All-In Podcast

Why Anthropic could be the most powerful monopoly ever made

Sacks compares Anthropic's current trajectory to Standard Oil; the SpaceX compute deal eases supply constraints and turns Elon into a hyperscaler.

DSDavid SackshostJCJason CalacanishostCPChamath Palihapitiyahost

May 8, 2026·1h 22m·Watch on YouTube ↗

🙋 FREQUENTLY ASKED QUESTIONS

Direct answers grounded in the episode transcript. Tap any timestamp to verify against the source.

1. 1

   Why did Chamath say Anthropic and OpenAI revenue is supply constrained?

   Chamath argues Anthropic and OpenAI are limited by power and data centers, not demand. He says their revenue performance has zero to do with demand and entirely to do with supply constraints, specifically power. With infinite power, he thinks revenue would be even more parabolic, so short-term forecast beats or misses mean little compared with a five-year view. The same bottleneck makes the SpaceX-Anthropic deal valuable. Chamath says about nine gigawatts were expected to come online that year, but almost fifty percent was being protested, which would tighten supply further if projects were turned off. Elon, in his view, secured power and built infrastructure before most people saw the constraint. That lets SpaceX monetize terrestrial capacity, subsidize Grok training, and strengthen the valuation story while Anthropic gets more compute.

   ▸ 6:01 in transcript
2. 2

   What is Elon Web Services in the SpaceX Anthropic deal?

   Elon Web Services is the hosts' shorthand for SpaceX becoming a hyperscaler-like compute supplier. Jason frames the deal as more than a one-off lease to Anthropic. He compares the opportunity with Amazon Web Services, Azure, and Google Cloud, then argues that Elon's strengths line up unusually well with the business: building large factories, deploying batteries, and using solar and distributed energy. In his version, data centers are basically giant factories, and SpaceX can extend compute from terrestrial facilities into homes through Powerwalls, Teslas, and Starlink-connected distributed systems. He also points to the long-term idea of moving compute into space, noting that the Anthropic announcement included interest in space. The practical implication is that SpaceX could add a neo-cloud business on top of Starlink and launch services, while giving Anthropic the compute it needs.

   ▸ 20:25 in transcript
3. 3

   Why did Sacks call Anthropic a potential AI monopoly?

   Sacks ties the monopoly worry to Anthropic's growth curve and the structure of tech markets. He says tech markets often consolidate into monopolies or duopolies, and he argues that only Anthropic and OpenAI were making substantial AI revenue at that point. In his framing, OpenAI's 3x to 4x growth was already incredible, but Anthropic's 10x annual growth was the exceptional signal. If that pace continued for another 18 months, he says Anthropic could become the most valuable company in human history and have unprecedented control over the most important technology of the era. He does not say Anthropic already has a monopoly. Later, he clarifies that competition should be the North Star, but warns that safety rhetoric and regulatory capture could strengthen a monopoly or duopoly if policy raises barriers for challengers.

   ▸ 27:20 in transcript
4. 4

   What was the FDA for AI controversy on All-In?

   Sacks says the FDA for AI framing overstated what the White House was actually considering. He argues that no senior official supported an FDA-style approval regime for AI models, and he says the idea spread through media commentary after reports about an AI working group and model review process. His distinction is between targeted rules and a broad Washington approval system. He says the White House had already released a national AI regulatory framework with legislative ideas, so the administration was not against every law or guardrail. But he rejects a giant power grab that would squash innovation. The real issue, in his view, was cyber. He says models like Mythos and an OpenAI model were becoming cyber capable, so systems needed to be hardened and code bases scanned before hackers gained the same capabilities.

   ▸ 43:40 in transcript
5. 5

   What does KYC mean for AI models like Mythos?

   KYC for AI models means identifying preview users before giving them access to powerful APIs. Jason defines KYC as know your customer, and Chamath applies it to Mythos by asking whether users should identify themselves so the labs can tell they are not state-sponsored actors or bad guys. Sacks agrees that basic KYC makes sense before handing a super powerful model API to a company or actor. His goal is not to block the tools broadly, but to get them quickly into the hands of more good guys, especially during the preview period. He says KYC is a predicate for that because the labs need to know who those good users are. After general release, he is less sure KYC matters as much, because many more people will have access.

   ▸ 48:10 in transcript

Answers are AI-generated from the transcript and may contain errors. Tap a question to verify against the source.