Anthropic has agreed to pay SpaceX nearly $45 billion over the next three years as part of a major computing infrastructure agreement designed to support the rapid expansion of its Claude artificial intelligence platform. The deal was disclosed in regulatory filings tied to SpaceX’s initial public offering and highlights the escalating race among AI companies to secure large-scale compute capacity.
Under the agreement, Anthropic is expected to pay approximately $1.25 billion per month through May 2029 for access to computing infrastructure across SpaceX’s Colossus and Colossus II AI data centers in Memphis, Tennessee. The filings noted that computing capacity will ramp during May and June 2026 at reduced pricing before reaching full scale. Either company may terminate the agreement with 90 days notice.
Anthropic had previously announced access to more than 300 megawatts of compute power from SpaceX’s Colossus 1 facility but did not disclose financial details at the time. The partnership has since expanded to include Colossus II, according to comments from Anthropic co-founder and Chief Compute Officer Tom Brown.
The agreement reflects the growing scarcity and strategic importance of AI infrastructure as demand surges for large language models and enterprise AI tools. Anthropic has experienced rapid adoption of Claude for coding, enterprise workflows, and business productivity applications, increasing its need for massive GPU clusters and power-intensive data center capacity.
For SpaceX, the deal provides a substantial recurring revenue stream tied to its expanding AI infrastructure business following the integration of Elon Musk’s xAI operations. Regulatory filings showed SpaceX’s AI division generated significant losses despite heavy investment in data centers and compute infrastructure, making large external compute agreements strategically important for the company.
