Anthropic announced on May 6, 2026 that it had agreed to a compute partnership with SpaceX and was raising usage limits for Claude. The unusual part is how directly the infrastructure deal is tied to the product experience.
Anthropic says it signed an agreement to use all of the compute capacity at SpaceX’s Colossus 1 data center, giving it access to more than 300 megawatts of new capacity and over 220,000 NVIDIA GPUs within the month.
What changed
Anthropic says Claude Code five-hour rate limits are doubling for Pro, Max, Team, and seat-based Enterprise plans. It is also removing peak-hours limit reductions for Claude Code on Pro and Max accounts, and raising API rate limits for Claude Opus models.
The company lists the SpaceX agreement alongside other compute commitments: up to 5 GW with Amazon, 5 GW with Google and Broadcom, a Microsoft/NVIDIA Azure capacity partnership, and a $50 billion U.S. infrastructure investment with Fluidstack.
Why this matters
For users, model quality is only half the product. Rate limits, peak-hour throttling, latency, and availability shape whether an AI tool can be trusted for real work.
Anthropic is effectively saying compute capacity is now a feature. If the company wants Claude Code and Claude API workloads to become daily infrastructure, it needs enough hardware to make high-intensity use feel normal rather than scarce.
What to watch next
Watch whether higher limits stay durable during demand spikes. Also watch whether compute concentration creates dependency risk. The AI model race increasingly looks like a power, data-center, and supply-chain race.
