SpaceX's Wednesday IPO filing with US regulators disclosed that Anthropic is paying $15 billion annually to access SpaceX data center capacity under a GPU lending agreement, establishing the largest known compute commitment between competing AI infrastructure providers. The arrangement, detailed in the S-1 documentation, positions SpaceX as a material compute lessor in the AI infrastructure market while Anthropic secures capacity outside its primary Amazon Web Services commitment.

The SpaceX agreement runs concurrently with Anthropic's $33 billion Amazon commitment and a newly disclosed $15 billion annual xAI arrangement ($1.25 billion monthly), bringing Anthropic's total disclosed annual compute obligations to approximately $63 billion. The SpaceX terms include fixed capacity guarantees with quarterly true-up provisions and variable GPU allocation rights based on SpaceX data center expansion timelines through 2027. SpaceX's filing noted the Anthropic revenue as a material line item under "Commercial Services — Advanced Computing," representing roughly 18 percent of SpaceX's projected non-launch revenue for the current fiscal year.

The arrangement matters because it reveals Anthropic's willingness to contract with direct AI competitors for compute access, a shift from the prior industry norm of exclusive cloud provider relationships. SpaceX operates an estimated 400,000 H100-equivalent GPUs across Texas and Nevada facilities, with planned expansion to 850,000 units by Q4 2025 according to the filing. Anthropic gains access to 120,000 reserved GPU slots with priority queuing rights, suggesting the company is securing capacity ahead of anticipated Claude model scaling requirements that its AWS Reserved Capacity agreements may not fully cover. The deal also demonstrates SpaceX's infrastructure monetization strategy ahead of its public offering, providing a revenue stream that offsets Starlink capital expenditure while establishing compute-as-a-service as a material business line.

For family offices and allocators, the arrangement signals that frontier AI labs are moving toward multi-vendor compute strategies to avoid single-provider dependency, even when it means transacting with competitors. The $15 billion annual commitment represents roughly 40 percent of Anthropic's estimated $37 billion post-Series D valuation, an unusually high revenue-to-valuation ratio that underscores the capital intensity of maintaining frontier model development. The SpaceX IPO filing also disclosed that Anthropic's contract includes a clause allowing capacity transfer to a yet-unnamed third party if Anthropic's liquidity falls below $8 billion, suggesting the company negotiated flexibility in case of funding constraints or strategic pivots.

Allocators should monitor SpaceX's April roadshow presentations for updated data center utilization metrics and whether additional AI customers are disclosed beyond Anthropic. Anthropic's next funding round, expected in Q3 2025, will clarify whether the company can sustain $63 billion in annual compute commitments without margin compression. The xAI agreement's monthly payment structure ($1.25 billion) suggests Anthropic may be managing liquidity through staggered obligations rather than annual prepayment, a detail worth tracking in any future term sheets.

SpaceX's S-1 lists Anthropic as a "significant customer" under Item 1.03, which triggers enhanced disclosure requirements if the contract is modified or terminated before the IPO closes in late May.