Elizabeth Holmes and Theranos, AI Data Center Forecast Jumps, Poppi’s Pepsi Deal | Diet TBPN

TBPN

AI-assisted summary of Elizabeth Holmes and Theranos, AI Data Center Forecast Jumps, Poppi’s Pepsi Deal | Diet TBPN — original source. Check the original for full context.

Source published: 2026-10-20

Key ideas

Critics attempting to rehabilitate Elizabeth Holmes ignore that medtech is an established industry where over 60 diagnostic companies have sold for $100M+ without fraud, while Billy Evans's movie points deal highlights potential workarounds to bypass her court-ordered restitution. (0:00)
SemiAnalysis tracking shows data center commitments targeting an unprecedented 50 gigawatts of power, with heavy physical construction and long-term utility deals locked in regardless of short-term public market volatility. (12:31)
A leaked internal Anthropic survey shows employees predict an average of 17.1% annual real GDP growth by 2030, presenting a radically positive-sum economic forecast compared to the FOMC baseline of roughly 2%. (22:17)
PepsiCo slashed Poppi's contingent $300 million earn-out payout down to $12 million, yet the beverage startup's founders still walked away with $1.9 billion in guaranteed upfront cash at transaction close. (28:00)

Detailed notes

## Background This episode of Diet TBPN was published on YouTube by TBPN on October 20, 2026. Hosted by John Coogan and Jordi Hays, the show recaps the best segments of their live tech broadcast, covering Elizabeth Holmes discourse, massive AI data center forecasts, Brett Adcock's Hark launch, and Poppi's acquisition earn-out dispute.

## TL;DR John Coogan and Jordi Hays dissect the public's bizarre attempt to rehabilitate Elizabeth Holmes following recent documentary coverage, emphasizing her fundamental failure as an operator alongside the mechanics of her partner Billy Evans's movie deals. The hosts then examine the explosive 50-gigawatt AI data center buildout, debating how enterprise budgets rather than consumer subscriptions must fund massive inference costs while reviewing Brett Adcock's buggy, privacy-invasive Hark pizza promotion. They scrutinize Anthropic's wildly optimistic internal forecast of 17% annual GDP growth and spotlight how AI-generated products already saturate Walmart aisles. Finally, they debunk the narrative surrounding Poppi's Pepsi acquisition, explaining why walking away with over $1.9 billion in upfront cash makes missing a $300 million earn-out an unambiguous success.

## Atmosphere & style The discussion carries the fast-paced, irreverent, and banter-heavy rhythm of an insider Silicon Valley morning show. Coogan and Hays bounce between acerbic skepticism—mocking Holmes's lack of basic scientific literacy and Hark's intrusive permissions—and sharp commercial pragmatism when analyzing enterprise compute and M&A earn-out traps. Listening feels like sitting in on an unfiltered tech lounge conversation where absurd founder behavior is gleefully roasted while the cold financial realities of venture capital and infrastructure are laid bare.

## Bulletin - [00:00](https://www.youtube.com/watch?v=WN_3je_KDHA&t=0s) Critics attempting to rehabilitate Elizabeth Holmes ignore that medtech is an established industry where over 60 diagnostic companies have sold for $100M+ without fraud, while Billy Evans's movie points deal highlights potential workarounds to bypass her court-ordered restitution. - [12:31](https://www.youtube.com/watch?v=WN_3je_KDHA&t=751s) SemiAnalysis tracking shows data center commitments targeting an unprecedented 50 gigawatts of power, with heavy physical construction and long-term utility deals locked in regardless of short-term public market volatility. - [22:17](https://www.youtube.com/watch?v=WN_3je_KDHA&t=1337s) A leaked internal Anthropic survey shows employees predict an average of 17.1% annual real GDP growth by 2030, presenting a radically positive-sum economic forecast compared to the FOMC baseline of roughly 2%. - [28:00](https://www.youtube.com/watch?v=WN_3je_KDHA&t=1680s) PepsiCo slashed Poppi's contingent $300 million earn-out payout down to $12 million, yet the beverage startup's founders still walked away with $1.9 billion in guaranteed upfront cash at transaction close.

## Key takeaways - [00:00](https://www.youtube.com/watch?v=WN_3je_KDHA&t=0s) Elizabeth Holmes was fundamentally an inept operator rather than a visionary victim, attempting to treat FDA-regulated diagnostic hardware like an unregulated consumer SaaS product. - [00:00](https://www.youtube.com/watch?v=WN_3je_KDHA&t=0s) More than 60 diagnostic device companies have completed acquisitions above $100 million since Theranos launched, demonstrating that legitimate point-of-care diagnostic innovation thrives without fraudulent claims. - [04:15](https://www.youtube.com/watch?v=WN_3je_KDHA&t=255s) Billy Evans's backend equity in Nathan Fielder's documentary could allow their household to collect millions in production upside while shielding those earnings from Holmes's court-ordered restitution obligations. - [10:00](https://www.youtube.com/watch?v=WN_3je_KDHA&t=600s) Sunny Balwani received a longer sentence (12 years, 11 months) than Holmes (10 years), yet neither founder extracted secondary liquidity before the $11 billion valuation collapsed. - [10:00](https://www.youtube.com/watch?v=WN_3je_KDHA&t=600s) Founders commonly extract personal wealth without raising negative signaling flags by taking out nine-figure equity-backed credit lines and later selling secondary stock under the guise of paying off personal debt. - [12:31](https://www.youtube.com/watch?v=WN_3je_KDHA&t=751s) Satellite tracking and land contracts indicate a 50GW wave of AI data center capacity is coming online over the next 12 months, detached from short-term software ARR hesitations. - [14:31](https://www.youtube.com/watch?v=WN_3je_KDHA&t=871s) Consumer AI assistants cannot generate enough margin to justify infrastructure capex; long-term monetization relies on enterprises granting six-figure inference budgets to employees to multiply productivity. - [20:00](https://www.youtube.com/watch?v=WN_3je_KDHA&t=1200s) Brett Adcock's AI assistant Hark triggered severe privacy backlash during a $1 million promotional pizza giveaway due to sweeping OAuth requests to read, edit, compose, and delete user emails, spreadsheets, and contacts. - [20:00](https://www.youtube.com/watch?v=WN_3je_KDHA&t=1200s) Spinning Hark out from Figure mirrors Elon Musk's xAI strategy of establishing a clean capitalization table to attract premier AI researchers with founder-level equity. - [22:17](https://www.youtube.com/watch?v=WN_3je_KDHA&t=1337s) Anthropic internal staff forecasts project an average 17.1% annual GDP growth by 2030, with upper estimates reaching 28%, reflecting immense confidence in AI's macroeconomic leverage. - [25:00](https://www.youtube.com/watch?v=WN_3je_KDHA&t=1500s) AI-generated physical consumer products are already flooding retail shelves at Walmart and on viral social platforms, driven by automated drop-shippers rather than elite enterprise software rollouts. - [25:00](https://www.youtube.com/watch?v=WN_3je_KDHA&t=1500s) Tech and education leaders like Tim Sweeney and Grant Sanderson view generative AI as an enabling tooling evolution comparable to past transitions like moving from C to C++ or 2D to 3D animation. - [28:00](https://www.youtube.com/watch?v=WN_3je_KDHA&t=1680s) Poppi's founders secured a massive liquidity win with $1.9 billion in guaranteed cash at closing, making Pepsi's decision to pay only $12 million of the remaining $300 million earn-out largely inconsequential. - [28:00](https://www.youtube.com/watch?v=WN_3je_KDHA&t=1680s) Startup founders entering corporate acquisitions should value performance earn-outs at zero, as post-acquisition parent companies inevitably shift resources, retail placement, and strategic priorities.

## Detailed walkthrough ### [00:00](https://www.youtube.com/watch?v=WN_3je_KDHA&t=0s) Elizabeth Holmes, the TED Talk Era, and the Reality of Medtech

Reflecting on the wave of online discourse following the release of the film *You Can See Everything*, John Coogan and Jordi Hays discuss how viewers seemed surprisingly eager to rehabilitate Elizabeth Holmes. Many people came away rationalizing her behavior, wondering if she had been repeatedly manipulated or excusing her fraud as standard Silicon Valley “fake it till you make it” culture. Rejecting that defense, the hosts argue that even if you strip away the criminal fraud entirely, Holmes was fundamentally an inept operator. As John puts it: “It seems like she's just a bad founder. Like put aside the fraud. Like it just seems like, like, total slop in terms of the execution of that business and gives me absolutely no faith in a Theranos 2.0.”

