Anthropic turns its first profit, OpenAI files for IPO, and SpaceX exposes a $45B compute deal
Three things happened at once. Anthropic told investors it expects $10.9 billion in Q2 revenue and its first-ever quarterly operating profit of $559 million. OpenAI is preparing to file a confidential S-1 with the SEC as early as today, targeting a September public listing at an $852 billion to $1 trillion valuation. And SpaceX’s S-1 prospectus dropped with a buried detail: Anthropic is paying SpaceX $1.25 billion per month through May 2029 for GPU compute at the Colossus data centers, a deal worth roughly $45 billion total.
All three stories are connected. The IPO filings and profitability disclosures are not coincidental events. They are a coordinated sequence designed to anchor public market valuations before the math gets harder.
Anthropic’s profit milestone
The numbers are blunt. Q1 revenue was $4.8 billion. Q2 is projected at $10.9 billion. That is a 130% jump in one quarter. The company went from a $9 billion annualized run rate at the end of 2025 to a $43.6 billion run rate if Q2 holds. Operating income of $559 million would be Anthropic’s first black quarter ever.
The drivers are specific. Claude Code and its agentic coding tools have become the default for enterprise software teams, generating recurring API revenue at high margins. Compute efficiency improved: Anthropic spent 71 cents on compute for every dollar of revenue in Q1, projected to drop to 56 cents in Q2. Enterprise customers spending $1 million or more annually doubled from 500 to over 1,000 between February and April.
CEO Dario Amodei recently called revenue growth “too hard to handle” at a developer conference. He was not being modest. The constraint is supply-side. Which brings us to the SpaceX deal.
The $45 billion compute contract
SpaceX’s S-1 filing, first reported by WIRED and confirmed by Reuters and Bloomberg, discloses that Anthropic has agreed to pay SpaceX $1.25 billion per month through May 2029 for access to both Colossus and Colossus II data centers. That is $15 billion per year. Over three years, the total exceeds $45 billion.
Anthropic confirmed the figures. Capacity is ramping in May and June 2026 at a reduced fee before the full monthly rate kicks in. The deal gives Anthropic access to over 300 megawatts of GPU capacity.
The irony writes itself. Anthropic, founded on principles of AI safety, is now sending $15 billion a year to Elon Musk’s company to keep training frontier models. The Colossus clusters were built for xAI’s Grok models. Now Anthropic is the anchor tenant.
SpaceX’s S-1 also reveals broader financials: $1.75 trillion target valuation, 10.3 million Starlink subscribers, and a $4.3 billion Q1 loss. The AI infrastructure revenue from Anthropic is a meaningful line item for a company otherwise losing money.
OpenAI’s IPO move
OpenAI is working with Goldman Sachs and Morgan Stanley on a confidential IPO filing, according to CNBC, Reuters, Axios, and Bloomberg all reporting the same timeline. The filing could come as early as Friday, May 22. A confidential S-1 lets OpenAI begin the SEC review process without public financial disclosure. The typical timeline is two months for SEC comments, then a public S-1, then a roadshow, then pricing. That puts a September listing on schedule.
The target valuation ranges from $852 billion to $1 trillion depending on the source. OpenAI has not disclosed its own profitability timeline.
Why this week matters
The sequencing is the story. Anthropic’s profitability announcement gives the AI sector its first proof point that frontier model companies can make money while still training. That proof point is designed to support OpenAI’s IPO pricing. And the SpaceX compute disclosure, buried in an S-1 that most people would read for Starlink numbers, reveals just how much capital is flowing through this ecosystem just to keep the models running.
Ed Zitron published a skeptical piece arguing that Anthropic’s profitability claim is a “swindle” because the $559 million operating profit is dwarfed by capital expenditure commitments like the SpaceX deal. CNBC published its own warning that cheap AI from open-source models could undermine the $800B+ valuations both companies are targeting. Chinese AI models already account for 60% of all usage on OpenRouter, up from 1% in 2024.
The IPO window is open right now. SpaceX is targeting a June listing. OpenAI in September. Anthropic reportedly in October. The race to go public before the next market rotation or the next cheap model that erodes margins is on.
Sources: CNBC, Reuters, WSJ, WIRED, Bloomberg, Axios, Business Insider