Anthropic has cut a compute deal with SpaceX so Claude users can get higher limits instead of running into the same annoying wall during heavy work.
The company is taking capacity from SpaceX’s Colossus 1 data center, and the deal gives it access to more than 300 megawatts of new power tied to over 220,000 Nvidia (NVDA) GPUs within the month.
The first changes are already aimed at Claude’s busiest users. Claude Code now gets double the five-hour usage limits on Pro, Max, Team, and seat-based Enterprise plans.
Anthropic is also ending the peak-hour limit cut for Claude Code on Pro and Max accounts. API users get a bigger ceiling too, with higher rate limits for Claude Opus models.
Anthropic said the SpaceX deal will feed directly into Claude Pro and Claude Max capacity. That matters because Claude Code has become one of the company’s main products for developers, especially users who keep long coding sessions open and rely on Claude for debugging, refactoring, file review, agent-style tasks, and API-heavy workflows.
The SpaceX capacity does not sit alone. Anthropic has been piling up infrastructure deals across cloud firms, chipmakers, and data center partners. It has an agreement with Amazon (AMZN) for up to 5 gigawatts of capacity, with nearly 1 gigawatt expected by the end of 2026.
It also has a 5-gigawatt agreement with Alphabet’s Google (GOOGL) and Broadcom (AVGO), with capacity due to start coming online in 2027.
There is also a strategic partnership with Microsoft (MSFT) and Nvidia (NVDA) that includes $30 billion of Azure capacity. On top of that, Anthropic has a $50 billion plan for U.S. AI infrastructure with Fluidstack.
Claude runs across several kinds of AI hardware. Anthropic uses AWS Trainium, Google TPUs, and Nvidia GPUs for training and serving its models.
The SpaceX agreement also includes a stranger future path. Anthropic said it has interest in working with SpaceX on multiple gigawatts of orbital AI compute capacity.
Anthropic is also adding capacity outside the United States because large enterprise users have strict data rules. The company said customers in financial services, healthcare, and government need local infrastructure for compliance and data residency. Its Amazon deal includes extra inference capacity in Asia and Europe.
The company said it wants new sites in democratic countries with legal and regulatory systems that can handle very large AI infrastructure. It also pointed to secure supply chains for chips, networking gear, and facilities.
Now comes the stock market headache. Bank of America (BAC) warned that the possible public listings of SpaceX and Anthropic may not be the easy jackpot some investors expect. Wall Street has been preparing for major IPOs, helped by Trump’s push to loosen rules around public and private markets under the phrase “make IPOs great again.”
Index providers have also been changing rules for faster entry into indexes and adjusting free-float math. Savita Subramanian, Bank of America’s equity and quant strategist, said those steps “smack of late-stage machinations.”
Savita warned that giant IPOs could send a huge batch of new shares into the market and pressure prices. She wrote, “Say goodbye to the ‘equity shrinkage’ bull case.” She also added, “Today, an issuance deluge may be imminent.”
The old bull case relied partly on fewer public shares. U.S. publicly traded securities fell to about 4,000 last year from more than 8,000 in the 1990s, based on Center for Research in Security Prices data.
The S&P 500 rose more than 10% in April, while the equal-weighted version gained only 6%. The Magnificent Seven now make up about one-third of the index, and passive funds hold heavy exposure to those names. Savita said about 60% of U.S.-domiciled assets are passively managed.
Her warning was simple enough: “This implies, all else equal, passive funds will be forced to free up capital for new issues, creating downward pressure on existing holdings.”
Don’t just read crypto news. Understand it. Subscribe to our newsletter. It’s free.