SAN FRANCISCO, May 8, 2026, 14:02 PDT
Anthropic has turned to Elon Musk’s SpaceX for a large new block of AI computing power, agreeing to use all capacity at the Colossus 1 data center in Memphis as the Claude maker tries to ease usage limits that have frustrated paying users and developers.
The deal matters now because demand for Claude, especially the company’s Claude Code programming tool, has run ahead of Anthropic’s infrastructure. Chief Executive Dario Amodei said the company had planned for 10-fold growth, only to see revenue and usage grow 80-fold in the first quarter on an annualized basis, a surge he called “just crazy” and “too hard to handle.” Fortune
Compute — the processing power used to train and run AI models — has become one of the industry’s main choke points. Anthropic said the SpaceX agreement gives it access within a month to more than 300 megawatts of new capacity and more than 220,000 Nvidia GPUs, the chips widely used to train and serve large AI systems.
SpaceXAI said Colossus 1 includes Nvidia H100, H200 and GB200 accelerators and can handle training, fine-tuning and inference, the process of running an already-trained model for users. Anthropic said the added capacity would directly improve service for Claude Pro and Claude Max subscribers.
Anthropic moved quickly to turn the deal into product changes. The company said it was doubling Claude Code’s five-hour rate limits for Pro, Max, Team and seat-based Enterprise plans, removing peak-hour reductions for Pro and Max users, and raising API limits for Claude Opus models. Rate limits are caps on how many requests a user or developer can make in a set period.
“We’ve had difficulties with compute,” Amodei said on stage at Anthropic’s developer conference, according to Business Insider. “We’re sorry if sometimes it takes some time, but we’re gonna keep going to acquire as much as we can.” The report said Anthropic had suffered outages and imposed usage limits in recent months, while some developers had shifted work to OpenAI’s Codex in part because OpenAI had secured more capacity earlier. Business Insider
For SpaceX, Reuters reported, the agreement gives an IPO-bound company a high-profile AI customer as it tries to show investors a broader role in the AI infrastructure race. Musk wrote on X that he approved the lease after meeting Anthropic leaders and that “No one set off my evil detector,” a notable shift after earlier public criticism of the company. Reuters
The pairing is unusual because Anthropic competes with Musk’s own AI operation, xAI, whose Grok chatbot was the original reason Colossus was built. It also puts Anthropic in the same scramble for chips, power and data-center space as OpenAI and Google, with the contest shifting from model launches to who can keep enough infrastructure online.
Anthropic used the San Francisco event to pitch more than capacity. It also introduced a “dreaming” feature for Claude Managed Agents, AI programs that can carry out tasks with limited human direction. The feature is meant to let agents review prior work between sessions, find patterns and update memory files that store user preferences and context. Business Insider
Boris Cherny, Anthropic’s head of Claude Code, framed the shift as a move toward more automated software work. “The default isn’t, ‘I’m going to prompt Claude Code.’ The default is now, ‘I will have Claude prompt Claude Code,’” he said at the event, according to Reuters. Reuters
Anthropic’s wider infrastructure push now spans several large suppliers. The company cited agreements with Amazon for up to 5 gigawatts of capacity, Google and Broadcom for another 5 gigawatts beginning in 2027, Microsoft and Nvidia for $30 billion of Azure capacity, and a $50 billion U.S. AI infrastructure plan with Fluidstack.
But the risks are plain. Reuters said the idea of orbital AI data centers would be capital intensive and technically difficult, while Data Center Dynamics noted that the Colossus site has drawn criticism from civil-rights groups over natural-gas turbines. If demand keeps outpacing fresh capacity, Anthropic may still face the same problem it is trying to solve: more users than available compute.
