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
Travis Kalanick joins; NYC pied-à-terre tax and the politics of targeting second homes
The episode opens with Travis Kalanick returning to the show as the hosts react to NYC’s proposed pied-à-terre tax on high-value second homes. They debate whether it improves affordability or destroys demand and development, and touch on the risks of publicly “calling out” specific wealthy owners.
- •Proposed annual pied-à-terre tax aimed at second homes over ~$5M and its likely demand impact
- •Argument that taxing the most “elastic” buyers could crash luxury demand and reduce new construction
- •Debate over “doxxing” wealthy property owners and potential for inciting violence
- •Claim that adding housing units (Austin example) reduces rents despite migration
- •Comparison to London’s experience with stamp duties and non-dom policy changes
Blue-city real estate policy spillovers: transfer taxes, development incentives, and capital flight
The group expands the discussion to broader blue-city property policies like mansion and transfer taxes in LA and SF. They argue high transaction costs freeze markets, reduce mobility, and encourage capital to relocate to friendlier jurisdictions.
- •LA/SF mansion/transfer taxes cited as drying up transaction volume and flipping activity
- •Concern that retroactive or escalating taxes make property ownership feel politically unsafe
- •Claim that wealthy absentee owners pay taxes while using minimal city services
- •Role of “whale” buyers in making new developments pencil out (subsidizing projects)
- •View that hostile policies push global capital to alternative cities/countries
OpenAI’s leaked memo and “identity crisis”: enterprise pivot, Anthropic rivalry, and valuation skepticism
Jason tees up a leaked internal OpenAI memo criticizing Anthropic’s revenue accounting and safety posture, while outlining OpenAI’s push toward enterprise and the agent platform layer. The panel debates whether OpenAI should focus on consumer dominance or chase higher-ARPU enterprise adoption.
- •Memo alleges Anthropic run-rate is inflated via revenue-share/accounting; rhetoric about “fear and restriction”
- •OpenAI’s stated priority: enterprise customers and the agent/platform layer; talent moves from open source
- •Investor critique: OpenAI is unfocused despite ChatGPT’s massive consumer footprint
- •Discussion of competitive traffic/share shifts (Claude/Gemini gaining)
- •Valuation pressure: secondary markets allegedly pricing Anthropic above OpenAI (the “flippening”)
AI flywheels: growth, network effects, and when capital subsidy loses to revenue efficiency
Travis frames the competition like marketplace wars: growth rate and scale create compounding advantages. Friedberg and Sacks add that cadence of releases and revenue quality may outlast pure fundraising power, especially when compute costs bite.
- •Travis: growth is “king” because scale drives reinforcement learning and token-volume advantages
- •Friedberg: Anthropic’s release cadence and operational momentum feel unprecedented
- •Sacks: efficiency and contribution margins can beat capital subsidy over time
- •Coding use-cases as the current enterprise wedge where willingness-to-pay is highest
- •Risk of over-raising and outrunning revenue if public markets won’t buy the story
Big Tech compute dominance and the coming infrastructure squeeze for frontier labs
The hosts explore whether hyperscalers (and players like xAI/Meta) can kneecap frontier labs by controlling compute supply. They argue the frontier labs are hitting a phase change where owning infrastructure becomes strategic, not optional.
- •Hyperscalers reportedly control a majority of compute, shaping frontier labs’ access and pricing
- •Frontier labs face strategic dependency if they must “tin cup” for capacity from cloud providers
- •Compute throttling compared to the “Friendster was slow” failure mode
- •xAI/Colossus and Meta cluster scale cited as examples of compute arms race
- •Idea: overbuild compute to privilege your own models and sell surplus capacity
Compute constraints meet product constraints: token budgets, “vibe-coded slop,” and enterprise ROI scrutiny
Chamath argues the next phase is customers refusing to subsidize unlimited tokens, forcing hard ROI conversations. The group discusses how agent-driven development can generate low-quality output unless properly managed and budgeted.
- •Enterprises hitting token budgets; costs may shift from labs to customers as subsidies end
- •Agent outputs can be “slop” without guardrails, taste, and human-in-the-loop workflows
- •Uncertainty about how to redesign org structures, budgeting, and governance around agents
- •Coding tools both generate revenue and accelerate internal product development (double flywheel)
- •Physical constraints (compute/electricity) become the limiter as usage scales
Allbirds’ “AI pivot” stock surge and what it signals about datacenter scarcity
A comedic setup about Allbirds pivoting from sneakers to AI transitions into a serious conversation about market behavior and compute scarcity. Chamath claims the market is correctly pricing extreme compute constraints and entitlement/power bottlenecks.
