The deal hands Anthropic access to more than 220,000 Nvidia GPUs, including H100, H200 and next-generation GB200 accelerators, housed at the Colossus 1 facility, which SpaceX describes as one of the world’s largest AI supercomputers. The company said the capacity would come online within the month.

Anthropic moved immediately to pass the benefit to customers. The company said it was doubling Claude Code’s rate limits for paid plans, removing peak-hour usage caps for Pro and Max subscribers, and sharply increasing the volume of requests developers can make to its Claude Opus models via the API.

The announcement came alongside Anthropic’s developer day, at which the company also unveiled a feature called “dreaming”, a tool that allows AI systems to review work between sessions, identify patterns and update files storing user preferences and context.

The compute deal is the latest in a rapid series of infrastructure agreements as Anthropic scrambles to keep pace with demand it has publicly acknowledged is straining its systems.

The company has also secured an agreement with Amazon for up to 5 gigawatts of capacity, including nearly 1GW coming online by the end of 2026; a 5GW deal with Google and Broadcom that begins in 2027; a Microsoft and Nvidia partnership covering $30 billion of Azure capacity; and a $50 billion US infrastructure commitment with Fluidstack.

The partnership also carries an orbital dimension. Anthropic has expressed interest in working with SpaceX to develop multiple gigawatts of space-based compute capacity, a strategic priority for SpaceX and a central component of its pitch ahead of a planned IPO expected in the coming weeks. Adding Anthropic as a named infrastructure customer strengthens SpaceX’s case as an AI compute provider, with the deal establishing a disclosed revenue line beyond its launch and Starlink businesses.

The agreement is notable given the history between Elon Musk and Anthropic. Musk has previously accused the company of hating “Western Civilisation” and predicted it was “doomed to become the opposite of its name.” His tone shifted following what he described as a week of meetings with the Anthropic leadership team.

Musk wrote on X that “no one set off my evil detector,” while adding that SpaceX reserved the right to reclaim capacity if Anthropic’s AI “engages in actions that harm humanity.” He also noted that SpaceXAI, the entity formed from SpaceX’s merger with his own AI startup xAI, had already migrated its training workloads to Colossus 2, freeing the Memphis facility for Anthropic’s use.

The deal arrives as Anthropic navigates a difficult relationship with the US government. In March, the Pentagon declared the company a supply chain risk and blacklisted it from military contracts after negotiations over permitted model uses broke down.

Anthropic has filed suit against the Trump administration in two jurisdictions to overturn the decision, and that litigation remains ongoing. Separately, the company is reported to be in discussions with investors about a funding round that would value it at $900 billion.

RELATED STORIES

Why Amazon is investing $25bn into Anthropic in a major AI deal

Anthropic to shoulder price hikes from its US data centres