Anthropic secures access to 220,000 NVIDIA GPUs via SpaceX Colossus 1 for its models

Anthropic said on May 6 that it had signed an agreement with SpaceX to use all of the compute capacity at the Colossus 1 data center, a deal the company said would give it access to more than 300 megawatts of capacity and more than 220,000 NVIDIA GPUs within a month. The company tied the announcement directly to product changes, saying it was doubling Claude Code’s five-hour rate limits for several paid plans, removing peak-hours limit reductions for some users, and raising API rate limits for Claude Opus models. The arrangement puts Anthropic’s newest infrastructure expansion in a facility tied to Elon Musk, whose xAI competes with Anthropic in the market for large language models. CNBC reported that Colossus 1 is in Memphis, Tennessee, and that Anthropic said the deal would directly improve capacity for Claude Pro and Claude Max subscribers. ### How much compute did Anthropic actually say it secured? (anthropic.com) Anthropic’s own statement put the number at “more than 300 megawatts” and “over 220,000 NVIDIA GPUs,” with the capacity expected within a month. The company said it had agreed to use all of the compute capacity at SpaceX’s Colossus 1 data center. Data Center Dynamics, citing the same announcement, reported that Anthropic planned to use all of the compute capacity at xAI’s Colossus 1 site and said the cluster would directly support Anthropic’s paid Claude offerings. (cnbc.com) ### Why did Anthropic announce user-facing changes at the same time? Anthropic said the compute deal, together with other recent infrastructure agreements, let it raise usage limits immediately. (anthropic.com) The company said the changes effective May 6 included doubling Claude Code’s five-hour rate limits for Pro, Max, Team and seat-based Enterprise plans, ending peak-hours reductions for Pro and Max accounts, and increasing API rate limits for Claude Opus models. (datacenterdynamics.com) The company framed those changes as aimed at “our most dedicated customers,” and said the added SpaceX capacity would directly improve service for Claude Pro and Claude Max subscribers. ### Where does this fit in Anthropic’s broader compute buildout? Anthropic said the SpaceX agreement joins several other recent compute commitments. In the same May 6 post, the company cited an up to 5 gigawatt agreement with Amazon, including nearly 1 gigawatt of new capacity by the end of 2026; a 5 gigawatt agreement with Google and Broadcom beginning to come online in 2027; a Microsoft and NVIDIA partnership that includes $30 billion of Azure capacity; and a $50 billion investment in U.S. (anthropic.com) AI infrastructure with Fluidstack. Amazon said on April 20 that its expanded Anthropic agreement secures up to 5 gigawatts of new capacity for training and deploying Claude, with significant Trainium2 capacity coming online in the first half of 2026 and nearly 1 gigawatt of Trainium2 and Trainium3 capacity online by the end of 2026. Amazon also said Anthropic would commit more than $100 billion over 10 years to AWS technologies, while Amazon would invest $5 billion immediately and up to an additional $20 billion later. (anthropic.com) Anthropic said in November 2025 that its Microsoft and NVIDIA partnership included a commitment to purchase $30 billion of Azure compute capacity and to contract additional compute capacity up to one gigawatt. Anthropic also said at the time that Amazon remained its primary cloud provider and training partner. ### Why is SpaceX involved when xAI is a rival? (anthropic.com) CNBC reported that Musk merged SpaceX with xAI earlier in 2026 and that the Colossus 1 agreement therefore links Anthropic to infrastructure owned by a competing AI group. The report also said Musk had previously criticized Anthropic in public posts before saying on May 6 that he was “impressed” after meeting senior members of the team. (anthropic.com) Data Center Dynamics reported that Musk said SpaceX was willing to lease Colossus 1 after training had already moved to Colossus 2. That report also said Colossus 1 includes H100, H200 and GB200 accelerators. ### What comes next after the Colossus 1 deal? Anthropic said the Colossus 1 capacity would come online within a month of the May 6 announcement. (cnbc.com) The company also said it had “expressed interest” in working with SpaceX to develop multiple gigawatts of orbital AI compute capacity. Anthropic’s next disclosed infrastructure milestones sit on a longer timeline. Amazon said nearly 1 gigawatt of Trainium2 and Trainium3 capacity is due by the end of 2026, while Anthropic said its Google and Broadcom compute agreement is set to begin coming online in 2027. (datacenterdynamics.com) (anthropic.com 1) (anthropic.com 2)