What proved most infuriating about her pitch was its utter disconnection from the actual medical device industry. Holmes presented the Theranos Edison machine as a magical, cloud-connected black box that could test blood, treating standard diagnostic concepts as unprecedented science. In reality, building diagnostic hardware is an established, highly active commercial space. Over sixty companies in that exact diagnostic category have sold for upwards of $100 million since Theranos was founded, all without their executives going to federal prison for a decade. The hosts point out that pitching a blood analyzer as an unheard-of breakthrough is the equivalent of claiming to have invented a revolutionary tool because it stores data in a database accessible via a browser—it is simply a SaaS app, something built by hundreds of conventional companies every year.

The root of Holmes's persona traced back to the post-*Social Network* tech boom between 2011 and 2015, which birthed the era of the “TED Talk entrepreneur.” During this period, an ecosystem of university accelerators, hard-tech showcases, and prodigy-focused venture networks tolerated young founders parading fraudulent products, fudged studies, and falsified metrics under the guise of vision. While other founders eventually adapted when that era ended, Holmes kept running the same playbook in medicine—the most heavily regulated industry in the world. As the hosts emphasize, medical devices hinge entirely on FDA marketing authorization: building hardware is not inherently illegal, but selling, marketing, or publicly promoting unapproved claims on podcasts and news broadcasts is strictly prohibited by law.

- Over 60 diagnostic device companies have completed acquisitions of $100 million or more since Theranos launched. - The 2011–2015 startup boom created an apparatus around young prodigies that normalized presenting fudged data and fake hardware prototypes on conference stages. - FDA clearance functions as marketing authorization; founders are legally barred from making unverified claims about clinical efficacy to the public or media prior to regulatory approval.

[04:15](https://www.youtube.com/watch?v=WN_3je_KDHA&t=255s) Billy Evans’ Movie Deal, Restitution Schemes, and Theranos 2.0

Turning to the behind-the-scenes drama of Nathan Fielder's project, the conversation shifts to Holmes’s partner, Billy Evans. During production, Evans negotiated a lucrative side deal to receive points on the back end of the film’s profits. After the project wrapped, he attempted to withdraw from the arrangement because of how damaging it looked publicly, though by then the documentary was already guaranteed to become a massive hit. The hosts unpack the financial incentives behind this move: Holmes was ordered to pay hundreds of millions of dollars in court-ordered restitution, meaning any personal earnings would be seized by the court for victims and investors. Because there is persistent ambiguity over whether Holmes and Evans are legally married, routing the backend points directly into Evans’s personal accounts could allow the household to capture millions in revenue while bypassing her restitution obligations.

The hosts also break down the absurdity of Evans attempting to pitch "Theranos 2.0." John shares that he spoke with Evans on the phone the previous year about his social media operations and was met with the pitch for a revival company. John’s stance was that if anyone was going to buy an Elizabeth Holmes redemption narrative, she needed to prove she was actually a serious student of the science: “My point that I posted about was was like I want to see Elizabeth Holmes give an incredible take about peptides and how that market will evolve and then watch that happen and be like, ‘Oh wow, she was early.’ I want her to be a good biotech analyst, I want her to be smart on the science.” Yet throughout the documentary, she never displays deep scientific insight, leaving Evans to wander around trying to grasp basic concepts like mitochondria. The hosts joke that instead of vague sci-fi slogans, she should publish a biotech market map, a 2027 AI/biotech thesis, or a tier list of diagnostic acquisitions to prove she understands the competitive landscape.

To illustrate how real medical hardware innovation actually works, the hosts highlight the playbook that Holmes failed to run: taking a standard diagnostic workflow and making it faster and smaller for point-of-care hospital use. Instead of trying to invent an impossible universal blood test, practical biotech startups focus on a single critical diagnostic need—such as testing for antibiotic-resistant bacteria like MRSA on-site rather than sending samples to a centralized lab. By shrinking down a dishwasher-sized machine into a desktop unit that returns results in one hour instead of two days, companies create enormous clinical and financial value.

```mermaid flowchart LR accTitle: Practical Medtech Playbook vs Theranos accDescr: Flowchart comparing a targeted diagnostic product pipeline to Theranos's failed approach.

subgraph Legitimate["Real Diagnostic Startup Playbook"] A["Identify Hospital Pain Point (e.g., MRSA / Antibiotic Resistance)"] --> B["Shrink Lab Hardware (Dishwasher size to desktop box)"] B --> C["Secure FDA Authorization (e.g., cobas vivoDx MRSA in 2019)"] C --> D["Acquisition by Roche ($425M outcome on $25M raised)"] end

subgraph Theranos["Theranos Approach"] E["Universal Blood Test Pitch (Hundreds of tests from a micro-drop)"] --> F["TED Talk Marketing & Cloud Slogans"] F --> G["Falsified Test Data & Regulatory Evaporation"] G --> H["Criminal Conviction & Prison"] end ```

- In a legitimate medtech success story referenced by the hosts, a company focused specifically on detecting antibiotic-resistant bacteria raised $25 million to build a point-of-care box. - The company was acquired for $425 million, and Roche subsequently secured FDA authorization in 2019 for the device as the cobas vivoDx MRSA test. - Shifting bacterial diagnosis turnaround times from two days to one hour directly improves hospital patient outcomes and creates hundreds of millions of dollars in enterprise value without relying on unproven claims.

[10:00](https://www.youtube.com/watch?v=WN_3je_KDHA&t=600s) Theranos Sentences, Secondary Sales, and the Founder Debt Hack

The hosts discuss the fallout from Theranos, focusing on the stark reality of the prison sentences handed down to Elizabeth Holmes and Sunny Balwani. One host recalls walking away from a recent film under the distinct impression that Balwani had cut a side deal and managed to escape serious punishment, pinning the blame on Holmes while she took the fall with roughly 11 years. However, Balwani ended up receiving an even harsher sentence: 12 years and 11 months, compared to Holmes’s sentence of 10 years (which was reduced from an initial 11 years and three months).

What baffles the hosts even more is that neither Balwani nor Holmes appeared to have cashed out by selling secondary shares along the way. Both went down with the sinking ship without banking any personal windfalls from an $11 billion valuation. One host marvels at the sheer absurdity of running a multi-billion-dollar scheme without a financial cushion: “Least shrewd person in the world potentially no secondary stuff 11 billion fraud purely for the love of the game what are you doing what are you doing and it's not like I came away being like it wasn't a fraud like it was.” The other host agrees that it embodies the tragic parody of the true believer tech entrepreneur who goes behind bars purely “for the vision.”

The conversation then shifts to a clever, tongue-in-cheek financial loophole founders use when they want to pocket substantial wealth without sending a negative signal to venture markets: - A founder scales their startup until it reaches a unicorn valuation in the billions. - Selling $10 million or $100 million of secondary shares outright would look like cashing out or losing faith in the company. - Instead, the founder takes out an immense line of credit—say, a $100 million personal loan backed by their equity. - Later, they quietly sell $100 million of secondary stock and announce to the public that they are merely liquidating shares to pay off personal debt. - The public mistakenly assumes they are dealing with standard credit-card-level personal debt, completely unaware it was a pre-planned, equity-backed credit line designed to extract liquidity cleanly from the outset.

---

[12:31](https://www.youtube.com/watch?v=WN_3je_KDHA&t=751s) The 50GW Compute Boom and SemiAnalysis Supply Forecasts

The hosts pivot to the surging scale of artificial intelligence infrastructure, addressing reports that data center capacity is targeting an unprecedented 50 gigawatts of power. While prominent tech investors like Brad Gerstner have debated whether the market is building out too much compute too fast, research firm SemiAnalysis has been cross-referencing power contracts and tracking physical development using satellite imagery.

