Claude gains lift from SpaceX compute

Anthropic has struck a major compute agreement with SpaceX that will give the maker of Claude access to all capacity at the Colossus 1 data centre, adding more than 300 megawatts of power and over 220,000 Nvidia GPUs within a month.

The deal is aimed at easing capacity pressure on Claude as demand rises for AI assistants, developer tools and enterprise services. Anthropic said the extra resources will directly improve access for Claude Pro and Claude Max users, while allowing higher limits for Claude Code and its application programming interface. The arrangement also opens a more speculative track: possible collaboration with SpaceX on multiple gigawatts of orbital AI compute capacity.

Colossus 1, based in Memphis, Tennessee, has become one of the most prominent symbols of the AI industry’s race for chips, power and data-centre scale. For Anthropic, the agreement offers an immediate route to more inference capacity at a time when top AI companies are competing not only on model quality but also on availability, speed and reliability. Claude Code, which has gained traction among software developers, is a particular beneficiary, with five-hour rate limits doubled for Pro, Max, Team and seat-based Enterprise plans.

The agreement also removes peak-hour limit reductions for Claude Code on Pro and Max accounts and raises API rate limits for Claude Opus models. These changes matter because AI companies increasingly face a constraint that is less about software demand than physical infrastructure: access to advanced chips, electricity, cooling, networking and large-scale facilities.

For SpaceX, the agreement gives Elon Musk’s company a marquee customer for AI infrastructure as it develops broader commercial ambitions beyond rockets, satellites and connectivity. The deal also reflects an unusual alignment in the AI market. Musk’s xAI competes directly with Anthropic through its Grok models, yet SpaceX is now providing capacity to a rival AI lab. Musk has indicated that SpaceX’s own AI training work has moved to Colossus 2, leaving Colossus 1 available for outside use.

The partnership marks a shift in tone between Musk and Anthropic. He had criticised the company earlier this year over perceived ideological bias in its models, but changed his stance after discussions with Anthropic’s leadership on AI safety and the company’s approach to making Claude beneficial for users. The détente comes as Musk remains locked in legal conflict with OpenAI over claims tied to its founding mission and corporate direction.

Anthropic’s compute strategy has been expanding across several fronts. The company has announced large-scale arrangements with Amazon, Google, Broadcom, Microsoft, Nvidia and Fluidstack, reflecting a deliberate effort to avoid reliance on a single hardware or cloud provider. Its infrastructure mix includes AWS Trainium, Google TPUs and Nvidia GPUs, a diversified approach designed to support both model training and high-volume customer use.

The SpaceX deal is distinctive because of its orbital-compute dimension. Space-based data centres remain technically difficult and commercially unproven, but the idea is gaining attention as AI growth strains terrestrial power grids and data-centre sites. Advocates argue that orbital infrastructure could one day draw on abundant solar power and reduce some land-based constraints. Sceptics point to launch costs, maintenance complexity, latency, radiation exposure, thermal management and the challenge of returning or upgrading hardware in space.

The immediate business case, however, is on Earth. AI coding assistants have become a central battleground for Anthropic, OpenAI, Google and other model developers. Enterprise customers are adopting such tools to accelerate software development, automate routine engineering tasks and integrate AI agents into business workflows. Claude Code’s usage growth has made capacity a strategic priority rather than a back-office concern.

Anthropic is also pushing deeper into regulated sectors, including financial services, healthcare and government, where customers require stronger assurances on compliance, data residency and security. The company has said part of its infrastructure expansion will be international, with additional inference capacity planned in Asia and Europe. That expansion reflects a wider industry shift as customers seek lower latency, localised compliance and more predictable service levels.