Anthropic shipped a one-two punch on May 6: significantly higher usage limits for Claude Code and the Claude API, and a new compute partnership with SpaceX that adds 300+ megawatts of capacity and over 220,000 NVIDIA GPUs within the month. The compute deal is the supply story; the limit changes are the immediately-visible developer story. Both shipped in the same announcement on purpose.
What Actually Changed for Developers
Three concrete changes, all effective the day of the announcement and all aimed at the heaviest users of Claude:
First, Claude Code's five-hour rate limits are doubled for Pro, Max, Team, and seat-based Enterprise plans. That's the limit most developers actually feel during a real coding session—the one that turns a productive afternoon into a wait-and-resume. Doubling it is the difference between hitting the wall and not.
Second, the peak-hours limit reduction on Claude Code for Pro and Max is gone. Until now, the cap quietly tightened during the busiest US business hours, which is exactly when most developers were trying to use it. Removing it means Pro and Max accounts get the same limits at 2pm Pacific as they do at 2am.
Third, API rate limits for Claude Opus models are raised considerably. Anthropic published the updated Opus rate-limit table on its API rate-limits docs page; the practical effect is more headroom for any agent that fans out Opus calls or runs long-context Opus jobs at scale.
The SpaceX Colossus 1 Deal
The deal that's funding all of this: Anthropic signed an agreement to use all of the compute capacity at SpaceX's Colossus 1 data center. That's more than 300 megawatts of new capacity, over 220,000 NVIDIA GPUs, available within the month. Anthropic said the additional capacity will directly improve capacity for Claude Pro and Claude Max subscribers—translation: the limit changes above are not theoretical, they're funded.
Colossus 1 sits on top of an already-stacked compute portfolio. Anthropic listed the joiners: an up-to-5 GW agreement with Amazon (with nearly 1 GW coming online by end of 2026), a 5 GW agreement with Google and Broadcom starting in 2027, a strategic partnership with Microsoft and NVIDIA that includes $30 billion of Azure capacity, and a $50 billion investment in American AI infrastructure with Fluidstack. Anthropic also flagged that it has expressed interest in partnering with SpaceX to develop multiple gigawatts of orbital AI compute capacity—a sentence that would have read as science fiction a year ago.
The Hardware Diversity Bet
Anthropic was deliberate about the framing: it trains and runs Claude on AWS Trainium, Google TPUs, and NVIDIA GPUs, and is continuing to explore additional capacity. That tri-vendor stance matters strategically—it's the bet that no single hardware bottleneck (Nvidia allocation, TPU availability, Trainium roadmap) can stall the company's roadmap. Most of Anthropic's frontier competitors are still effectively single-vendor on training.
International expansion is the other dimension. Enterprise customers in financial services, healthcare, and government increasingly need in-region infrastructure for compliance and data residency, so part of the new capacity is going outside the US. The Amazon collaboration in particular includes additional inference in Asia and Europe.
Why It Matters for Web Developers
For solo developers and small teams paying for Claude Code Pro or Max, this is the most direct quality-of-life upgrade Anthropic has shipped in months. The five-hour cap was the single most-felt limit in real coding sessions—doubling it changes how much you can actually do in one stretch. Removing peak-hour throttling means the limit doesn't quietly get tighter at 2pm Pacific when half of San Francisco is also running an agent.
For platform builders and anyone running serious agentic loads against the Opus API, the rate-limit increases are the unblocker. The combination of higher API limits and the Multi-Agent Orchestration cost cut shipped the next day means the same dollar buys substantially more Opus throughput than it did at the start of May. The compute deal is the part that says it'll keep buying more—Colossus 1 is online within the month, the Amazon GW comes by year-end, and the Google and Microsoft commitments start landing in 2027.