```mermaid flowchart LR accTitle: Data Center Buildout Momentum accDescr: Diagram illustrating how inked contracts and physical construction resist short-term market fluctuations. subgraph Commitments ["Long-Term Pipeline"] A["50GW Inked Deals & Land"] --> B["Satellite-Tracked Construction"] end subgraph Volatility ["Market Sentiment"] C["Short-Term Nasdaq Dips (5%)"] -.->|"Cannot stop"| B end subgraph Delivery ["Capacity Output"] B --> D["Massive Supply Online (Next 12 Mos)"] end ```

The hosts review a note from SemiAnalysis circulated via RevCap, noting the firm's paradoxical reassurance to investors worried about stalling software ARR: you do not need to worry about demand because an even more staggering wave of supply is about to hit the market over the next 12 months. SemiAnalysis remains extraordinarily bullish on the leading AI labs, and the physical reality on the ground confirms that buildouts cannot simply be paused on a whim. While a 5% drop on the Nasdaq might shake public market confidence, massive industrial projects with earthmovers and concrete already poured will keep rolling forward regardless of short-term interest rate jitters.

---

[14:31](https://www.youtube.com/watch?v=WN_3je_KDHA&t=871s) AI Agent Monetization and Enterprise vs. Consumer Unit Economics

While infrastructure supply is undeniably materializing, the hosts debate where the recurring demand and actual willingness to pay will come from. One host is an avid user of Dot, an autonomous AI assistant, highlighting how transformative agentic tools are for day-to-day productivity. The agent quietly monitors inbound email, follows up on outstanding requests without being asked, and was even deployed successfully to cancel a Tesla Roadster reservation.

Interacting with the open web via a large language model feels vastly superior to modern web browsing because it strips out the clutter: - No aggressive cookie consent pop-ups. - No intrusive programmatic ad displays. - Synthesized answers containing exact source quotes, letting users bypass having 25 browser tabs open at once.

Despite this superior user experience, monetization remains the critical hurdle. If the entire agentic commerce market caps out around $40 billion on interchange and take rates, consumer subscriptions cannot shoulder the massive capital expenditures required for 50GW of data centers. One host jokes about a viral video featuring a woman claiming to run a "26-hour day" to stay online with her agents, but they note that everyday consumers will not spend $1,000 annually on personal inference:

> “She's not crazy. You're crazy. You're crazy for thinking that it takes 24 hours to complete a full day.”

Instead, the return on investment must originate from enterprise diffusion. The hosts point to Jensen Huang’s thesis: if a software engineer earns $300,000, a company should readily give them an equivalent $300,000 budget in AI tokens if it multiplies their output. The same calculus will eventually apply to enterprise sales reps receiving $50,000 a month in inference budgets to generate $100,000 in incremental pipeline. For the broader market to sustain the infrastructure buildout, the typical $10 million top-line business currently spending $100,000 on software seat licenses will have to scale its inference spending tenfold to $1 million per year.

---

[19:06](https://www.youtube.com/watch?v=WN_3je_KDHA&t=1146s) Hark's $1 Million Pizza Promotion PR Backlash

The discussion turns to consumer marketing stunts in the AI agent ecosystem, specifically Brett Adcock’s rollout of Hark, a personal AI tool competing alongside agents like Dot and Instinct. To juice downloads, Hark ran a promotional campaign offering to give away $1 million worth of free pizza through their newly launched application.

Users were instructed to download the app and place their food orders directly through the Hark interface. However, the promotion swiftly deteriorated into an operational and public relations headache. While the marketing push succeeded in driving a massive spike in app installations, users ran into technical failures and voiced deep privacy reservations about the sweeping data access permissions requested by the application during setup. Joking about the intrusive access requirements needed just to claim a slice, one host quips: “In order to get this pizza, I will need to edit your tax return.”

[20:00](https://www.youtube.com/watch?v=WN_3je_KDHA&t=1200s) Hark’s Free Pizza Debacle and Permissions Backlash

The conversation turns to the aggressive permissions requested by Hark, a new venture from Brett Adcock. The hosts run through the staggering list of access rights demanded during onboarding, which include permissions to view, edit, create, and permanently delete contacts, Google Sheets spreadsheets, calendars, and even read, compose, and send emails directly from a user’s Gmail account, alongside scraping an organization's internal workplace directory. While some of these might technically be optional bypasses in standard OAuth permission flows—comparable to opting out of push notifications on iOS—the optical severity of the permissions list triggered immediate alarm bells across the tech community.

The launch day was further complicated by Hark's viral promotion: an advertised $1 million free pizza giveaway launched on X on October 16, 2026, where users could supposedly claim a pie simply by messaging the app. When the service predictably bottlenecked and failed, one user named Armand instructed the AI agent to draft and file a formal grievance with the Federal Trade Commission. The AI initially attempted to talk him down, suggesting that launch-night flakiness was better handled by bug reports to the Hark engineering team, but Armand insisted on federal escalation. Reflecting on Brett Adcock's broader corporate structure, the hosts question why Hark was spun out as a separate entity rather than built inside Figure. They note that Adcock appears to be following the playbook Elon Musk established with xAI: establishing a fresh capitalization table to attract premier AI researchers with founder-level equity grants, an incentive structure that legacy hardware balance sheets like SpaceX or Figure cannot easily accommodate.

> “On October 16, 2026, Hark publicly advertised a $1 million free pizza giveaway.”

