SpaceX IPO Filing Reveals Anthropic Pays $15B a Year for AI Compute

SpaceX's long‑awaited IPO filing landed this week — and inside the S‑1 was a number that redefines how expensive AI infrastructure has become. Anthropic, the maker of Claude, has agreed to pay SpaceX $1.25 billion per month for AI compute capacity through May 2029. That's $15 billion per year, or roughly $45 billion over the life of the contract.

The deal gives Anthropic access to more than 220,000 NVIDIA GPUs across both of SpaceX's AI training clusters — Colossus and Colossus II — representing over 300 megawatts of capacity, according to Yahoo Finance. Either party can exit with 90 days' notice, per the Reuters report on the filing.

Anthropic expanded the deal just as the filing went public, adding Colossus II to what was originally just Colossus I. “We're expanding our partnership with SpaceX, and will be scaling up on GB200 capacity in Colossus 2 throughout June,” Anthropic co‑founder and chief compute officer Tom Brown posted on X, as reported by Axios.

The compute deal is not a luxury purchase — it's survival math. Anthropic CEO Dario Amodei admitted the company's growth “far exceeded internal expectations,” creating infrastructure strain that required rival‑level spending to resolve, per Yahoo Finance. Claude's paid subscribers were hitting five‑hour rate limits, and code generation on paid plans was capped — both restrictions lifted once the SpaceX capacity came online.

The competitive irony is sharp: SpaceX's data centers were originally built for xAI, Elon Musk's own AI company and a direct Anthropic competitor. Anthropic is renting compute from infrastructure built for its rival. Musk, for his part, posted on X that SpaceX is in discussions with other companies about “offering AI compute as a service at significant scale,” according to Reuters.

SpaceX stated in the filing it has “sufficient capacity to provide compute for our own AI models, including support of our training and inference demands, and to satisfy the obligations under these agreements,” and expects to sign additional similar contracts, per.²

SpaceX's IPO documents reveal a company that is dominant in rockets but bleeding money in AI. For Q1 2026, SpaceX posted $4.69 billion in revenue against a $4.28 billion GAAP net loss, according to Reuters. Full‑year 2025 showed $18.7 billion in revenue and a $4.94 billion loss.

The AI segment is the biggest drag: it lost $2.5 billion from operations in Q1 on just $818 million in segment revenue. The Anthropic deal represents a critical revenue stream — the $1.25 billion monthly payments are larger than the entire AI segment's quarterly revenue. SpaceX is seeking to raise $75 billion at a valuation between $1.5 trillion and $2 trillion, which would make it the largest IPO in history, Axios reported.

While SpaceX burns cash, Anthropic is about to turn the corner. The company is projecting Q2 2026 revenue of at least $10.9 billion — more than double its Q1 figure of $4.8 billion — and its first‑ever operating profit of $559 million, according to Reuters. The surge is driven by enterprise adoption of Claude for coding and the Mythos security model.

Anthropic's annualized revenue now sits near $45 billion — roughly 35% higher than OpenAI's estimated $33 billion ARR, according to reporting from Sherwood News. The revenue has quintupled from $9 billion at the end of 2025. The first quarterly profit is a milestone that few AI labs have reached — OpenAI, by contrast, projects $14 billion in losses for 2026.

The IPO filing has reignited speculation about a SpaceX‑Tesla merger. Musk reportedly discussed combining the companies with colleagues, and an early SpaceX investor told CNBC that a post‑IPO tie‑up is “only a matter of when.” The filing revealed SpaceX bought $697 million worth of Tesla Megapack battery systems in 2024‑2025 to power its data centers, showing existing operational overlap.

Musk will retain 85% voting control of SpaceX after the IPO, per The LA Times, giving him the power to drive consolidation regardless of minority shareholder preferences. A merged SpaceX‑Tesla would create an entity spanning rockets, EVs, AI compute, social media (X), and solar energy — the “everything app” Musk has teased for years.

The Anthropic-SpaceX deal is a signal about where AI infrastructure is heading. Nvidia recently projected that AI infrastructure spending could approach $4 trillion annually, and the SpaceX IPO documents confirm that trajectory, according to HPC Wire. OpenAI expects to spend roughly $115 billion over the next four years.

For builders, the implications are straightforward: AI compute is becoming the single largest cost line item for frontier labs, and that cost will flow downstream. When Claude's maker pays $15 billion a year for GPUs, those costs eventually surface in API pricing, rate limits, and subscription tiers. The companies that own infrastructure — SpaceX, Google, Microsoft, Amazon — are positioning themselves as the landlords of the AI economy.

The 90‑day termination clause adds another layer: either side can walk away on short notice. For Anthropic, that means the compute that powers Claude's availability and rate limits could evaporate if the relationship sours. For SpaceX, it means the $1.25 billion monthly revenue stream is renewable, not guaranteed.