In a move that has surprised many in the AI industry, Anthropic has struck an unexpected deal with Elon Musk-owned SpaceX to expand computing capacity for its Claude AI models.
The agreement gives Anthropic access to 300 megawatts of new compute power from SpaceX’s Colossus 1 data center in Memphis, Tennessee. That facility was originally built to support xAI’s Grok chatbot and reportedly contains more than 220,000 Nvidia processors. The extra capacity is expected to become available within a month.
This deal comes as Anthropic faces growing demand for products such as Claude Code, its AI coding assistant. The company has also discussed partnering with SpaceX on future orbital data center projects that could eventually deliver several gigawatts of compute power.
A Shift in Tone
The agreement stands out because top AI firms rarely rent infrastructure to direct rivals in the race for advanced AI models. Musk has repeatedly criticized Anthropic in the past and previously accused the company of drifting away from its original mission. In a post on X, he once said he believed Anthropic risked becoming “misanthropic.”
But his tone changed this week after meetings with Anthropic executives. Musk said he decided to approve the deal after being impressed by the company’s efforts to ensure Claude remains “good for humanity.” He added that no one at Anthropic had triggered his “evil detector,” though he warned he would remove any client from xAI facilities if their AI systems harmed humanity.
New Business Strategy for SpaceX
The move has drawn attention because it could signal a new business strategy for SpaceX and xAI. Reports suggest xAI has already shifted much of its internal AI training work to the newer Colossus 2 system, leaving room to lease capacity to outside firms. Musk also recently signed a similar compute agreement with AI coding startup Cursor.
SpaceX’s entering the cloud computing business could create another revenue stream ahead of a potential public offering. Investors are also watching whether Musk’s companies can turn massive AI infrastructure investments into stable commercial income.
Anthropic’s Developer Conference
Anthropic announced the partnership during its first developer conference in San Francisco. The company said the added capacity would allow it to double Claude Code rate limits for paid users and remove peak-hour usage caps on some plans.
Anthropic also introduced a new “dreaming” feature designed to help Claude learn between sessions by reviewing user workflows and updating stored context automatically.
“We are thrilled to work with SpaceX on this, it allows us to serve more users and improve reliability,” an Anthropic spokesperson said. The company expects the deal to help it handle growing demand without needing to build its own data centers from scratch.
