The $40 Billion Clause Nobody Noticed in the SpaceX S-1

The SpaceX S-1 reads like a space company prospectus on the surface, but the detail that changes the valuation math is a contingent compute deal with Anthropic worth up to $40 billion.

The Biggest IPO Ever Has an AI Problem

SpaceX filed its S-1 with the SEC ahead of what could be the largest IPO in history, targeting a Nasdaq listing under SPCX with Elon Musk occupying CEO, CTO, and Chairman simultaneously. Revenue growth slowed to 15% in Q1 — modest for a company whose prospectus claims to have identified "the largest TAM in human history." The filing leans heavily on Starship ambitions and AI infrastructure bets, but that 15% growth rate is the number investors will circle.

The Anthropic Deal's Fine Print

Buried in the disclosure is a compute agreement with Anthropic worth up to $40 billion — a number that immediately became its own headline. The "catch" reported by The Information suggests the commitment is contingent rather than firm, which matters enormously for how investors model forward revenue. Anthropic, for its part, is projecting its first profitable quarter with revenue more than doubling to roughly $10.9 billion in Q2 — a trajectory that at least makes the compute commitment plausible. Anthropic is also expanding to Colossus2 on GB200 infrastructure, signaling that its hardware appetite isn't slowing down.

Jensen Huang's Next $200 Billion Claim

Nvidia CEO Jensen Huang declared that CPUs purpose-built for AI agents represent a brand-new $200 billion market — a pivot from GPU training dominance toward full-stack compute. The logic is straightforward: if agents need persistent, low-latency processing rather than burst training runs, the bottleneck shifts away from Nvidia's existing stronghold. Huang naming this publicly almost certainly means Nvidia is already deep in development, but it also signals the company knows it can't coast on H100 momentum forever.

OpenAI Solves Something — With Receipts

OpenAI claims its reasoning model disproved a discrete geometry conjecture that had stood since 1946. What separates this from prior AI math theater: the same mathematicians who publicly torpedoed OpenAI's last embarrassing false breakthrough are backing this one. Falsifying a conjecture is also logically cleaner than proving one — a single verifiable counterexample is all it takes — which may be exactly why this result survived scrutiny where others haven't.

Incumbents Pay the Bill

While AI infrastructure companies post record forward projections, Intuit is cutting 17% of its workforce as revenue growth hits its lowest point since 2024. Intuit sells financial software to the same small businesses that AI tools now serve directly. That's not coincidence — it's the disruption mechanism running through the income statement of a company that moved too slowly.

The throughline is that AI infrastructure is solidifying its capital structure — Anthropic's compute commitments, Nvidia's next product pitch, SpaceX's revenue ambitions all pointing the same direction — while the disruption it enables is already showing up in someone else's layoff announcement.