Anthropic has inked a major agreement with Elon Musk’s SpaceX to take over the entire computing capacity of the Colossus 1 data center in Memphis, Tennessee, in a move that instantly reshapes the Artificial Intelligence infrastructure race.
The agreement, revealed by Anthropic Chief Product Officer Ami Vora at the company’s developer conference in San Francisco, on Wednesday, May 6, comes as global demand for AI compute strains existing infrastructure and forces rivals to strike unconventional alliances.
According to officials, the deal gives Anthropic exclusive access to more than 220,000 Nvidia GPUs and 300 megawatts of new capacity by the end of the month, addressing urgent compute shortages as demand for its Claude models and Claude Code platform surges.
The partnership turns two AI rivals into uneasy collaborators. Colossus 1 was originally built by Musk’s xAI to train Grok models before xAI was acquired and folded into SpaceX in February. With SpaceX now shifting its own frontier training to the larger Colossus 2 facility nearby, the company is leasing out Colossus 1 rather than using it internally.
Musk said he became comfortable with the deal after meeting Anthropic’s team, noting “no one set off my evil detector.” The agreement also includes Anthropic’s stated interest in partnering with SpaceX to develop multiple gigawatts of orbital AI compute capacity — solar-powered data centers in space.
For Anthropic, the Colossus deal is a lifeline. The company has struggled to secure enough GPUs to keep pace with explosive growth of Claude Code and API usage.
Executives said the added capacity will “substantially increase” limits for developers and enterprise customers. The timing is critical: Anthropic is also building its own $50 billion data center network across Texas and New York with Fluidstack, but those sites won’t come online until later in 2026. Colossus 1 plugs the gap immediately.
The agreement also signals a shift in SpaceX’s AI strategy. As the company prepares for a potential public listing, it is monetizing existing infrastructure while pushing toward Musk’s long-term vision of space-based compute. Colossus 1 sits in a former Electrolux factory in Memphis, chosen for its reliable power grid, water access, and proximity to the Mississippi River. With a third building called MACROHARDRR now acquired and Colossus 2 under construction, Musk says total training compute for SpaceX will soon approach 2 gigawatts — among the most energy-intensive AI projects in the world.
The deal also carries economic externalities. Colossus 1 and the nearby Colossus 2 facility in Memphis have drawn complaints from local residents over pollution from gas turbines powering the sites, Business Insider reported. Anthropic acknowledged the energy footprint in its announcement, noting a recent commitment to cover any consumer electricity price increases caused by its US data centers and saying it is exploring ways to extend that pledge internationally.
Buried at the bottom of Anthropic’s press release was a line that may matter more in five years than it does today. The company has “expressed interest” in partnering with SpaceX to develop “multiple gigawatts of orbital AI compute capacity.” No timeline, no dollar figure, no technical detail. But the sentence signals that both companies are at least exploring whether data centers in orbit could eventually supplement ground-based infrastructure.
The immediate effect for users is that Anthropic will double the five-hour rate limits on Claude Code for its Pro, Max, Team, and seat-based Enterprise subscribers. It is also lifting peak-hour usage restrictions for Pro and Max accounts and raising API rate limits for Claude Opus models, according to the company’s announcement.
The SpaceX partnership adds to a compute portfolio Anthropic has been assembling at speed. The company listed several other agreements in its announcement: up to 5 gigawatts with Amazon (including nearly 1 GW of new capacity by the end of 2026), a 5 GW deal with Google and Broadcom beginning in 2027, a strategic partnership with Microsoft and Nvidia covering $30 billion of Azure capacity, and a $50 billion infrastructure investment with Fluidstack.
Taken together, these commitments represent tens of billions of dollars flowing into data center construction and chip procurement across multiple countries. For the global economy, the implications run beyond Silicon Valley. Anthropic stated it plans to expand internationally, adding inference capacity in Asia and Europe through its Amazon collaboration, targeting enterprise customers in regulated industries such as financial services, healthcare, and government that require in-region infrastructure for compliance.
