Outsider I found the book [145:32] really wonderful really compelling I [145:34] also think I previously had only read [145:38] the ride of a lifetime and watched the [145:40] Disney plus special about sort of the [145:42] history of Disney and Disneyland I had a [145:44] one-sided view of Michael based on just [145:47] his short tenure at Disney yeah I was [145:50] yeah what a great connection with you [145:51] know Acquired and Disney and totally [145:54] Iger and what kicked it off was doing [145:56] the in recent episodes and you know [145:58] hearing about how they base it on CIA so [146:00] especially if you like those episodes or [146:02] if you like the Disney episodes or if [146:03] you're a movie fan or if you like these [146:06] classic CEO business stories who is [146:08] Michael loitz was just an awesome read [146:10] so cool well and you know like all the [146:12] media that we grew up on probably even [146:14] more so cuz we were kids but it was the [146:15] adult movies and the kids movies too but [146:18] you know when you're a kid in like the [146:19] adult movie that you really want to see [146:21] young to see and just all these such [146:23] Classics like Good Fellas like that was [146:25] just my previous car out or Jurassic [146:27] Park or just everything that they [146:29] packaged it's cool to hear how it came [146:31] to be super cool I don't think we told [146:33] you at the beginning but you can join [146:35] our slack acquire. fm/ slack come hang [146:38] out with other talented smart [146:41] good-looking people like yourselves and [146:43] with that listeners feel free to share [146:47] the show with a friend shout it from the [146:48] social media hilltops yeah you always [146:51] sometimes you say but uh I'll chime in [146:53] here too like seriously you know it's [146:54] funny podcasting is this weird thing [146:57] right there's no viral loop it's not [146:59] like you know you can share please you [147:01] know share it from social media we love [147:02] that that's great if you love this [147:03] episode you think it's interesting you [147:04] think what we do is cool here but really [147:06] the way this gos is Word of Mouth that [147:08] is it people tell their friends they [147:10] listen to this episode they thought it [147:11] was cool they think that their friends [147:13] would really enjoy learn from listening [147:16] to this too so share a thing you liked [147:17] share a thing you disagree with us on [147:19] whatever it is is if you feel that way [147:21] please do that if you don't feel that [147:23] way get in touch with us and tell us why [147:26] all right listeners we will see you next [147:28] time we'll see you next time who got the [147:32] truth is it you is it you is it you who [147:36] got the truth now huh [147:39] [Music]