Elon's Anthropic Deal, The Next AI Monopoly?, "FDA for AI" Panic, Trading the AI Boom

TL;DR

XAI leased Colossus 1 to Anthropic solving its compute crisis, raising monopoly fears and triggering White House debate over AI safety regulation.

Key Points

• 1.XAI leased its entire Colossus 1 data center to Anthropic. The deal added 220,000+ Nvidia GPUs and 300 megawatts of energy, immediately doubling Claude code rate limits and removing peak usage caps for paid users.
• 2.Elon is building 'Elon Web Services' (EWS) to compete directly with AWS, Azure, and GCP. Brad Gerstner estimates the deal generates an incremental $4–5B in revenue this year on top of analyst mid-$20B estimates, subsidizing XAI's Grok training costs.
• 3.Chamath predicted the Elon-Anthropic deal on the previous episode, five days before it happened. He argued Anthropic and OpenAI's revenue is 100% supply-constrained — unlimited power would make their growth even more parabolic.
• 4.Anthropic grew from $10B ARR to $44B ARR in just four months — from January to April. Sacks called it unprecedented even by Silicon Valley exponential standards, with the company now tracking toward ~$100B ARR by year-end.
• 5.Sacks raised the specter of Anthropic becoming 'the biggest monopoly in human history.' He used the John D. Rockefeller / 'Safe Oil' analogy to argue that AI safety rhetoric could serve as cover for regulatory capture that entrenches the duopoly.
• 6.Brad pushed back hard, arguing the monopoly framing is premature and dangerous. He noted Google, Amazon, and other hyperscalers have hundreds of billions in free cash flow to compete, and warned DC picking winners at the AI starting line would be disastrous.
• 7.The White House is reportedly considering an 'FDA for AI' to vet new models for safety. The catalyst was Anthropic's Mythos model, which spooked officials; Kevin Hassett confirmed studying an executive order on Fox Business, though the White House chief of staff later appeared to dismiss it.
• 8.Both Sacks and Gerstner said the 'FDA for AI' is largely fake news amplified by the New York Times. Sacks spoke to Hassett directly and said no senior official supports a pre-approval regime; he pointed to a March 20th national AI regulatory framework he helped craft as the actual policy.
• 9.The panel agreed KYC (Know Your Customer) wrappers for frontier model APIs are a reasonable guardrail. Sacks argued the goal should be rapidly getting powerful cyber-capable tools like Mythos into the hands of vetted cybersecurity firms like Palo Alto Networks and CrowdStrike.
• 10.Chamath argued tech leaders are earning a 'D-minus trending to F' on communicating AI's benefits. He said the regulatory backlash is inevitable when society sees three people gaining trillion-dollar net worths without visible broad-based reinvestment in America.
• 11.Data center protest campaigns are blocking ~50% of the ~9 gigawatts of capacity set to come online this year. Brad argued these are highly organized national activist campaigns — not organic local opposition — analogous to anti-nuclear protests that halted U.S. reactor construction 30 years ago.