SpaceX and Anthropic AI have announced a major compute deal granting Anthropic full access to the Colossus 1 data center in Memphis, Tennessee, a supercomputer built by xAI and now operated by SpaceX. The announcement was made during Anthropic’s second annual “Code with Claude” developer event on Wednesday.
The agreement gives Anthropic access to over 220,000 Nvidia GPUs and more than 300 megawatts of compute capacity, immediately boosting its AI infrastructure to meet surging demand for its Claude models, especially Claude Code and Claude Opus.
In a statement, Anthropic said the deal will allow it to double rate limits for paid users of Claude Code on Pro, Max, Team, and Enterprise plans, remove peak-hour usage reductions, and significantly increase API rate limits for Claude Opus. The company also revealed it has “expressed interest in partnering with SpaceX to develop multiple gigawatts of orbital AI compute capacity,” signaling long-term ambitions for space-based data centers.
Anthropic emphasized this partnership supplements existing agreements with Amazon, Google, Microsoft, and Nvidia, with those larger-scale commitments expected to come online in late 2026 or early 2027.
SpaceX confirmed that Colossus 1—featuring H100, H200, and GB200 Nvidia accelerators—was built in just 122 days and is one of the world’s largest AI supercomputers. The company noted that xAI has transitioned its training operations to Colossus 2, freeing up Colossus 1 for external partners.
The space and AI company framed the deal as part of its broader strategy to monetize underused compute capacity and position itself as a leading AI infrastructure provider, with potential future developments in orbital computing to address Earth’s power and cooling constraints.
SpaceX CEO Elon Musk endorsed the deal on X, writing: “Same here. By way of background for those who care, I spent a lot of time last week with senior members of the Anthropic team to understand what they do to ensure Claude is good for humanity and was impressed. Everyone I met was highly competent and cared a great deal about doing the right thing. No one set off my evil detector. So long as they engage in critical self-examination, Claude will probably be good.”
He added: “After that, I was ok leasing Colossus 1 to Anthropic, as SpaceXAI had already moved training to Colossus 2,” and reserved the right to reclaim compute if Anthropic’s AI ever harms humanity. Musk also noted SpaceX will offer compute to other AI firms under “fair terms and pricing,” similar to its satellite launch model.
Musk had previously criticized Anthropic, calling it “misanthropic” and its AI “evil,” and accused it of bias against Western civilization. This shift follows xAI’s merger with SpaceX earlier in 2026, forming “SpaceXAI,” which is reportedly preparing for an IPO.
The deal comes amid a legal battle between Anthropic and the Trump administration’s Department of Defense, which seeks unrestricted military use of Claude—something Anthropic has resisted on ethical grounds.
