When you need GPUs badly enough, you'll take them from just about anyone.
On Wednesday, Anthropic announced a deal to lease the entire Colossus 1 data center from SpaceX, giving it access to over 220,000 Nvidia GPUs and more than 300 megawatts of compute capacity. xAI confirmed the arrangement as well. The facility will be available within the month.
SpaceX also owns xAI, Elon Musk's AI lab behind the Grok chatbot, which he folded into SpaceX earlier this year. xAI competes directly with Anthropic's Claude. In other words, Anthropic just leased a data center from one of its biggest rivals' parent company.
The arrangement makes a lot more sense when you look at what's happening on both sides.
Anthropic is growing faster than it can build infrastructure. CEO Dario Amodei said the company projected 10x growth and experienced an 80-fold jump in both revenue and usage instead. Claude Code alone is generating over $2.5 billion in annualized revenue while authoring 4% of all public GitHub commits worldwide. When demand is that far ahead of supply, you take GPUs wherever you can get them.
SpaceX had the opposite problem: too many GPUs and not enough to do with them. All 11 of xAI's original co-founders have departed the company, and Musk acknowledged in March that xAI's coding tools weren't competitive with Claude Code or OpenAI's Codex. The lab's GPU utilization rate across its 550,000 chips sat at roughly 11%, compared to over 40% at Meta and Google. Leasing idle compute to a paying customer beats letting it depreciate, especially with SpaceX reportedly targeting a $1.75 trillion IPO this summer.
Musk appeared comfortable with the partnership. "I spent a lot of time last week with senior members of the Anthropic team to understand what they do to ensure Claude is good for humanity and was impressed," he posted on X. "No one set off my evil detector." That's a notable comment from someone who has previously called Anthropic "evil."
For people who use Claude, the payoff is immediate. Anthropic announced three changes effective the same day:
- Claude Code rate limits doubled for Pro, Max, Team, and Enterprise users
- Peak-hour throttling removed for Pro and Max subscribers
- API rate limits for Opus models raised substantially, with some tiers seeing up to five times the previous capacity
This caps an already massive week for Anthropic, which also launched an enterprise joint venture with Blackstone, Goldman Sachs, and Hellman & Friedman to deploy Claude across large organizations, as we covered on Monday.
This deal is the clearest signal yet about where the AI race actually stands. Anthropic's models are in such high demand that it will rent a data center from a direct competitor's parent company, and xAI's infrastructure failed so completely that leasing it out was the smarter business decision. The next phase of this industry won't be defined by whoever builds the smartest model. It'll come down to whoever can get enough compute to actually serve the people who want to use it. Right now, Anthropic's biggest problem isn't Claude. It's keeping up with everyone trying to use Claude, and Elon Musk just became the fix.