> “The only thing is Elon had sold a lot of cars and launched a lot of rockets prior to doing that.”

```mermaid flowchart TD accTitle: Hark Enterprise Spin-Out Rationale accDescr: Diagram illustrating the structural reasoning behind spinning out Hark from Figure. subgraph Core ["Figure Robotics"] A["Hardware & Robotics Focus"] B["Constrained Equity Structure"] end

subgraph SpinOut ["Hark Standalone Venture"] C["Fresh Cap Table"] D["$1M Consumer Pizza Promo"] E["Founder-Level Equity for AI Talent"] end

Core -.->|"Musk Playbook"| SpinOut E -->|"Attracts"| F["Top AI Researchers"] ```

[22:17](https://www.youtube.com/watch?v=WN_3je_KDHA&t=1337s) Anthropic’s 17% GDP Forecast and AI Super-Growth Debate

The hosts shift to an internal poll leaked from Anthropic regarding future economic growth spurred by artificial intelligence. The baseline premise rests on whether the forecasts account for real versus nominal gross domestic product, noting that nominal growth paired with runaway inflation would result in negative real expansion. Assuming real GDP is the benchmark, the consensus inside Anthropic skews wildly optimistic: the average staff prediction sits at an astonishing 17.1% annual GDP growth by 2030, with one bullish outlier projecting 28% and the absolute lowest internal forecast coming in at 6%.

This stands in stark contrast to traditional macroeconomic projections. The Federal Open Market Committee holds steady at the historical 2% baseline, broad surveys of American adults land at 5.4%, and external website visitors settle around 12.5%. The hosts joke that an annual 17% economic explosion would turn the economy into an overwhelmingly positive-sum environment where roads, infrastructure, and universal healthcare could easily be funded out of sheer surplus. The extreme optimism has drawn intense criticism online, notably from commentator Spore, who dismissed the 20% projections as detached from sanity. The hosts playfully riff on the Palantir executive Shyam Sankar’s infamous anecdote on CNBC/Bloomberg about AI researchers chain-smoking cigarettes because they genuinely believe artificial intelligence will eradicate lung cancer before tumors can form.

> “Anthropic staff are obviously like the most bullish.”

> “If we grow at 17%, it is so positive sum.”

> “Spore says this is borderline mental illness.”

| Group / Forecaster | Predicted Annual GDP Growth | |:--- |:--- | | Anthropic Outlier (Highest) | 28.0% | | Anthropic Internal Staff Average | 17.1% | | Site Visitors Poll | 12.5% | | Anthropic Lowest Forecast | 6.0% | | General U.S. Adults Survey | 5.4% | | Federal Open Market Committee (FOMC) Baseline | ~2.0% |

[25:00](https://www.youtube.com/watch?v=WN_3je_KDHA&t=1500s) AI Slop in Walmart and the Optimism of Sweeney and 3Blue1Brown

The hosts examine how artificial intelligence is permeating the mainstream consumer economy, bringing up an interview between Tyler Cowen and Dwarkesh Patel on Cowen's podcast. When Cowen pressed Patel on when everyday shoppers would walk into a Walmart and see a physical product resulting from artificial intelligence, Patel estimated around 2028. The hosts counter that Cowen and Patel are missing the obvious: AI-generated physical consumer goods are already dominating retail shelves today. From bizarre algorithmically rendered beach towels in Walmart aisles to viral TikTok inflatable tank pool floaties generated by AI video slop, low-end automated commerce is already flourishing.

Rather than looking at high-minded corporate research labs, the hosts argue that tech observers should look to the drop-shippers and course-sellers speeding around Miami in rented Lamborghinis to see where AI is making real-time financial impact. Parallel to this retail wave, industry leaders outside the frontier labs are expressing genuine long-term optimism. Tim Sweeney, founder of Epic Games, recently argued that generative technology will ignite a golden age of game development by making lavish, resource-heavy game designs economically viable for the first time. Similarly, Grant Sanderson of the educational channel 3Blue1Brown released content cheering AI’s role in mathematics, viewing this shift not as an existential threat, but as another normal technological migration akin to moving codebases from C to C++ or transitioning animation pipelines from 2D to 3D.

> “You can go into Walmart and get AI products now.”

> “The AI slop economy is thriving right now.”

> “It’ll be a new golden age of gaming.”

[28:00](https://www.youtube.com/watch?v=WN_3je_KDHA&t=1680s) Poppi’s $2.2B Pepsi Deal and the Earn-Out Trap

The conversation moves to the major corporate acquisition of prebiotic soda maker Poppi by PepsiCo. Initial headlines framed the transaction as a $2.2 billion buyout, but financial disclosures revealed that $300 million of that total valuation was tied up in performance-contingent earn-outs. When Pepsi recently announced it would only pay out a mere $12 million of the remaining $300 million pool, social media critics rushed to frame the haircut as a catastrophic loss for the beverage founders.

The hosts reject this framing, pointing out that Poppi’s founders collected $1.9 billion in guaranteed cash at closing, giving them more than 90% of the top-line transaction value. With the additional $12 million payout, the founders walk away with $1.912 billion, marking an unambiguous triumph in founder liquidity. A genuine disaster would have looked like the reverse: collecting a tiny upfront cash payment paired with an earn-out that gets slashed to near zero. Echoing commentary from Ridge Wallet CEO Sean Frank, the hosts emphasize that founders should treat any secondary earn-out payment as worth zero on paper. The moment an acquired brand enters a conglomerate's balance sheet, original management forfeits ultimate control over operational budgets, retail shelving leverage, and marketing priority to corporate handlers who have entirely different strategic agendas.

> “$300 million of that $2.2 billion was an earn out.”

> “Now, Pepsi announced that instead of paying 300 million based on the performance of the brand, they will only be paying 12 million.”

> “You shouldn’t go into one of these deals expecting to get the full earn out because companies have lots of resources.”

> “You’re no longer the master of your own destiny.”

## Notable quotes > [25:00](https://www.youtube.com/watch?v=WN_3je_KDHA&t=1500s) The AI slop economy is thriving right now.

## Mentioned 3Blue1Brown, Anthropic, Billy Evans, Brad Gerstner, Brett Adcock, Cisco, Console, CrowdStrike, Diet TBPN, Dot, Dwarkesh Patel, Elizabeth Holmes, Epic Games, Federal Open Market Committee (FOMC), Federal Trade Commission (FTC), Figma, Figure, Food and Drug Administration (FDA), Grant Sanderson, Hark, Instinct, Jensen Huang, John Coogan, Jordi Hays, Mark Cuban, Mark Zuckerberg, MongoDB, NYSE, Nathan Fielder, Palantir, PepsiCo, Poppi, Public, Railway, Ramp, RevCap, Ridge Wallet, Roche, Sam Altman, Satya Nadella, Sean Frank, SemiAnalysis, Shopify, Shyam Sankar, Sunny Balwani, TBPN, Tesla Roadster, The New York Times, Theranos, Tim Sweeney, Tyler Cowen, Walmart, You Can See Everything, xAI

## Video description Sign up for TBPN’s daily newsletter at TBPN.com

Diet TBPN delivers the best of today’s TBPN episode in 30 minutes. TBPN is a live tech talk show hosted by John Coogan and Jordi Hays, streaming weekdays 11–2 PT on X and YouTube, with each episode posted to podcast platforms right after.

Described by The New York Times as “Silicon Valley’s newest obsession,” the show has recently featured Mark Zuckerberg, Sam Altman, Mark Cuban, and Satya Nadella.

TBPN is made possible by: Ramp - https://ramp.com Public - https://public.com Cisco - https://www.cisco.com Console - https://www.console.com CrowdStrike - https://www.crowdstrike.com Figma - https://www.figma.com MongoDB - https://www.mongodb.com NYSE - https://www.nyse.com Railway - https://railway.com Shopify - https://www.shopify.com/

Follow TBPN: https://TBPN.com https://x.com/tbpn https://open.spotify.com/show/2L6WMqY3GUPCGBD0dX6p00?si=674252d53acf4231 https://podcasts.apple.com/us/podcast/technology-brothers/id1772360235 https://www.youtube.com/@TBPNLive

Chapters: 00:00 Elizabeth Holmes, You Can See Everything + The Theranos Backlash 04:15 Billy Evans’ Movie Deal + Can Theranos 2.0 Redeem Holmes? 12:31 The 50GW AI Data Center Boom + The Supply Explosion 14:31 Who Pays for AI Agents? Dot, Compute + Enterprise Demand 19:06 Hark’s $1M Pizza Giveaway + The Privacy Backlash 22:17 Anthropic Predicts 17% GDP Growth + AI’s Real-World Impact 28:00 Poppi’s Pepsi Deal + The $300M Earn-Out Collapse

[Source](https://www.youtube.com/watch?v=WN_3je_KDHA)

中文简报

