Anthropic takes all of SpaceX's Colossus 1 compute, raises Claude Code limits

Anthropic signed a deal to use the entire compute capacity of SpaceX's Colossus 1 data center: more than 300 megawatts, roughly 220,000 NVIDIA GPUs, available within weeks rather than years.

What separates this from a routine infrastructure post is how directly it reaches users. Alongside the deal, Anthropic doubled Claude Code rate limits for Pro, Max, Team, and seat-based Enterprise plans across their five-hour windows, and removed the peak-hours limit for Claude Code on Pro and Max accounts entirely. API rate limits for Opus models also went up. Lifting a restriction signals more confidence than raising a quota: it suggests Anthropic sees the new capacity as real and close, not a distant promise.

The agreement sits alongside Anthropic's other infrastructure commitments with Amazon, Google, Microsoft/NVIDIA, and Fluidstack. For anyone whose daily work leans on Claude Code, the part that matters is not the megawatt figure. It is the quiet removal of the peak-hours cap. The full details, including a table of the new API limits, are in Anthropic's announcement.

Why it matters

If you hit Claude Code limits during the workday, this changes your routine now. The peak-hours throttle on Pro and Max is gone and rate limits are doubled, so the habits you built around busy-hour slowdowns are no longer needed.