All-In PodcastAnthropic S-1 Shocker, Greatest Business Ever? Google’s $500B Wipeout, Midterm Predictions
CHAPTERS
- 0:00 – 3:20
Besties reunite: “super intelligence” talk, travel updates, and card-game roasting
Jason opens the show by welcoming listeners and celebrating David Sacks’ return, then the group riffs on media coverage of their terminology and a “fake news” article. They swap personal updates—Friedberg’s work in Kenya, a long card game, and jokes about who’s allowed back at the table—before pivoting to the main agenda.
- •Jason and Sacks argue over an article claiming the show cooled on the term “super intelligence,” calling it inaccurate
- •Friedberg shares he’s in Kenya with Oholo and mentions farmers trying “true potato seed”
- •The group jokes about a marathon card game and complaints about Jason talking nonstop
- •Chamath banters about enterprise sales, quotas, and learning to focus on “singles and doubles”
- 3:20 – 5:52
Anthropic’s S-1 and IPO framing: massive raise, valuation expectations, and founder control
Jason tees up Anthropic’s S-1 as a potentially historic IPO, citing timing, ticker, and extremely large capital targets. He highlights rapid reported revenue growth, a huge total addressable market claim, and an unusual voting-control structure concentrated with founders and a nonprofit-related setup.
- •Jason describes the S-1 as positioning Anthropic for a huge IPO with an “Ant” ticker
- •He cites very large capital-raise expectations and a potential $1.5–$2T valuation range
- •Jason emphasizes reported revenue scale and growth rates, plus a claimed $40T TAM
- •He notes a control structure where founders retain majority voting power and mentions a nonprofit/benefit-style entity angle
- 5:52 – 11:12
Sacks’ read: the “loss” narrative collapses, capital efficiency, and improving unit economics
Sacks argues the S-1 refutes a common media storyline about massive losses, framing it as accounting effects from convertible notes rather than operating cash burn. He stresses Anthropic’s capital efficiency, improving margins, and the company’s “compute multipliers” concept—algorithmic and architectural gains that reduce costs while expanding capability.
- •Sacks says the widely repeated loss figure was driven by convertible-note accounting as valuation rose, not cash loss
- •He claims Anthropic raised enormous capital but spent relatively little, leaving major cash reserves
- •Sacks highlights improving gross margins and year-over-year efficiency gains in delivering tasks
- •He explains Anthropic’s model: scaling raw compute plus “compute multipliers” like algorithm/data/architecture improvements
- 11:12 – 18:00
Friedberg’s takeaway: enterprise penetration is early, and the whole AI value stack is profitable
Friedberg focuses on signs of real commercial adoption, citing the ramp in customers spending $100K+ annually and arguing this implies early penetration against a large potential base. He shares a value-stack sketch (chip economics to infrastructure to model economics), concluding that profitability appears strong across layers and may reflect real productivity gains downstream.
- •Friedberg points to rapid growth in customers spending $100K+ and argues this suggests low market penetration so far
- •He estimates the number of businesses for whom $100K is a small fraction of spend to frame potential customer base
- •He describes margins across layers—from chip production economics to infrastructure depreciation to model gross margin
- •He flags uncertainty about customer ROI but says continued spend implies customers are seeing value
- 18:00 – 36:08
Chamath’s S-1 method: risk-factor word clouds, compute as the binding constraint, and regulation close behind
Chamath walks through how he dissects an S-1, starting with risk factors and what management emphasizes most. He says Anthropic’s dominant risk is compute—especially data centers and power—followed by regulation, privacy/IP, safety, and only later competition or governance, implying strong demand confidence.
- •Chamath says the most-mentioned risk category is compute, including data centers, power, chips, and grid constraints
- •He ranks other concerns as regulation/export controls, privacy/IP and data retention, safety, then profitability and competitors
- •He argues the risk emphasis implies the company isn’t worried about customer traction as much as supply constraints
- •Jason challenges Chamath’s political interpretation of incentives around accelerating or slowing data-center buildout
- 36:08 – 42:55
Data centers, power, and politics: government vs private AI, and why capital could trigger backlash
The discussion widens from Anthropic to the broader “AI complex,” with concerns that compute buildout is constrained by permitting, utilities, and equipment timelines. Friedberg argues governments and organizations compete for resources and power, predicting future regulation of compute access as AI organizations attract more capital and influence.
- •Chamath argues the U.S. economy is increasingly levered to a handful of AI-linked companies and their CapEx
- •Jason and Chamath debate whether public sentiment against “oligarchs” could impede compute buildout
- •Friedberg frames governments as competing organizations that may try to subsume or control AI through compute regulation
- •The group discusses narrative warfare—each side claiming it benefits society more—as the conflict intensifies
- 42:55 – 46:41
Influence and “doomer” tension: lobbying, regulatory capture fears, and Anthropic’s RSI push
The group debates whether Anthropic is publicly championing political allies versus working through PACs and a revolving door of government hires. Sacks then argues that Anthropic’s actions—especially accelerating recursive self-improvement—conflict with “doomer” rhetoric, suggesting a coming split between safety-aligned narratives and frontier racing.
- •Chamath argues Anthropic should be more publicly political; Sacks counters they already fund PACs and built a major government affairs operation
- •Sacks says former government officials have moved into Anthropic, describing it as a strong influence strategy
- •Sacks cites reporting and Anthropic materials to claim they’re leading on recursive self-improvement to create efficiency gains
- •He suggests Anthropic’s “pacing” rhetoric may function as an alibi while they continue racing the frontier
- 46:41 – 54:29
How to price the IPO: customer ROI scaling vs a systemic bond-market “crowding out” risk
Chamath returns to an investor-style framework: the IPO price depends on how quickly customers expand from smaller spends to large multimillion-dollar commitments, which requires credible ROI proof. Separately, he argues the AI funding appetite could compete with U.S. Treasuries, creating higher-rate scenarios that either raise AI borrowing costs or raise economy-wide rates—both introducing market risk.