伊丽莎白·霍姆斯与Theranos、AI数据中心预测激增、Poppi的百事收购交易 | Diet TBPN John Coogan和Jordi Hays深入剖析了在近期纪录片播出后公众试图为伊丽莎白·霍姆斯（Elizabeth Holmes）洗白的怪异现象，强调了她作为运营者的根本性失败，并分析了其伴侣比利·埃文斯（Billy Evans）电影交易背后的运作机制。随后，两位主持人审视了规模达50吉瓦的爆发式AI数据中心建设潮，讨论了庞大的推理成本必须依赖企业预算而非消费者订阅来支撑，同时评价了布雷特·阿德科克（Brett Adcock）漏洞频出且侵犯隐私的Hark披萨推广活动。他们还仔细审视了Anthropic内部预测年均GDP增长达17%的极其乐观的预期，并指出AI生成的产品已充斥沃尔玛货架。最后，他们破除了围绕百事收购Poppi交易的传闻，解释了为何即使错失3亿美元的盈利能力对赌补偿，但凭借超过19亿美元的预付现金套现，仍是一次毋庸置疑的成功。试图为伊丽莎白·霍姆斯洗白的批评人士忽视了医疗科技是一个成熟的行业，其中有超过60家诊断公司在没有欺诈的情况下以逾1亿美元出售；同时，比利·埃文斯的电影分成交易凸显了规避法院判决其赔偿金的潜在漏洞。 (0:00) SemiAnalysis的追踪显示，数据中心承诺的目标功率达到了前所未有的50吉瓦，无论短期公开市场的波动如何，大规模实体建设和长期公用事业协议均已被锁定。 (12:31) Anthropic一份泄露的内部调查显示，员工预测到2030年实际GDP年均增长率将达到17.1%，与联邦公开市场委员会（FOMC）约2%的基准预测相比，呈现出一种极其激进的正和经济预测。 (22:17) 百事公司将Poppi价值3亿美元的附带条件盈利对赌款削减至1200万美元，但这家饮料初创公司的创始人在交易完成时仍拿到了19亿美元的有保障预付现金。 (28:00)

Reading summary

TBPN's John Coogan and Jordi Hays critique the persistent rationalization of Elizabeth Holmes's fraudulent execution, contrasting it with real-world biotechnology successes, while examining the infrastructure pressures of artificial intelligence and the financial realities of high-profile corporate acquisitions. Theranos, Regulatory Realities, and Restitution Maneuvers (0:00) Following the release of *You Can See Everything*, the hosts push back against public attempts to rehabilitate Elizabeth Holmes under standard Silicon Valley tropes. Rather than demonstrating visionary execution, Holmes acted as an inept operator out of touch with the commercial diagnostic landscape. More than sixty diagnostic companies have secured acquisitions exceeding $100 million since Theranos began, doing so legally within FDA regulatory bounds that prohibit marketing unapproved clinical claims. Real medtech progress focuses on shrinking targeted diagnostic workflows—such as a $25 million point-of-care bacterial diagnostic startup acquired for $425 million that secured FDA clearance under Roche—rather than promoting universal blood-testing promises. Founder Liquidity Schemes and Sentence Disparities (4:15) The hosts highlight the legal and financial fallout of Theranos, observing that partner Sunny Balwani received a harsher prison term (12 years and 11 months) than Holmes (10 years, reduced from 11 years and three months). Curiously, neither founder extracted secondary liquidity from their $11 billion paper valuation. Meanwhile, Holmes's partner Billy Evans negotiated back-end points on Nathan Fielder’s film project, creating a channel where household revenue might bypass Holmes's court-ordered restitution debt due to legal ambiguities surrounding their marriage. The 50GW Compute Buildout and Unit Economics (12:31) In AI infrastructure, research firm SemiAnalysis reports that data center development targeting 50 gigawatts of capacity will proceed despite software market volatility, as locked power contracts and satellite-tracked physical construction cannot simply be halted. However, the hosts question whether consumer agent applications can generate sufficient revenue to cover these massive capital outlays. Consumer willingness to pay remains modest, meaning infrastructure investments must rely on enterprise diffusion, such as companies significantly expanding software seat and token budgets to augment developer and sales productivity. AI Growth Projections, Launch Headwinds, and Retail Impact (20:00) Market expectations around AI vary wildly: while the Federal Open Market Committee projects historical baseline GDP growth around 2%, a leaked internal Anthropic poll revealed an average staff expectation of 17.1% annual GDP growth by 2030, drawing external skepticism. At the application layer, Brett Adcock's agent Hark faced severe backlash over broad data access permissions and an operational breakdown during a $1 million pizza promotion. Concurrently, AI's physical integration is already visible in retail, spanning algorithmic consumer products in Walmart aisles and low-end digital commerce, alongside long-term technical optimism from figures like Epic Games founder Tim Sweeney. Poppi’s Acquisition and Earn-Out Dynamics (28:00) Addressing PepsiCo's acquisition of prebiotic soda brand Poppi, the hosts challenge public narratives framing the deal as a failure after Pepsi paid only $12 million out of a contingent $300 million earn-out pool. Because the founders secured $1.9 billion in guaranteed upfront cash—capturing more than 90% of the headline $2.2 billion valuation—the transaction represents an overwhelming financial win. Once an acquired asset is folded into a conglomerate, original founders lose control over shelf prioritization and budgets, meaning earn-outs should always be treated as zero on paper. Across venture-backed biotech, commercial AI deployments, and consumer acquisitions, the hosts illustrate that operational control, regulatory adherence, and guaranteed upfront cash consistently outweigh promotional narratives and speculative earn-outs.

精炼阅读摘要

TBPN 主持人 John Coogan 与 Jordi Hays 探讨了科技与商业领域的现实落差，涵盖 Elizabeth Holmes 医疗欺诈案对创始人生态的警示、AI 算力基础设施与实际商业化之间的脱节，以及大型并购案中对赌条款的财务本质。 Theranos 骗局的本质与医疗创新的真实路径 (0:00) 主持人反驳了纪录片《You Can See Everything》后试图为 Elizabeth Holmes 脱罪的舆论。他们认为，即使抛开刑事欺诈不谈，Holmes 本质上也是一名执行力极其低劣的创始人，将成熟的常规诊断硬件包装成前所未见的神奇黑盒子。自 Theranos 成立以来，同一诊断赛道已有超过 60 家公司以 1 亿美元以上的估值完成收购，且高管均未触犯联邦法律。医疗硬件的核心在于 FDA 的营销授权，法律严禁在获得审批前向公众或媒体宣传未经证实的临床疗效。真正的医疗设备创新路径通常是针对单一具体痛点进行微型化改良。例如，有实际案例显示，一家初创公司专注于在床旁快速检测耐药菌（MRSA），仅融资 2500 万美元便将原本如洗碗机大小的实验设备缩小为桌面级仪器，将出结果时间从两天压缩至一小时，最终以 4.25 亿美元被 Roche 收购，其产品随后在 2019 年作为 cobas vivoDx MRSA 测试获得 FDA 授权。相比之下，Holmes 的伴侣 Billy Evans 试图兜售“Theranos 2.0”复兴概念，却连线粒体等基础生物学常识都未掌握，缺乏对科学与竞争格局的深刻认知。刑期定谳、偿债漏洞与创始人套现手段 (10:00) 针对 Theranos 案件的最终裁决，Sunny Balwani 最终被判处 12 年 11 个月监禁，刑期超过 Holmes 被减刑后的 10 年（最初为 11 年 3 个月）。令人费解的是，在公司估值曾达 110 亿美元的高峰期，两人均未通过二级市场抛售股票套现，最终以“真信徒”的荒唐姿态锒铛入狱。针对 Holmes 面临的数亿美元法院指定赔偿金，主持人指出 Billy Evans 在参与纪录片制作时私下商谈了后端利润分成，其后因舆论压力试图退出；由于二人法律婚姻状态存疑，将分成直接汇入 Evans 账户可能成为规避个人赔偿金判决的财务操作。此外，主持人揭露了一种独角兽创始人常用的隐秘套现机制：创始人不在二级市场直接抛售股权以避免引发市场恐慌，而是以股权为抵押借入上亿美元的个人信用额度，随后再以清偿“个人债务”为由平价减持股票，使外界误以为其仅是在处理常规私人负债。 50GW 算力狂潮、供给过剩与企业端商业化分歧 (12:31) 在人工智能基础设施领域，研究机构 SemiAnalysis 监测显示，数据中心签约与规划电力容量正瞄准 50 吉瓦（GW）。即便短期二级市场如纳斯达克回调 5% 或软件 ARR 增长放缓，由于土地合约与混凝土浇筑等物理施工已无法随意中止，未来 12 个月内仍将有巨量算力供给涌入市场。然而，算力基础设施的巨额资本开支难以仅靠个人消费者订阅来支撑。主持人认为，即便像 Dot 这样能够静默处理邮件和取消订单的 AI Agent 提供了优于传统网页的用户体验，但若整个 Agent 商务市场仅在抽成与费率上达到 400 亿美元规模，商业化依然面临瓶颈。日常消费者不可能为个人推理年均支付 1000 美元，AI 的投资回报必须依靠企业级渗透。例如英伟达 CEO 黄仁勋的观点，企业会为年薪 30 万美元的软件工程师匹配等额的 AI 代币预算以放大产能；一家年收入 1000 万美元、传统软件席位支出 10 万美元的企业，未来需要将推理支出扩大十倍至每年 100 万美元，才能支撑当前的算力扩张。消费端推广翻车与宏观经济超高速增长预测 (20:00) 消费级 AI Agent 在获取用户时频遭挫折。Brett Adcock 推出的 Hark 应用在 2026 年 10 月 16 日于 X 平台开展了 100 万美元免费披萨推广活动，要求用户授予读取和删除 Gmail 邮件、日历、Google 表格及公司目录等极度侵入性的权限，引发严重隐私警惕并导致服务崩溃，甚至有用户指示该 Agent 向联邦贸易委员会（FTC）提交正式投诉。主持人分析，Adcock 效仿马斯克建立 xAI 的做法，将 Hark 从硬件机器人公司 Figure 中独立拆分，主要是为了通过全新股权架构吸引顶级 AI 研发人才。在宏观前景方面，Anthropic 内部泄露的民调显示，员工对 2030 年年化 GDP 增长率的平均预测达到极其乐观的 17.1%，最高预测甚至达 28%，最低预测也有 6%；这与美联储公开市场委员会（FOMC）约 2.0% 的基准增长率、美国成年人普遍预期的 5.4% 形成鲜明对立。与此同时，AI 生成商品已通过低端算法渲染的毛巾和充气玩具渗透至沃尔玛货架；Epic Games 创始人 Tim Sweeney 与 3Blue1Brown 创作者 Grant Sanderson 则对 AI 表达了理性乐观，认为其将推动游戏开发的黄金时代，或成为类似于 C 语言迁移到 C++ 的正常生产力工具更迭。 Poppi 饮料 22 亿美元并购中的业绩对赌实质 (28:00) 针对百事公司（PepsiCo）收购益生元苏打水品牌 Poppi 的交易，初期宣传的总金额为 22 亿美元，但其中 3 亿美元属于与业绩挂钩的对赌款项（Earn-out）。近期百事宣布仅兑现该 3 亿对赌款中的 1200 万美元，引发社交媒体上关于创始人遭遇损失的议论。主持人驳斥了这种负面解读，明确指出创始团队在交易完成时已稳稳拿到了 1.9 亿美元的确定性现金，最终现金收益累计达 19.12 亿美元，占总交易价值的 90% 以上，属于无可置疑的巨大财务成功。主持人援引 Ridge Wallet 首席执行官 Sean Frank 的观点指出，在企业并购谈判中，被收购创始人理应在账面上将任何非确定性的后期对赌金额视为零。一旦标的资产被并入大型跨国巨头的资产负债表，原创始人便彻底丧失了对运营预算、货架资源分配和营销优先级的控制权，跨国集团的战略重心和利益诉求通常会导致原定业绩目标难以达成。本期节目通过对比 Theranos 的虚妄炒作与扎实医疗工程的差异，剖析了当下 AI 领域基础设施过度扩张与企业级盈利模式尚未成熟的矛盾；同时通过对 Poppi 并购案例的拆解，提示创始人在面对资本运作与技术浪潮时，必须看清底层现金流与商业控制权的真实归属。

Source quotes

[25:00] The AI slop economy is thriving right now.

Original transcript

[0:00] - Yeah, this weekend, every time I would refresh, it would be, you can see everything takes, everyone had all these different opinions and takes. The one that stuck out to me was that there's this, there's a little bit of a, Like... You watch the movie and most people came away thinking more positively of Elizabeth Holmes being like, oh, maybe she wasn't. Maybe she's not that bad. Maybe there's more to the story. Maybe she was manipulated. She seems to be manipulated multiple times, potentially. Like maybe maybe she's more of a. Some people might be like. She's bad, but... She'd probably be nice to me if I interacted with her. Yeah, yeah, a little bit of that. There's some sort of suspicion that I feel some people in Silicon Valley are having where it's like, oh, it's just like the fake it till you make it thing. It's not that big of a deal. I just can't stop thinking about it. It seems like she's just a bad founder. Like put aside the fraud. Like it just seems like, like, total slop in terms of the execution of that business and gives me absolutely no faith in a Theranos 2.0. Totally. And the whole time I was like almost yelling at the movie screen because The way she describes Theranos is like this box and like it's in the cloud. It's just like it's so interesting. It's so unaware of the medical device industry, which is so frustrating to me because she she's making this point where it's like we even we have an invention and our invention allows you to test blood for things. And like we test all this stuff and she's making it sound like it's like magical science and it's like. I looked it up. There have been 60 plus companies since