SpaceX filed for what could be the largest IPO in history this week, targeting a self-assigned valuation of $1.25 trillion and a potential raise of $50 to $75 billion. The filing also marks the first time the company has opened its books. Here are ten things worth flagging.
1. The full-year loss was $4.9 billion, against a $791 million profit in 2024. Revenue grew 33% to $18.7 billion, but capex nearly doubled from $11.2 billion to $20.7 billion, driven almost entirely by AI buildout.
2. AI is now 17% of revenue and burning more cash than the rest of the business combined. AI capex alone hit $12.7 billion in 2025, more than Space and Connectivity put together. Q1 2026 AI capex was $7.7 billion in a single quarter.
3. The Anthropic contract is bigger than the entire AI segment. SpaceX signed Cloud Services Agreements with Anthropic in May 2026 for $1.25 billion per month through May 2029. Annualized, that single contract is roughly $15 billion against an AI segment that did $3.2 billion in all of 2025.
4. COLOSSUS II came online in 91 days. Industry benchmark for a 100 MW greenfield data center is roughly two years. COLOSSUS II is currently training Grok 5, and the next expansion phase adds at least 220,000 GB300 processors and over 400 megawatts.
5. The orbital AI plan targets 100 gigawatts of compute launched to orbit per year, starting as early as 2028. AI compute satellites are framed as an evolution of the Starlink bus, with solar power in sun-synchronous orbit removing the terrestrial energy bottleneck that the S-1 identifies as the binding constraint on AI growth.
6. Starship reusability is not required for the orbital AI thesis to start. V3 broadband satellites, V2 Mobile satellites, and initial AI compute satellites can deploy without full reusability, giving the orbital pitch a partial path forward even if Starship slips against its current timeline.
7. Musk's January 2026 performance grant vests against market cap milestones up to $7.5 trillion. Six times the company's own self-assigned valuation at filing, and a number that only makes sense if you are underwriting a multi-trillion-dollar outcome for SPCX itself.
8. Class B shares elect 51% of the board, forever. Ten votes per share, and the board election right holds as long as any Class B remains outstanding. Public shareholders have limited governance recourse on anything.
9. Cursor is an option and not an acquisition. SpaceX has the right but not the obligation to acquire Cursor at a $60 billion implied equity value, payable in Class A common stock. If SpaceX walks, Cursor receives a $1.5 billion termination fee plus an $8.5 billion deferred services fee.
10. Terafab is a framework. The chip joint venture with Tesla and Intel targets one terawatt of annual compute production, but the S-1 is explicit that no specific projects, timelines, or capital commitments are locked in. SpaceX expects to continue sourcing significant compute hardware from third parties.
If the offering prices anywhere near the company's self-assigned $1.25 trillion valuation, it would clear the path for the next wave of large tech listings, with OpenAI reportedly preparing to file confidentially in the coming weeks and Anthropic also expected to test the market. Cerebras opened the window last week with a 68% first-day pop, the largest tech IPO since 2019. SPCX is the test of how much further that window can open.
