Anthropic is dramatically loosening the reins on Claude usage starting today. The Claude Code rate limit — previously capped at 5 hours — is being doubled across Pro, Max, Team, and Enterprise tiers. Pro and Max users will no longer see their speeds throttled during peak hours, and API limits for the Opus model are getting a big bump too.
What makes all this extra capacity possible? Anthropic just signed a deal for the entire compute output of SpaceX's Colossus 1 data center — that's over 220,000 NVIDIA GPUs drawing more than 300 megawatts — coming online within the month and flowing directly to Claude Pro and Max. The two companies also hinted at a longer-term ambition: jointly developing multi-gigawatt-scale orbital AI compute.
Toss in Anthropic's other infrastructure deals — up to 5GW with Amazon (nearly 1GW by year's end), 5GW with Google and Broadcom (delivery from 2027), $30 billion in Azure compute from Microsoft and Nvidia, and $50 billion in infrastructure investment from Fluidstack — and you get an expansion campaign running on three hardware fronts: AWS Trainium, Google TPUs, and NVIDIA GPUs.