Theranos started that sold for over $100 million in that category. It's like someone coming to you and being like, I have a magical, we invented something. It takes information and stores it on a database that you can access through a website. It's like, So you make a SaaS product. Literally hundreds of people make those. And there are thousands of people that work in the medical device industry, and they've made boxes that look kind of like a Theranos box, and it runs a test, And they don't go to prison for 10 years. Yeah, they don't go to prison. And she never just says, like, look, this stuff happens all the time. It's not crazy that I'm going to be able to build a box that tests for something. Like this happens all the time. Instead, it's like, I'm the only one that could possibly do this, which is so weird. It's such a bad frame to not just admit that. what's really happening on the field, which is that diagnostic boxes get built and sold every few years for hundreds of millions of dollars. You could just do that. But instead, it has to be this pie in the sky nonsense. And it really, really frustrates me, as you can probably tell. broke down a little bit of this concept. He said, "One of the things that hasn't been talked about with Elizabeth Holmes, she was created in the era of the TED Talk entrepreneur." And this is exactly what I was picking up on. So between 2011 and 2015, after the release of the social network, there was an entire apparatus built around child prodigies, college accelerators, young entrepreneur organizations. I watched countless founders of hard tech companies present fraudulent products at conferences. They had fake studies with fudged numbers. Most of the adults in the room excused it as faking it till you make it. But Holmes was the one who rode the wave better than anyone else. When the TED Talk era came to an end, when people finally came to their senses, she was still writing the same playbook. And in the most scrutinized field, there is medical technology. it in And she got caught when the world suddenly woke up. There's other things, which is that this is strictly regulated by the FDA. So if you're a founder of a medical device company, what you are going for is FDA authorization. That's marketing authorization. That's literally the ability to sell the product. It's not illegal to build a medical device. It's illegal to sell it, to market it. And so that also includes, like, you typically can't go on podcasts and media and say, "Oh, yes, we have this thing that does this thing." We have to say, oh, we're working on this progress. Like, we can't say that it actually works because we haven't been approved to say that by the FDA. Yeah, and we'll just talk away. What spoiler do you want to talk about? So... Billy does this side deal. Yeah. to get a cut of... profit on the back end of the movie. You would probably know what that's exactly called. Yeah, points on the back end. Points on the back end. So he does this thing. Incredibly good. And then later, after the movie's totally been... Yeah. like, finished, he decides, ah, this is going to look pretty bad, backs out, right? Damage has already been done, right? Totally. Like, he looks, he doesn't look great. I And Elizabeth is trying to say, like, "Oh, I don't really know." [5:00] much about the details. So one thing like the damage was already done and now the movie is a massive hit and it's going to be continue to be a massive hit. It seems like it'll be like a cult classic and I can't imagine... being 60 years old, being like, I'm going to turn on some Nathan Fielder. Just take me back to the good old days. But the funny thing is, do you think it made sense for him to do the side deal? Because she has... She claimed in the show she has hundreds of millions of dollars of restitution to pay. So, of course, she wouldn't want to make... Even if she wanted to make money on it, she wouldn't get any benefit of it. So she would want to route it. - That's a good question. There's some ambiguity over whether or not they're married. I couldn't get to the bottom of this, But I think they might not be legally married. in which case, It would make sense for him. to get those profits and then use that for their kids and their family, but then it doesn't flow to her bank account, which immediately gets grabbed by the investors. That's what I'm saying. I think you might be on to something there. She gets a lot of the benefit in that it's sort of funding... Yes. And so is your point that if she's in on the deal sitting there being like, we want to get paid, but just pay to Billy's account, don't pay. That's more tricky from a restitution perspective. Totally. I imagine a judge would be like – Oh, that's an interesting wrinkle. I hadn't thought about that. Yeah, a judge would be like, yeah, you're earning money. I don't care. Yeah, I don't care that you did it. Billy has to like go behind her back and make it really public that she's – Yeah. He's going behind her back. Huh? Yeah, good point. Good point. I mean, the wildest part about The whole you can see everything there knows thing is that one I've talked to Billy on the phone about The Twitter that he was running and all that stuff answered all those questions. Nothing that really was different than what you saw in the movie. But he sort of gave me the pitch for Theranos 2.0, and I was, like, pushing back, being like – So you should go for like a specific diagnostic, go for like do this. Like I have a buddy who ran this exact playbook, made $400 million in a very short amount of time. Like this is tractable. And then he was like, when did you have this conversation? I talked to him like last year. And so my pitch was, I said, look, if we're going to take this Elizabeth Holmes redemption arc seriously, I want to see. some great calls about biotech because this is sort of what's happened with other individuals who have been rehabilitated uh from a branding perspective my point that i posted about was was like i want to see elizabeth holmes give an incredible take about peptides and how that market will evolve and then watch that happen and be like oh wow she was early i want her to be a good biotech analyst i want her to be smart on the science i want her to crazy thing about the movie I can't remember a single point where she was locked in on anything science related. I completely agree. It was just, it was Billy. That's the point. Learning about mitochondria. Yeah, that's my point. Like, like if, if. If a genius scientist goes to jail wrongfully, I would assume that they continue reading science literature and that they come up with interesting theses around that. - One simple trick that the medical industry hates. - You can actually do that, that happens. And I want Elizabeth Holmes to throw down, give me your biotech 2027, give me your AI 2027 thesis. Let's watch it play out if you're right. - At the very least, A market map. At the very least, a market map. Show me that you understand what Roche Farmers are doing. Or a tier list. A tier list. A tier list of biotech acquisitions in the medical device space. Yeah, show me you're a student of the industry you're operating in. Right? Yeah. Watching Nathan have that light bulb moment where they're pitching him Theranos 2.0. Yeah. And them not realizing that people have been using the phrase Theranos 2.0 to describe like massive things that they perceive might be massive fraud. And it was basically just a lab testing machine, previously size of a dishwasher at a lab down the street. You shrink it down to the size of a box and you put it at the actual hospital, and then the hospital can test for a specific thing. They were testing for antibiotic resistant bacteria. They developed a chemistry. It's not rocket science. Like it is hard science, but like it's not magic. It's not future sci-fi magic. And if you can get that data in an hour as opposed to two days, everything in the hospital is just going to move a little bit smoother and people are going to be a little bit healthier. and you're going to sell your company for $425 million, which is exactly what happened. They raised $25 million. It was a great outcome for founders, employees, Roche eventually. Roche eventually. got the product approved by FDA. So in 2019, it got it got it got authorized for THE COBOSS VEVO DX MRSA TEST, WHICH IS LIKE NOT GOING TO BE ON THE STAGE AT TED ANYTIME SOON, BUT IT SHOULD BE. [10:00] - And it actually moved the needle forward. - That being said, I came away from the movie, right? feeling like, okay, Sonny had done some side deal and got way less. You mean Philly? No, no, no, Sonny Balwani, the other co-founder. My impression was that he was getting way less jail time. Elizabeth is going to jail for like 11 years. Yeah, it seemed like Sonny... They were positioning it as like-- Sonny was able to pin it on her, he got out. Yeah. Or at least he got a reduced sentence or whatever. Yeah, yeah, yeah. And you told me afterwards Sonny is in prison for like 12 years. Yeah. It's about