All-In PodcastWhy the datacenter backlash could cap America's AI lead
Ratepayer groups, doomer activists, and local politicians are blocking permits; Allbirds' AI-pivot stock surge shows how scarce compute has become.
CHAPTERS
- 0:00 – 10:02
Travis Kalanick joins: NYC pied-à-terre tax and the fallout for housing & cities
The besties welcome Travis Kalanick, then dive into NYC’s proposed pied-à-terre tax and what it signals about taxing wealth and second homes. They debate impacts on demand, development incentives, city revenues, and whether these policies improve affordability or hollow out cities.
- •What the pied-à-terre tax targets (second homes over ~$5M) and why elasticity matters
- •How taxing the most mobile buyers could depress high-end demand and broader development
- •Doxing/pointing at billionaire homes and the risk of escalating political violence
- •Comparisons to London’s hollowed-out neighborhoods and capital flight
- •Austin/Texas as a contrasting case where building supply reduced rents despite migration
- 10:02 – 11:23
Blue-state real estate friction: mansion taxes, transaction volume, and ‘property isn’t safe’
The conversation broadens from NYC to LA and San Francisco’s mansion/transfer taxes and their effects on transaction volume and investment behavior. They argue that high transaction costs and retroactive-feeling policies chill liquidity and push capital elsewhere.
- •LA and SF transfer/mansion taxes and how they change flipping and transaction volume
- •Argument that wealthy owners pay taxes but use few city services, making them net contributors
- •Developer underwriting: “whale” buyers can make projects pencil and subsidize new supply
- •Claim that arbitrary/retroactive taxes increase perceived political risk of owning property
- •Capital reallocation to more hospitable jurisdictions (Zurich/Lugano/Milan examples)
- 11:23 – 15:23
OpenAI’s leaked memo & enterprise pivot: competing narratives vs Anthropic
Jason introduces a leaked OpenAI memo criticizing Anthropic’s revenue accounting and framing Anthropic’s approach as restrictive and fear-based. The panel discusses OpenAI’s pivot toward enterprise customers and “agent platform” ambitions amid growing competition.
- •Leaked memo themes: Anthropic revenue-share ‘inflation’ and ideological positioning
- •OpenAI’s push to win the agent/platform layer and enterprise adoption
- •Investor complaints about OpenAI focus vs consumer dominance of ChatGPT
- •Market-share context: more overall users but more competition from Gemini/Claude/Meta
- •How internal memos can be used as PR/valuation warfare
- 15:23 – 26:42
Frontier model race: growth rates, network effects, and the ‘flippening’ debate
They assess OpenAI vs Anthropic using a startup competition lens: growth, scale, and flywheels. Travis emphasizes network effects and scale; others discuss cadence of releases and organizational momentum as potential differentiators.
- •Why growth is ‘king’ in frontier AI and how scale can compound advantages
- •Travis’s network-effects analogy from Uber: volume drives defensibility
- •Friedberg’s observation: Anthropic’s rapid release cadence and internal adoption shift
- •Debate over valuations (OpenAI ~$850B) and secondary markets favoring Anthropic
- •Risk of ‘running off the cliff’: capital deployment outrunning sustainable revenue
- 26:42 – 31:12
Compute dominance & hyperscaler leverage: why infrastructure control shapes winners
The discussion shifts to compute as a strategic chokepoint: hyperscalers control a large share of capacity, which can throttle frontier labs. They argue frontier labs may need their own infrastructure to avoid dependency and ensure reliable scaling.
- •Compute concentration: hyperscalers’ control and the game theory of throttling rivals
- •Why frontier labs may need vertical integration into data centers and power
- •Enterprise token budgets and the risk of cost pass-through exposing ‘slop’ outputs
- •Friendster analogy: performance constraints can kill leaders despite product demand
- •Elon/xAI/Colossus and Meta clusters as examples of aggressive capacity build-out
- 31:12 – 32:32
Mythos/Opus scarcity and ‘doomer NIMBYism’: was safety rhetoric also a compute constraint?
They explore the theory that Anthropic’s withholding/positioning of certain models may reflect serving cost and compute limits as much as safety concerns. The panel connects safety politics, marketing, and infrastructure realities.
- •Theory: holding back a costly model can create scarcity + safety PR while managing compute
- •Possible trade-off: ‘thinking time’ reductions as a signal of capacity constraints
- •Anthropic’s earlier alignment with anti-data-center sentiment vs future need to build
- •How model serving costs can dictate product strategy and launch timing
- •Speculation on whether policy/doomer narratives shift when incentives change
- 32:32 – 33:53
Polymarket gag: ‘Anthropic buying All-In’ and the episode’s meta-humor break
A comedic interlude erupts when Jason cites a Polymarket contract claiming odds of Anthropic buying the All-In Podcast. The group questions whether the market is real and riffs on private jets and ‘trickle-down avionics.’
