220,000 NVIDIA GPUs and double the Claude Code limits: Anthropic’s SpaceX deal explained
Follow Us
For someone suing OpenAI for abandoning their original mission, Elon Musk also doesn’t seem to be able to stick to one opinion. On the one hand, he spent weeks bashing Anthropic calling them evil, “misanthropic”, a company that hates western civilisation and a bunch of other things. On the other, he just signed a compute partnership with them giving them access to the entire Colossus 1 data centre.
SurveyAlso read: Forts, monkeys, and a telephoto lens: The Vivo X300 FE experience in Jaipur
This gives Anthropic over 300 megawatts of new capacity, backed by over 220,000 NVIDIA GPUs, with the company expecting it to come online within the month. This has also come at the perfect time for Anthropic as the demand for Claude models just seems to keep increasing.
What’s actually changing for users
The deal isn’t just a headline number. Anthropic has tied the SpaceX partnership directly to product improvements that have come into effect immediately. Claude Code’s five-hour rate limits are being doubled across Pro, Max, Team, and seat-based Enterprise plans, a change that will be immediately noticeable for developers running long agentic sessions. On top of that, the peak hours limit reduction on Claude Code has been scrapped entirely for Pro and Max accounts, meaning users will no longer hit a wall during high-traffic periods. API rate limits for Claude Opus models have also been raised considerably, making the company’s most capable models more accessible for production workloads.
Part of a much bigger compute push
Also read: Anthropic’s finance agents: Claude can now build pitchbooks, run KYC, and close the books
However, the SpaceX agreement is part of a bigger picture. Anthropic has been putting together one of the most extensive compute portfolios ever seen. In addition to up to 5 gigawatts from Amazon, Anthropic is working with Google and Broadcom to deliver another 5 gigawatts from 2027 and entering a $30 billion capacity deal for Azure via a strategic collaboration with Microsoft and NVIDIA. Additionally, there’s another $50 billion commitment to invest in U.S. AI infrastructure in conjunction with Fluidstack.
The SpaceX agreement also includes an expression of interest in developing multiple gigawatts of orbital AI compute, a longer-horizon bet that signals how seriously Anthropic is thinking about the infrastructure constraints of the next decade.
The company is expanding its efforts beyond American borders, targeting regulated industries like finance, health care, and government that demand regional presence and data security. Apart from the Amazon agreement that includes additional inference capacity in Asia and Europe, Anthropic claims it will expand only in stable, democratically governed nations.
In a notable gesture, Anthropic has committed to covering any consumer electricity price increases caused by its US data centres, and is exploring extending that pledge internationally.
For subscribers and developers, the immediate takeaway is pretty simple – more headroom, fewer throttles, and a company that is clearly betting big on the infrastructure needed to back it up.
Also read: AI Sovereignty to space: Pixxel and Sarvam are building India’s first orbital data centre satellite