a year long. 12 years, 11 months was the sentencing. Elizabeth got a 10-year sentence, reduced originally from 11 years and three months. According to the internet, Sonny Didn't sell secondary either. I know. That's ridiculous. None of them. They both go down with the ship. They both don't make money. That was a jaw-dropping moment for me. I was like, least shrewd person in the world potentially no secondary stuff 11 billion fraud purely for the love of the game what are you doing what are you doing and it's not like i came away being like it wasn't a fraud like it was yeah it was very clear that like everyone there sonny heard they knew it didn't work that really is they didn't have the revenue the profits that's really like the ted entrepreneur yeah it goes to prison for the love of the game for the love of the game insane just just Just for the vision. If you do want to sell secondary, there's one neat trick. One neat trick. This is a new thing. So if you're a founder and you want to sell $10 million a secondary, $100 million a secondary, you don't want it to be a negative signal. Here's what you do. Before, so you've grown your company to the point where you're valued in the billions. you don't want to sell secondary because everyone will be like, oh, he's like selling out. She's selling out like that's not good. good. That's not a good sign. So what you do is you take out an immense amount of debt. So you get like $100 million of debt. Then you sell the secondary, sell another $100 million of secondary, pay off the debt, and then you can tell everyone, look, I just sold secondary to pay off my debt, my personal debt. And everyone assumes it's like, oh, 50, like, yeah, yeah. They think it's credit card debt. They don't realize that it was an equity-backed $100 million line of credit from a apart. But you can still get away with 100 million out of the gate. It's possible. I wonder if anyone's doing that. It might be happening. Who knows? Who knows? Anyway. to 50 gigawatts, wow. 50 gigawatts. 50 gigawatts. That's a lot. That's a lot of compute. And people are going back and forth on this, right? Did Gerstner say it was too much? Something like that? It seems like semi-analysis has the data and has been pulling together a whole bunch of satellite photos and stuff. So we looked at the data. They looked at the satellite photos. I mean, a lot of these things... Like they were papered a long time ago. Even if the market's like having a hiccup right now, that's not going to stop people from actually pulling out of – I sent you this This morning. What do you say? From RevCap. Okay. which is semi-analysis saying if you're worried about demand, ARR, don't be, because there's going to be even more supply than you realize. Of course, semi-analysis is also saying that they have some of the most bullish forecasts for the labs. But it's funny that in the midst of everyone being worried about ARR growth over the last two weeks – that semi-analysis is coming out and saying, don't worry. there's an obscene amount of supply coming online within the next 12 months. YEAH. I believe them on supply. That seems very cut and dry. I don't think anyone has a stronger read on what's actually getting built than them. I think if you look at the satellite photos and you look at the deals that have already been inked, that even though, oh, the Nasdaq sold off 5% last week, like that's not enough to be like, okay, pack up the dump trucks. We're not building this thing. Like maybe in months, if things continue to get bad and the rates go up, like some projects will be delayed or stalled or deals will get unwound. But like right now, It's not that it's full speed ahead. It's just like it's a train. You can't slow it down that quickly. Like it's just going to keep going. On the demand side... I feel like the agentic stuff, the personal agentic stuff is going to be high demand for compute. I'm less clear on what the willingness to pay will be. HOW MANY PEOPLE WILL be willing to pay the equivalent of $1,000 a year for inference related to like their semi-personal prosumer life shopping and all the little agentic features. Now, I think [15:00] Tech is incredible. I've been using Dot a ton and I think having an agent that just runs in the background and just tries to go find little opportunities for you is amazing. It just catches things in my email that I never would have... gone through or taken the time to do and it just sends me a little message being like, Like, you asked for this, and you should follow up, because they said that they would do it. You used it to cancel your roadster order, too, right? Yeah, yeah, yeah. So just little things like that. It's just a better interface. I love interacting with the open web through an LLM. It's amazing. No cookie pop-ups, no ads. It's truly the best consumer experience. There's a lot of controversial... uh, stuff downstream of that. But like, and turn as a consumer, if I'm trying to learn about something instead of having like 25 tabs open where I have to accept cookies 25 times, make sure I'm logged into 25 different things and like scroll, find the fact it's so much better to just have it. Oh, put it all together with the exact quotes. I can click out to the link if I need to. And that's starting to happen in the personal agent space. The question is like, how do you get the money from people? Is it the ads? Is it the interchange? Like there's that sort of bearish take that even that the agentic commerce market, if you assign all these take rates, it's like a $40 billion market or something like that. It's like not big enough to justify. No, I think what needs to happen is something closer to, going from automating tasks to automating certain roles or extending the capabilities of people, right? Because we're in this era of Obviously there's a viral clip going on over the last 24 hours talking about somebody's having a 26-hour day. Um, because they need to stay up. She's changing and manipulating time. Yeah. Because they need to... be online and available to keep their... agents are you saying are you saying she's crazy I'm not saying. She's not crazy. You're crazy. You're crazy for thinking that it takes 24 hours to complete a full day. That's right. Not 26. But anyways, it's like... The next leg up will probably be... She's a special case. She's an entrepreneur. Yeah, the question is like for the 90% of Americans that aren't in... But that's not what we're seeing. We're seeing that one... We're seeing that a very, very small number of companies are driving all of the adoption, and so there's probably not all the adoption, but a massive amount of it. And so there's another leg up there, just with the capabilities increase. And then what you're saying is like, The other 99% of businesses... how are they going to be, how is their spending going to ramp? Will it ramp beyond the, you know, $20, $100, $200 a month subscription? Yeah, yeah. I mean, in the business side, like the Jensen take of like, you know, a software engineer who makes $300K should also get $300K in tokens or something. Like that seems to be happening, basically. It's like, yeah, like a software engineer can deploy AI very efficiently and like at a high ROI. It's like, okay, what should your actual token, your inference budget be? If you gave a sales rep $50,000 a month in inference, like, is there something that they could do with the best models to actually generate $100,000 in value that they wouldn't have otherwise? Like, those are the economic equations that happen across the business. I'm talking more on the personal side, the consumer side. Like, okay. People are using the LLMs. No, and I'm just saying we don't pay for all this. Like, the CapEx is not going to be paid for. By consumers. Yeah. Yeah, I think you're right. So you're saying that is not the leg up that needs to happen. It has to be diffusion into the enterprise and B2B beyond what it is. It has to 10x. So like the small company that has basically everyone on a $200 a month plan and they're maybe spending $100,000 on maybe 10 mil top line, they need to go to a million dollars in for instance a year, even though they're a $10 million business. Maybe they go to 15 million of top line. Speaking of $1 million budgets... What we got? Brett Adcock launched Hark. which is something like a muse or an instinct or a dots. And they gave away or tried to give away a million dollars of pizza. I believe it was on Friday. You just had to download an order from the Hark app. And... I don't know that it went well. I think the summary we have here is an app called Hark tried to do a pizza giveaway. but it quickly turned into a PR disaster. That's the thing. - Yeah, I don't know. I mean, clearly they got a ton of people to download the app. Like, I think it wasn't working. - People just don't want to give them the data? I know people were giving them the data. You had to give them – In order to get this pizza, I will need to edit your tax return. C edit. [20:00] Download and permanently