Claude Doubles Usage Limits and Partners With SpaceX for Massive Compute Expansion.

Anthropic announced a partnership with SpaceX that will give the AI company access to more than 300 megawatts of additional compute capacity, enabling immediate increases to usage limits across Claude Code and the Claude API.

The deal gives Anthropic exclusive access to all available capacity at SpaceX's Colossus 1 data center, which the company expects to deploy within the month. The compute expansion arrives as demand for Claude's most capable models continues to strain existing infrastructure, particularly during peak usage windows.

Effective immediately, Anthropic has doubled the 5-hour rate limits for Claude Code across Pro, Max, and Team subscription tiers. The company also removed the peak-hours limit reduction that previously throttled Claude Code access for Pro and Max subscribers during high-traffic periods.

For API users, Anthropic substantially raised rate limits for Opus models, though the company did not publish specific numbers in its announcement on LinkedIn. The changes address a longstanding complaint from developers who hit rate limits during production workloads, forcing them to queue requests or switch to competing models.

The timing follows Anthropic's recent launch of Cowork, which brought Claude Code's automation capabilities to non-technical tasks. That product expanded the potential user base for compute-intensive features beyond software developers.

The SpaceX partnership represents a different approach to securing compute compared to typical cloud provider agreements. By taking all capacity at a single data center, Anthropic gains predictable access to a large block of infrastructure without competing for shared resources.

Colossus 1 is SpaceX's first dedicated AI training facility, originally built to support the company's Starlink satellite network optimization and other internal machine learning projects. The data center uses a custom cooling system designed for the high-density GPU clusters required for large language model training and inference.

Anthropic noted in its announcement that this deal is one of several recent compute agreements, suggesting the company is diversifying its infrastructure partnerships rather than relying solely on traditional cloud providers like AWS or Google Cloud.

Pro subscribers paying $20 per month now get 10 hours of Claude Code usage per rolling 5-hour window, up from 5 hours. Team plan subscribers, who pay $30 per user monthly, receive the same doubling. Max plan users, at $200 monthly, also see their limits double and no longer face reduced capacity during peak hours.

The removal of peak-hours throttling is arguably more significant than the raw limit increase for many professional users. Previously, Pro and Max subscribers might hit artificially low caps during business hours in US time zones, even if they hadn't exhausted their nominal 5-hour allocation. That constraint disappears with today's changes.

API developers using Claude 3 Opus, the most capable model in Anthropic's lineup, can now make more requests per minute without triggering rate limit errors. The company has not yet updated its public API documentation with new specific limits, but developers reported seeing immediate increases when testing after the announcement.

For small businesses and builders using Claude Code for automation tasks or HIPAA-compliant workflows, the expanded limits reduce friction around scheduling compute-intensive jobs during off-peak hours. Teams can now run more concurrent sessions without hitting walls or carefully timing their usage windows.