Anthropic boosts Claude limits after SpaceX Colossus access

Anthropic raised usage limits for its Claude family of models after securing access to SpaceX’s Colossus 1 data center in Tennessee. The company said the additional capacity lets it relax peak-time restrictions and increase API throughput for heavy users.

The firm doubled Claude Code’s five-hour rate limits for Pro, Max, Team and seat-based Enterprise plans so users can make more prompts and write more code in each rolling session. Peak-hour cap reductions for Pro and Max customers were removed. Anthropic also increased Claude Opus API rate limits by more than an order of magnitude; it provided an example of tier 1 API customers moving from 30,000 input tokens per minute to 500,000 input tokens per minute.

SpaceX has agreed to let Anthropic use all compute capacity at the Colossus 1 site, which was originally built for another AI project. Anthropic wrote that the facility will add more than 300 megawatts of capacity and over 220,000 NVIDIA GPUs within the month, and that the extra machines have already enabled the company to raise limits across its services.

The Colossus 1 agreement supplements multiple recent capacity deals. Anthropic has contracted with Amazon, Google, Broadcom, Microsoft and Fluidstack to expand compute. The company runs Claude on a mix of hardware, including AWS Trainium chips, Google TPUs and NVIDIA GPUs, and said some of the new capacity will be located outside the United States. The Amazon arrangement will add inference capacity in Asia and Europe.

Anthropic cited demand from enterprise customers for in-region infrastructure, noting that regulated industries such as financial services, healthcare and government often require local data residency and compliance controls. The company said it will add capacity in countries whose legal and regulatory frameworks support investments of this size and where the hardware and network supply chain can be secured.

On power and community impact, Anthropic and other U.S. AI firms agreed to cover the consumer electricity costs tied to their data center rollouts in the United States. The company said it is exploring ways to extend that commitment to additional jurisdictions.

The changes come as demand for model inference and training has risen across the industry. Companies have adjusted pricing and access models in response to higher compute costs; for example, GitHub recently shifted Copilot pricing toward consumption-based billing to address increased inference demand. Anthropic said the additional Colossus capacity will be brought online quickly to support developer workflows and enterprise customers that need predictable, high-throughput access.