Anthropic raises Claude limits and signs compute deal with SpaceX | Keryc
Anthropic announces changes aimed at improving the experience of heavy users and enterprise customers of Claude. What’s behind the news? More compute capacity and higher usage limits for key features, plus a big deal with SpaceX that accelerates that expansion.
What's changing in usage limits
The company applied three changes, effective immediately, designed for people who use Claude intensively or in enterprise environments:
The five-hour limit for Claude Code is doubled for Pro, Max, Team, and Enterprise plans with seats.
The peak-hour throttling for Claude Code on Pro and Max accounts is removed.
Rate limits on the API are significantly increased for the Claude Opus models (Anthropic shows a table with specific details on their website).
Why does this matter to you? If you use for development, analysis, or automation, you’ll notice fewer interruptions and more continuous productive time. For companies, it means fewer restrictions during busy periods.
Claude Code
The deal with SpaceX and what it brings
Anthropic signed an agreement to use the full compute capacity of SpaceX’s Colossus 1 data center. This adds more than 300 megawatts of new capacity and access to over 220,000 NVIDIA GPUs within a month. In practice, that extra capacity will mean better availability and performance for Claude Pro and Claude Max subscribers.
This announcement comes alongside other major agreements Anthropic has secured:
Up to 5 gigawatts with Amazon, with nearly 1 gigawatt additional planned by the end of 2026.
5 gigawatts with Google and Broadcom, which will start coming online in 2027.
A strategic partnership with Microsoft and NVIDIA that includes $30 billion of Azure capacity.
A $50 billion investment in AI infrastructure in the U.S. with Fluidstack.
Additionally, Anthropic is exploring the possibility of developing orbital compute capacity with SpaceX—an ambitious idea that aims to take massive compute off the planet.
Access to more GPUs and power means fewer bottlenecks, faster responses, and greater capacity for large workloads.
International expansion and local responsibility
Part of the expansion will be international to serve customers in regulated industries like finance, healthcare, and governments, which need infrastructure in-region for compliance and data residency.
Anthropic says it will pick partners in democracies with stable legal frameworks and secure supply chains. It also committed to covering any U.S. consumer electricity price increases caused by its data centers and is evaluating how to extend that commitment to new countries.
That approach isn’t just technical: it involves negotiating with local authorities and reinvesting in the communities that host these facilities.
What this means for users, companies, and developers
For Pro and Max users: fewer limits and less impact from peak hours. If you rely on Claude Code for long programming sessions or content generation, you’ll feel the difference.
For regulated companies: more in-region infrastructure availability, which makes it easier to meet data rules and maintain operational continuity.
For developers using the API: higher limits for Claude Opus let you integrate larger workloads and handle spikes without as many rate rejections.
And for the general public: a race for more capacity that also raises questions about energy consumption, local impact, and governance. Anthropic says it’s aiming to balance expansion with responsibility.
Plugins, tools and ecosystem
At the same time, Anthropic keeps expanding Claude’s ecosystem: new plugins for Cowork and Claude Code, integrations with Microsoft 365, additional connectors, and an MCP app focused on financial services and insurance. In other words, the extra capacity isn’t just about speed; it’s about supporting more integrations and enterprise features.
In short: Anthropic is betting big on scaling its infrastructure to improve experience and compliance, while opening the door to bold initiatives like orbital compute. Does it sound like a lot? Yes. Is it relevant to anyone who depends on AI today? Also yes.
