Anthropic has agreed to pay SpaceX $1.25 billion per month through May 2029 for access to the rocket company's Colossus data centers in Memphis, Tennessee. The staggering deal, revealed in SpaceX's IPO filing, amounts to $15 billion annually, nearly double the $18.7 billion in revenue SpaceX reported for all of 2025.
Why This Matters
This agreement underscores the insatiable demand for computing power needed to train advanced AI models. For Anthropic, securing massive compute capacity is essential to keep pace with rivals like OpenAI and Google. The deal also transforms SpaceX into a major AI infrastructure provider, turning its data centers into a lucrative revenue stream that rivals its core launch business.
Investors and industry watchers now have a rare window into the costs driving the AI arms race. The scale of this commitment shows that leading AI companies are willing to pay extraordinary sums to lock in the hardware they need.
Details of the Agreement
The S-1 filing reveals that Anthropic will pay SpaceX $1.25 billion each month until May 2029. The agreement covers access to both Colossus I and Colossus II training centers. It includes standard provisions that allow either party to adjust or exit the deal under certain conditions, though the specifics remain confidential.
The total commitment of $15 billion per year highlights the enormous scale of Anthropic's ambition. The company is betting that the long-term contract will guarantee the computational resources required for next-generation AI development.
Impact on the AI Industry
The partnership signals a broader shift in the AI sector. Companies are racing to secure long-term compute contracts to avoid capacity shortages that could slow development. For SpaceX, the deal diversifies its revenue beyond satellite launches and Starlink, positioning the company as a key player in the AI infrastructure market.
Other AI firms may now face pressure to strike similar deals, potentially driving up costs across the industry. The IPO filing provides unprecedented transparency into how much top AI companies are willing to spend on raw computing power, and it suggests that the era of cheap, abundant AI training capacity may be ending.