- •Allbirds stock spike framed as late-cycle/bubble-like “name change” behavior
- •Chamath: capital markets are reacting to massive compute constraints (power, land, approvals)
- •Local backlash: votes and elections overturning datacenter approvals; some regions banning builds
- •Behind-the-meter power solutions (gas/onsite generation) highlighted as a workaround
- •Frontier labs face a “five-alarm fire” if compute bottlenecks cap growth despite demand
Why datacenters are getting blocked: ratepayer fears, doomer strategy, and populist resentment
Sacks lays out multiple constituencies driving datacenter opposition—from grid rate concerns to organized “AI doomer” activism and political alliances. Another host adds that datacenters have become a symbol of elite wealth and unequal gains, fueling broader populism.
- •Community concern: datacenters drawing from the grid could raise residential rates
- •“Ratepayer protection” concept: new builds should bring their own power / be grid-neutral
- •Claim that doomer groups astroturf NIMBY arguments (water, local impacts) to slow AI progress
- •Critique of Anthropic’s political alignment with anti-datacenter coalitions as potentially self-defeating
- •Populist framing: datacenters as a physical emblem of elite advantage and job displacement fears
Game break: “The Price Is Wrong” ROUND 1 (overvalued startup trivia)
The hosts pivot into a light game segment guessing famously overvalued startups. The round lands on NFT-era and pandemic-era hype examples, with playful banter and scoring.
- •OpenSea guessed as the $13B NFT marketplace example
- •Clubhouse identified as the $4B audio social app boom-and-bust
- •Juicero named as the overhyped “juice iPhone” gadget startup
- •Discussion about how failed unicorns vanish from memory
- •Jokes about making the segment a recurring feature
Eric Swalwell exits CA governor race and resigns: what insiders knew and the “party machine” theory
The panel discusses allegations around Eric Swalwell and the timing of his exit from the governor race and Congress. Friedberg claims he heard similar rumors months earlier and is struck by how widely known—but withheld—the allegations were, suggesting coordinated political timing.
- •Friedberg recounts hearing allegations months prior from multiple sources
- •Emphasis that claims are allegations; due process and “allegeds” repeatedly stated
- •Question of why victims/insiders waited and why disclosure appeared coordinated
- •Sacks: CA “jungle primary” incentives push Democrats to consolidate to avoid two Republicans in runoff
- •Parallels drawn to Biden’s pressured exit and Pelosi’s influence within party leadership
Markets at highs amid Iran conflict: predicting de-escalation, ‘TACO’ optics, and risk-off signals
They debate why equities remain strong during geopolitical conflict, arguing markets are pricing a resolution. Chamath points to valuation indicators (Shiller P/E, Buffett indicator) suggesting caution even as momentum and dispersion complicate the read.
- •Sacks: markets appear to price a near-term resolution and limited military scope (as stated publicly)
- •Travis: Trump uses the stock market as a “weather vane,” creating a repeatable volatility pattern
- •Chamath: valuation metrics are near all-time highs, suggesting risk-off posture
- •Dispersion: a small set of companies drive highs while broad market lags, confusing signals
- •Concern that investors can cherry-pick data to justify biases in an uncertain regime
Is AI delivering profits yet? Productivity vs transformation friction in big enterprises
The hosts argue over whether AI gains are already translating into scaled profits for large companies. Travis emphasizes change management as the real bottleneck, while Jason argues early adopters are outcompeting laggards, especially in startups and tech-forward orgs.
- •Chamath: asks for concrete examples of scaled enterprise profit uplift, not just small wins
- •Jason: cites startups and portfolio companies seeing strong productivity and revenue traction
- •Travis: big-company complexity and undocumented processes make transformation slow
- •Sacks: bullish long-term; ROI improving at the model layer, application layer still forming
- •Consensus: agents aren’t “AGI”; they need human oversight and can fail basic reasoning (e.g., betting both sides)
Game break: “The Price Is Wrong” ROUND 2 + bonus (Theranos, Quibi) and closing plugs
A second game round continues the overvalued startup theme, adding notorious examples and a bonus for Travis. The episode closes with event plugs and final banter.
- •Theranos identified as the $9B valuation fraud case with founder prison sentence
- •Quibi identified as the $1.7B short-form streaming flop shut down in ~6 months
- •Travis gets a featured turn in the game segment
- •Promotion of Liquidity event capacity and All-In Summit ticket urgency
- •Closing jokes and sign-offs with the ‘besties’ dynamic