delete your contacts. Delete all of your... See, edit, create, and delete all of your Google Sheets spreadsheets. See, edit, share, and permanently delete all the calendars you can access. Read, compose, and send emails from your Gmail account. See and download your organization's workplace directory. I don't know if all of these were required. Yeah, sometimes you can just not check those and still get access. It's like when you download an app on the iOS app store, you can just say, Like, no thanks. I don't want any notifications. Yeah. Armand tried to get a free pizza and had some issues, and so he used the app to submit a formal complaint to the FTC. Hark said, "Fair, that's annoying." And I just tried again with no luck. I can file with the FTC, but a giveaway that's been flaky for a few minutes on launch night isn't something they'll act on. I'd rather report it straight to Hark's team now with all the details so they can fix it. Want me to send that? No, file a complaint with the federal government. It submits FTC complaint about a Hark pizza promotion. On October 16, 2026, Hark publicly advertised a $1 million free pizza giveaway. promoted on X and stated users could claim it by typing order a free pizza in the hard cap. Crazy. Yeah, interesting. I'm so curious why... Why... Brett didn't just do Hark as part of figure, because the whole... like I don't know. It doesn't make any sense to me. other than you wanted a fresh cap table that could raise. Yeah, I mean, like Elon kind of created the playbook, right? XAI was a separate thing. You raise fresh funds. You get – it's a way – it would be hard for – SpaceX. to go to a top AI scientist and say, we're going to give you $5 billion of equity. Now, Zuck has kind of broken that. and been like, I'm just going to do it. But the way you get the multi-billion dollar outcome is you give a co-founder role of a new company and then you acquire them in. So I think it's probably that. I don't know. The only thing is Elon had sold a lot of cars and launched a lot of rockets prior to doing that. makes more sense in the fullness of time. Now, did they specify if this is real GDP growth? Because they just said GDP growth. Because what if inflation is at like 40 percent? That would be the equivalent of negative 23 percent real GDP growth. You've got to clarify these things if you're going to be asking the tough questions. But let's assume they went with real GDP growth. You think that's low? You think that's high? There's someone at Anthropik who said it would be 28 percent. I like that. The lowest was 6%. Basically, no one there thinks that... GDP growth is going to be, you know, at normal levels. Everyone is expecting a boom within the building. Agree? Disagree? Bullish. Bullish. Drinking the Kool-Aid, I don't know. Just added the FOMC, the FOMC of course, thinks GDP growth will be 2%. That's the government suit. Anthropic staff are obviously like the most bullish. Like they are at 17.1% GDP growth. Reminder, America grows 2%. TYPICALLY. THEY'RE SAYING 17. HUGE ECONOMIC EXPLOSION, UNLIKE ANYTHING WE'VE EVER SEEN IN AMERICA. ABSOLUTELY FANTASTIC. If we grow at 17%, it is so positive sum. There's so much to go around. You can do everything, basically. You can build new roads, you can give people health care. Everything just solves itself because a rising tide lifts all boats. Where would you really put it, though? I'd put a – 200% GDP growth. Yeah. I want to be tripling every year. Yeah. Predict what you want to happen. Manifesting. This is manifestation. I want to manifest it. Site visitors said 12.5%. Again, staggeringly high. But interesting, U.S. adults predict 5.4%. just broadly. which is again, way above where we've been. So like everyone is bullish. - The American mind must be safe. - Spore doesn't buy it. Spore says this is borderline mental illness. I'm sorry, but 20% GDP growth in 2030 is a common belief in Anthropic. What are these people smoking? How does one get so divorced from reality? - They're smoking cigarettes. - They're smoking cigarettes. because lung cancer will be solved. That was reported from – we talked about this a little bit, but that was reported from Shamsankar and has not been widely confirmed beyond his anecdote that he shared, I believe, on Bloomberg or CNBC. I don't think this is – unique to anthropic [25:00] Yeah. I think it like clearly the average American is already insanely. Yeah, exactly. Average Americans are already more than two and a half times. And so if you go to the labs, anybody who like really truly believes in this current revolution, and that probably is like an average across the field. There was a funny exchange in the Tyler Cowen-Dawar Kesh Patel interview on Interesting Times. He was guest hosting for the now-departed show. NOT. passed away but he rossed out that laugh for I think CBS and the free press or something like that but Tyler Cowen and Dwarkesh Patel were going back and forth about AI and Tyler asks to our cash. Like when when do you think I'll walk into a Walmart and see a product that is the product of artificial intelligence? And Dwarakash is like, oh, it's really hard to say. Like, I think that there will be a real effect on the economy soon. 2028. I'm very I'm very bullish on this change in the economy, but I just I can't predict how it'll be. And I was like, There's already AI slop beach towels in Walmart. You can go into Walmart and get AI products now. There's so many pool floaty that looks like a tank that was sold on TikTok via AI video slop. We have signs of signs. Yes, we have signs of signs. The AI slop economy is thriving right now. But the AI slop is being sold. And a lot of people will say, oh, the San Francisco tech people, they need to touch grass. think if they want to get in touch, they need to touch the gas pedal of a Lamborghini Countach or Lamborghini Huracan in Miami. You got to get in to the mind of the core seller, the drop shipper to understand how AI is changing the real economy because that crew is selling AI slop. It's happening. Like there's definitely AI, AI instantiated products up and down the stack all over. Facebook, Instagram Reels. TikTok, this stuff's flying off the shelf. And Sweeney, founder of Epic Games, says the games we've long wanted, which didn't exist because they'd cost several times more to make than they'd earn back, will soon become possible. As will the game, the game's developers wanted to build but couldn't with the resources available. It'll be a new golden age of gaming. Yeah, I like that he is taking the white pill on AI. Three Blue and Brown also put out a big, he's a math YouTuber, does these incredible video explainers, these motion graphic explainers of difficult math concepts. And he put out a video saying like, it is the best time to go into math. Very, very optimistic about it. I saw another... post by him talking about like Look, I had to migrate from C to C++, from 2D to 3D. I've been through so many tech migrations. This is another one. I'm very excited for it. And he'll just adapt. He'll improvise, adapt, overcome. If people are rolling their own Fortnite clone in a web browser, decompiled or whatever, I'm sure he's going to find a way to participate in the upside in some way because he is innovative and flexible. What about this Poppy deal? Have you seen this? Yeah. Quite funny because everyone's trying to use it as an opportunity to try to dunk when it's like they sold their company for 1.9 billion. Yeah. and So wait, let me set the table. So Poppy is a beverage brand. It's a soda. probiotic soda company sold to Pepsi for $2.2 billion. $300 million of that $2.2 billion was an earn out. So they got $1.9 billion. on the day the deal closed, and then $300 million waiting for... Now, Pepsi announced that instead of paying 300 million based on the performance of the brand, they will only be paying 12 million. And the dunk is like, oh, that stings. They didn't get their earn out. But yeah, they got 1.9.12 billion dollars. Like, it's still an incredible amount of money. And people are like, yeah, like here's how you negotiate an earn out. It seemed like they already negotiated it. Like they negotiated it really well. The disaster would have been if the numbers were reversed. $300 million for the company and $1.9 in earn-out. And if they didn't get that earn-out, it would seriously change the economics. But it seems like they got – 90% of the deal value, still pretty good outcome for the founders of Poppy. Amazing outcome. But, yeah, Sean Frank chimes in and says a quarter is what you should expect in an earn out most of the time right off that second bit. that second byte to zero. You shouldn't go into one of these deals expecting to get the full earn out because [30:00] Companies have lots of resources. They control the budget, the distribution, all the success or failure. You're no longer in charge. You're no longer the master of your own destiny. And so... uh... If you put all the chips with them, they might bet them differently than you would expect to.