The artificial intelligence start-up Anthropic has struck a landmark computing partnership with Elon Musk’s SpaceX, securing exclusive access to the full capacity of the Colossus 1 data centre in Memphis, Tennessee — a facility originally built to power Musk’s own rival AI venture, xAI. The deal, announced Wednesday at Anthropic’s annual developer conference in San Francisco, grants the maker of the Claude assistant more than 300 megawatts of compute capacity backed by over 220,000 Nvidia chips, a move that underscores the escalating infrastructure arms race in generative AI.

The agreement is striking not only for its scale but for the unlikely bedfellows it brings together. In February, Musk publicly excoriated Anthropic, describing its models as “misanthropic and evil” and accusing them of hating “whites and Asians, heterosexuals and men”. Viewed from Silicon Valley, the partnership appears purely transactional: Anthropic needs vast computing power to relieve bottlenecks in its popular coding tool Claude Code, while SpaceX — which recently absorbed xAI’s operations — is commercialising its data centre assets to generate revenue. Analysts in London note that the deal also gives Anthropic a strategic hedge against reliance on cloud providers such as Amazon and Google.

Under the terms, Anthropic will immediately double rate limits for subscribers of Claude Pro, Claude Max, and Team plans, and raise limits for enterprise developers. Chief product officer Ami Vora told the conference that the additional capacity will be directed straight at relieving the constraints that have frustrated users as demand surges. For competitors including OpenAI and Google DeepMind, the development signals that Anthropic is serious about scaling its infrastructure to match the computing heft of its bigger rivals.

Yet the alliance may extend far beyond terrestrial computing. Anthropic has expressed interest, according to a company statement, in collaborating with SpaceX to develop “multiple gigawatts” of AI compute capacity in orbit. Viewed from Washington, such an ambition raises questions about concentration of critical AI infrastructure in the hands of a few billionaires, even as it plays directly into Musk’s long-term vision of space-based data processing. While orbital computing remains speculative, the partnership gives Anthropic a foothold in Musk’s broader ecosystem, from Starlink to Starship.

For Musk, the deal offers a pragmatic way to monetise hardware originally built for his own AI ambitions while maintaining a presence in a sector he has publicly derided. For Anthropic, it provides an immediate solution to its capacity crunch and a route to the next frontier. As demand for generative AI shows no sign of abating, such improbable alliances may become the new normal for an industry where compute is the ultimate currency.