Anthropic has increased usage limits for its paid Claude subscribers as the company scales up its computing resources through a partnership with SpaceX. The added capacity is expected to improve performance and availability for users on the Claude Pro and Claude Max plans, while the free tier will continue with existing limits.
Anthropic said its partnership with SpaceX will give the company full access to the computing resources available at SpaceX’s Colossus 1 data centre. Anthropic added that the deal is expected to bring in more than 300 megawatts of additional compute power, along with access to over 220,000 NVIDIA GPUs within the next month.
Anthropic expands Claude usage limits for paid users:
Anthropic has announced a fresh set of upgrades to Claude usage limits for paying customers, and the changes are already live. One of the biggest updates is a doubling of Claude Code’s five-hour usage allowance for users on the Pro, Max, Team, and seat-based Enterprise plans.
The company has also removed the temporary peak-hour restrictions that previously affected Pro and Max subscribers using Claude Code. Alongside this, Anthropic revealed expanded API rate limits for its Claude Opus models.
Under the revised limits, Tier-I users can now access up to 500,000 input tokens per minute, a major jump from the earlier cap of 30,000. Output token limits for the same tier have also been raised significantly, increasing from 8,000 to 80,000 tokens per minute.
Anthropic has sharply increased token limits across its higher usage tiers. For Tier-II users, input token capacity has been raised from 450,000 to 2 million tokens per minute, while output limits have more than doubled from 90,000 to 200,000 per minute.
Users in Tier-III will now be able to access up to 5 million input tokens along with 400,000 output tokens every minute. Meanwhile, Tier-IV users are getting the highest allocation yet, with support for up to 10 million input tokens and 800,000 output tokens per minute.
Anthropic expands global AI infrastructure
Anthropic said its agreement with SpaceX is part of a larger effort to strengthen and scale its AI infrastructure worldwide. The company also pointed to earlier partnerships focused on expanding computing capacity, including a deal with Amazon that could provide access to as much as 5 gigawatts of compute power. Of that total, nearly 1 gigawatt is expected to become available by the end of 2026.