- •Prediction-market contracts can be created by anyone, raising credibility questions
- •The show’s self-referential humor about being acquired by an AI lab
- •Jokes about liquidity events, upgrading planes, and ‘trickle-down’ themes
- •How hype cycles and speculative markets blur into entertainment
- •Transition setup into the next segment on bubble behavior
- 33:53 – 39:20
Allbirds’ ‘AI pivot’ and bubble behavior: name changes, ZIRP hangover, and capital markets signals
Allbirds’ stock surge after an AI/data-center pivot becomes a proxy for late-cycle market psychology. The group revisits 2021 ZIRP-era valuation excesses and why physical-world businesses were priced like software.
- •Allbirds’ brand asset sale vs prior IPO valuation as a case study in hype/mean reversion
- •Late-’90s dot-com parallels: name/pivot narratives driving short-term pops
- •Travis’s view: 2021 valuations extrapolated 2–3 years of growth, not 1 year
- •Confusion between revenue growth and unit economics/gross margin reality
- •Regulatory dynamics and how ‘pivots’ can exploit thematic manias
- 39:20 – 54:37
Datacenter backlash & compute scarcity: power, permits, local politics, and populist resentment
Chamath argues the market is waking up to severe compute constraints, and local opposition is increasingly blocking new data centers. Sacks adds categories of resistance—ratepayer fears, doomer astroturfing, and competitive politics—while another panelist frames data centers as a populist target representing elite wealth.
- •Constraints: power availability, interconnect delays, land/entitlements, permitting friction
- •Local politics: approvals reversed, boards ousted, and outright bans (e.g., Maine)
- •Ratepayer fear and the ‘bring your own power’/behind-the-meter strategy
- •Doomer groups using water/power narratives as ‘meet people where they are’ messaging
- •Populism thesis: data centers as a visible symbol of unequal tech gains and resentment
- 54:37 – 59:20
The Price Is Wrong (Round 1): overvalued startup trivia as a cautionary history
The show turns into a game segment identifying once-overhyped startups/unicorns. The humor underscores how quickly markets can overprice trends and then forget failures.
- •OpenSea as NFT-era valuation excess
- •Clubhouse as COVID-era social audio boom-and-bust
- •Juicero as hardware hype vs reality
- •Theme: collective delusion and narrative-driven valuation spikes
- •Entertainment format used to reinforce investing lessons
- 59:20 – 1:10:32
Swalwell exits: allegations, insider ‘machine’ dynamics, and California’s jungle primary incentives
They discuss Eric Swalwell dropping out and resigning amid allegations, focusing on why damaging information can be held until politically useful. Sacks argues party insiders may force exits to reduce risk in California’s top-two primary system and draws parallels to how Biden’s withdrawal unfolded.
- •Friedberg’s anecdote: rumors circulating months earlier and coordinated timing of release
- •Emphasis on ‘alleged’ and due process while analyzing political incentives
- •California’s jungle primary could yield two Republicans if Democrats fragment
- •Sacks’s theory: party power brokers ‘lance the boil’ to avoid late-stage scandals
- •Pelosi’s role as a perceived gatekeeper/mentor figure in Democratic machine politics
- 1:10:32 – 1:23:46
State of the market: Iran ‘peace pricing,’ valuation signals, and AI’s uneven ROI so far
The panel weighs why equities are strong despite geopolitical conflict, with Sacks citing markets pricing a resolution. Chamath highlights conflicting indicators (Shiller P/E, Buffett Index vs dispersion patterns) and debates whether AI is translating into profits at scale yet.
- •Sacks: market as prediction engine pricing in conflict de-escalation
- •Travis: ‘stock market is Trump’s weather vane’ and traders learning the pattern
- •Valuation caution: Shiller P/E and Buffett Index near highs vs bullish dispersion stats
- •Debate: AI productivity gains visible in pockets vs unclear scaled enterprise profit proof
- •Tension between model-layer revenue surge and application-layer ROI realization
- 1:23:46 – 1:26:30
AI in enterprises: change management, ‘agents aren’t that smart,’ and what comes next
They close by noting enterprise transformation is constrained by human and process complexity, not just tech. Travis describes CEOs accelerating shipping with AI, but warns agents remain error-prone and require guardrails and humans-in-the-loop.
- •Change management as the bottleneck in large organizations (‘the crats’)
- •Founder-led tech companies adopting AI faster than complex incumbents
- •Agents’ limitations: lack of taste, getting lost, making basic investing mistakes
- •Need for guardrails, budgeting discipline, and human oversight
- •Market implication: productivity is real but uneven and not yet ‘AGI’
- 1:26:30 – 1:30:56
The Price Is Wrong (Round 2) + outro plugs: Theranos, Quibi, and event promotions
A bonus round of the game features infamous startup collapses, then the hosts wrap with announcements about Liquidity and the All-In Summit. The episode ends with playful banter and sign-offs.
- •Theranos as the archetype of fundraising scale vs fraudulent reality
- •Quibi as Hollywood/Silicon Valley capital misread on product-market fit
- •How hype and capital can’t compensate for fundamentals
- •Event promotion: Liquidity and All-In Summit ticket urgency
- •Final bestie banter and closing catchphrases