- •Chamath frames a core execution question: when do customers justify moving from sub-$1M spend to larger enterprise commitments?
- •He describes a second category of risk as systemic: AI borrowers competing with Treasury issuance for capital
- •He outlines two downside pathways: higher coupons for AI debt cascades through the ecosystem, or higher Treasury yields raise broad borrowing costs
- •Jason counters that a dominant AI franchise could keep capital flowing to U.S. assets despite the scale
- 54:29 – 1:01:54
Demand, pricing, and adoption: Jevons effects, open-model threats, and enterprise security friction
The besties debate whether open models can commoditize key workloads and how Anthropic’s price cuts affect demand. They argue cheaper tokens can increase usage, and that newer applications (like agents) consume many more tokens than simple prompting. Jason adds that enterprise adoption is still limited, partly because companies restrict connecting internal systems due to security concerns.
- •Sacks discusses scuttlebutt of temporary revenue slowing and attributes some to limiting distillation and lowering token costs
- •Jason and Sacks describe Jevons-style dynamics: lower cost drives more usage, keeping overall revenue opportunity large
- •Sacks flags open-source commoditization as a risk, especially in segments like coding or other saturated tasks
- •Jason argues enterprise use is constrained by cyber fears and limited system integration, leaving a large adoption runway
- 1:01:54 – 1:03:30
Shift to Alphabet: a $500B market-cap hit and worries about search monetization
Jason transitions to Alphabet’s sharp selloff, attributing it to concerns about search revenue dynamics and the long-term disruption from AI interfaces that “just give the answer.” He notes Google Cloud strength but frames investor anxiety around whether the core cost-per-click engine is decelerating and whether free cash flow and click behavior are signaling trouble.
- •Jason says Alphabet dropped sharply after earnings and cites the market-cap impact
- •He attributes concern to consecutive misses in search revenue and a mismatch between search growth and ad clicks
- •Jason highlights Google Cloud strength while noting investor focus remains on the search engine funding the rest
- •He argues AI answer-style products and agents pressure the traditional search-and-ads model
- 1:03:30 – 1:07:49
Friedberg on Google: Cloud momentum vs the big question—what replaces consumer search economics?
Friedberg calls the situation “complicated,” pointing to strong progress in Google Cloud profitability while questioning the durability of core ad revenue and whether pricing changes are masking weakness. He says the key issue is whether Google can build a consumer AI revenue engine that ramps fast enough to offset any flattening or decline in core search.
- •Friedberg says valuation depends on reliability and growth of future cash flows, which hinges on the ad franchise trajectory
- •He highlights Google Cloud’s operating income improvement and the value of built compute infrastructure
- •He raises concerns that core services growth may reflect ad pricing/auction changes and could face advertiser pushback
- •He questions internal strategic clarity: whether Google is prioritizing models or renting compute, and how that shows up in earnings messaging
- 1:07:49 – 1:11:34
Chamath’s critique: ad-auction controversy, investor perception shift, and “keep Demis at Google”
Chamath focuses on claims circulating among advertisers about changes to Google’s auction mechanics and the idea that the company may be “pulling money from the future into the present.” He also argues Alphabet may be transitioning into a more diversified conglomerate profile, which changes the buyer base. He ends with a single stated priority: retaining Demis.
- •Chamath cites a widely viewed advertiser thread alleging changes in auction dynamics and degraded performance
- •He says if Google abandoned a Vickrey-style auction, it could be over-earning and risk backlash as advertisers notice
- •He argues Alphabet’s revenue mix shift could make it look more like a conglomerate, changing which investors want to own it
- •Chamath emphasizes keeping Demis at Google as the central strategic task and says he bought shares during the show
- 1:11:34 – 1:14:10
Sacks’ counterweight: search still growing, diversification helps, and the stock may be oversold
Sacks pushes back on the idea that search is already collapsing, noting that while search is a smaller share of revenue, its absolute revenue still grew. He argues that faster growth in other segments—especially Google Cloud—explains the mix shift, and that diversification is a long-sought goal. Chamath agrees on being long, while Friedberg notes Alphabet’s stakes in other major companies.
- •Sacks argues the narrative is overstated: search share is down but year-over-year search revenue is still up
- •He credits Google Cloud’s rapid growth as a key driver of the revenue-mix change
- •Sacks calls the selloff potentially an overreaction, while avoiding explicit recommendations
- •Friedberg adds that Alphabet owns meaningful stakes in other major ventures mentioned earlier
- 1:14:10 – 1:25:46
Midterm scenarios: distrust of polls, Polymarket debate, and forecasts of post-election chaos
The conversation turns to midterms, with Chamath saying he’s confused by misinformation but expects a “Democratic spritz” and outlines how congressional control could drive subpoenas and “FUD.” Sacks argues polls are biased or even fraudulent and urges ignoring media narratives, while also saying a normal midterm pattern would favor the out-of-power party. Jason predicts Democrats winning both chambers and lays out a contentious two-year outlook including investigations and impeachment efforts.
- •Chamath predicts a likely split outcome and says Democrats would use subpoenas, disclosures, and investigations to generate “FUD”
- •Sacks says polls are unreliable due to admitted hoaxes and recurring pro-Democrat polling error, and warns about media-driven turnout effects
- •Sacks says it’s typical for the out-of-power party to do well in midterms and expects something closer to normal than “tsunami” coverage
- •Jason predicts Democrats take Senate and House and anticipates investigations, impeachment moves, and heightened political conflict
