Anthropic on Wednesday said it reached a deal to tap the computing resources of Elon Musk’s SpaceX, marking a détente with its one-time critic and a boost for both companies in the high-stakes artificial intelligence race. Under the agreement, Anthropic will use the full computing power of SpaceX’s Colossus 1 facility in Memphis, Tennessee, which houses more than 220,000 Nvidia processors and will give the Claude chatbot maker 300 megawatts of new capacity within a month.
The deal gives the IPO-bound SpaceX a marquee customer as it looks to sell investors on its AI ambitions, while helping Anthropic ease capacity constraints following a surge in demand for products, such as its AI coding tool, Claude Code.
The announcement came as Anthropic held a developer day in San Francisco on Wednesday, where it unveiled a new Claude AI feature called “dreaming” that is meant to help its AI systems learn by reviewing work between sessions, spotting patterns and updating files that store user preferences and other context.
Buoyed by the extra capacity from the SpaceX deal and other similar recent agreements, Anthropic said it was doubling Claude Code’s rate limits for its paid plans, removing peak-hour usage caps for Pro and Max accounts and sharply increasing the volume of requests developers can make to its Claude Opus models.
The company said it is also interested in working with SpaceX to develop multiple gigawatts of space-based orbital data centres: one of Musk’s key goals and a major driver behind SpaceX’s initial public offering, as the endeavour is expected to be highly capital intensive and technically challenging.
Ryan Mallory, CEO of data centre operator Flexential, said, “The fact that serious companies are even discussing compute capacity in space tells you how aggressively the market is searching for power and scale.”
In a statement on X, Musk said that he made the decision to lease computing power after spending time with Anthropic leaders last week. Their work to ensure Anthropic’s Claude AI is “good for humanity” impressed him, he said.
“No one set off my evil detector,” wrote Musk, who is battling OpenAI and its CEO Sam Altman in court for allegedly breaching its mission to develop AI for the good of humanity.
He added SpaceX has moved its AI training efforts to Colossus 2 and that he would provide computing capacity to other AI companies that make similar efforts to favor humanity, like how SpaceX launches satellites for competitors with “fair terms and pricing.”
The comments marked a sharp shift from remarks Musk made in February, when he accused Anthropic’s AI of bias. “Frankly, I don’t think there is anything you can do to escape the inevitable irony of Anthropic ending up being misanthropic,” he wrote at the time in a post on X.
Anthropic said on Wednesday that its “dreaming” feature was available as a research preview and comes with its software for managing agents, or AI programmes that perform tasks with little human involvement.
The move is part of Anthropic’s push to win business customers, following a surge in popularity for Claude Code that has intensified competition and prompted OpenAI to scale back efforts, such as its Sora video-generation tool, to focus more on the fast-growing market for AI-powered coding.
On stage at the San Francisco event, Boris Cherny, Anthropic’s head of Claude Code, showed off how developers could set up “routines” that schedule Anthropic’s AI computer programmer to take action.
“The default isn’t, ‘I’m going to prompt Claude Code.’ The default is now, ‘I will have Claude prompt Claude Code,’” he said.
Cherny closed out the keynote with a message on AI’s productive power. “The capability is already here. The gap left is how fast we put it to work,” he said.