SpaceX Will Rent 220,000 GPUs to Anthropic, Immediately Doubling Claude Code Rate Limits
xAI's Colossus 1 supercomputer — originally built to train Grok — will now power Claude. The deal instantly doubles Claude Code's rate limits and signals a new era of cross-lab compute leasing.
SpaceX's AI division will provide Anthropic with access to Colossus 1, one of the largest GPU clusters ever assembled, in a compute-sharing deal that both companies announced Tuesday. The agreement takes effect immediately, with Anthropic confirming it is doubling Claude Code's five-hour rate limits and substantially raising API rate limits for Opus models. For developers who have been hitting Claude Code's usage ceiling mid-session — a persistent frustration since the tool's rise as a primary coding agent — the relief is tangible and instant.
The arrangement is unusual by any measure. Colossus 1 houses more than 220,000 NVIDIA GPUs, as NVIDIA itself noted in a congratulatory post: "Two frontier labs. One accelerated computing platform." The cluster was purpose-built by xAI over the past two years to train and serve Grok, and its rapid deployment became a point of pride for Elon Musk. Now, excess capacity on that same infrastructure will run inference for a rival's flagship model. It is, as far as publicly known deals go, the first time a frontier lab has rented significant GPU capacity directly to another frontier lab.
Get our free daily newsletter
Get this article free — plus the lead story every day — delivered to your inbox.
Want every article and the full archive? Upgrade anytime.
No spam. Unsubscribe anytime.