In a surprising twist for the AI arms race, Elon Musk’s SpaceX has opened up major data-center capacity to Anthropic, the maker of Claude. The deal gives Anthropic access to roughly 300 megawatts of compute at SpaceX’s Colossus 1 facility in Memphis, Tennessee — a site originally built to support xAI’s Grok — and is expected to come online within about a month.
What’s inside matters: Colossus 1 reportedly houses more than 220,000 Nvidia processors, a scale that can meaningfully boost Anthropic’s ability to train and serve large models. The extra capacity arrives as demand rises for Anthropic products like Claude Code, its AI coding assistant. Anthropic says the new compute allocation will let it double rate limits for paid Claude Code users and lift some peak-hour caps on subscription plans.
The partnership is striking because top-tier AI players seldom rent infrastructure to direct competitors. Musk has publicly criticized Anthropic in the past — even accusing the startup of steering away from its founding goals and quipping that it risked becoming “misanthropic.” That rhetoric softened after recent meetings between Musk and Anthropic executives. Musk said he approved the arrangement after being “impressed” by Anthropic’s efforts to keep Claude “good for humanity,” adding that no one at Anthropic had set off his “evil detector.” He cautioned he would cut off any client who used xAI facilities to build systems that harmed people.
SpaceX and xAI may be signaling a broader strategy shift. Reports indicate xAI has moved much of its internal training to a newer Colossus 2 cluster, freeing capacity in Colossus 1 to lease out. SpaceX has already signed a similar compute deal with AI coding startup Cursor, and Anthropic has reportedly discussed potential future collaborations on orbital data centers that could one day deliver several gigawatts of compute.
For investors and the market, the arrangement is notable: it points to SpaceX expanding into cloud and AI infrastructure as a potential revenue stream ahead of any public offering. If Musk’s companies can commercialize their huge AI-capex investments, it changes the economics of their AI ambitions.
Anthropic unveiled the deal at its first developer conference in San Francisco and also introduced a new “dreaming” feature for Claude. The feature helps the model learn between sessions by reviewing user workflows and updating stored context automatically — a move aimed at smoother, more context-aware assistance.
Why crypto readers should care: massive, centralized compute platforms like Colossus reshuffle where AI workloads live and who controls them — an important consideration for Web3 projects and decentralized infrastructure builders exploring AI integration or on-chain/off-chain hybrid models. For now, the headline is simple: one of the world’s most talked-about tech leaders just opened a large slice of GPU firepower to a previously skeptical rival, and the AI infrastructure market may never look the same.
Read more AI-generated news on: undefined/news
