Anthropic Reportedly Paying SpaceX $15 Billion Annually for Compute Power Anthropic’s race for more computing power has collided with SpaceX’s push to impress public-market investors, producing a colossal AI–infrastructure pact whose future remains deliberately flexible.

In early May, Anthropic quietly signed a deal to buy compute capacity from SpaceX’s Colossus data center in Tennessee to support its “booming AI business,” agreeing to pay $1.25 billion per month through May 2029. SpaceX’s S-1 IPO filing, released on May 20, revealed that the agreement covers both the original Colossus 1 and the newer Colossus 2 facilities and confirmed that Anthropic is getting discounted rates in May and June as usage ramps up.

With the filing, the scale of the arrangement came into focus: Anthropic’s payments total $15 billion annually, nearly doubling SpaceX’s reported 2025 revenue of $18.7 billion. Over the full term, SpaceX could earn more than $40 billion, if the agreement holds. The company framed the structure as a way to “monetize unused compute capacity” while retaining the option to redirect it for internal AI work.

On the AI side, Anthropic has been constrained by a lack of compute even as its revenue “is taking off,” pushing it to secure large-scale GPU access wherever possible. The Verge described the deal as a sign of how “desperate AI companies like Anthropic are for compute capacity,” especially as new data centers face local resistance across the US.

SpaceX, meanwhile, has poured money into AI since merging with Elon Musk’s xAI, spending $12.7 billion on AI-related capital expenditure in 2025 and $7.7 billion in just the first quarter of 2026, far more than on its space division. That buildout has produced heavy losses — $6.3 billion on AI operations in 2025 and $2.5 billion more in Q1 2026 — which the Anthropic contract could help offset.

The deal also plays into SpaceX’s growth narrative ahead of its IPO. One analysis noted that Grok, the AI system tied to Musk’s X platform, has seen downloads drop sharply and a key federal contract stall, potentially “undercutting SpaceX’s IPO growth story.” Against that backdrop, recurring, multi‑billion‑dollar AI infrastructure revenue from Anthropic bolsters the case that SpaceX’s compute business can stand alongside its launch and satellite operations.

Yet both sides have deliberately left themselves an exit. The contract includes a clause allowing either Anthropic or SpaceX to terminate with 90 days’ notice, a provision observers say reflects the “fast-moving nature of the AI industry.” If AI architectures, chip economics, or regulatory constraints shift, either party can pivot.

Publicly, Musk has highlighted the human effort behind SpaceX’s expanding ambitions. After the IPO filing drew praise for SpaceX’s achievements, he wrote, “The SpaceX team is incredible!” In another post, he amplified a story of an engineer working deep into the night, described as “happy” and “energized” because he was “directly influencing the course of a multi-planetary future for humanity.”

Taken together, the Anthropic–SpaceX agreement illustrates how the scramble for AI compute is reshaping tech alliances: an AI lab effectively underwriting a rocket company’s AI pivot, even as both keep one hand on the eject handle.

Continue reading https://foxvector.com