Anthropic has signed an agreement with SpaceX to access “Colossus 1,” a large-scale AI supercomputer, as part of efforts to expand computing capacity and improve performance for its Claude models.
According to the announcement, Colossus 1 is powered by more than 220,000 GPUs from NVIDIA, including H100, H200, and GB200 accelerators. The system is designed to support large language model training, multimodal AI systems, scientific simulations, and generative workloads.
The agreement provides Anthropic with over 300 megawatts of additional compute capacity within a month. The company said this will be used to enhance services for Claude Pro and Claude Max subscribers, reflecting growing demand for advanced AI capabilities.
Anthropic also indicated interest in working with SpaceX on future projects involving multiple gigawatts of orbital AI compute capacity. The company noted that the compute needs of next-generation AI systems are exceeding what traditional land-based infrastructure can support in terms of power, cooling, and space.
In parallel, Anthropic announced product updates, including increased API rate limits for Claude Opus models and expanded usage limits for Claude Code. The company said it is doubling the five-hour rate limits across Pro, Max, Team, and Enterprise plans, while removing peak-hour restrictions for Pro and Max users.
The SpaceX agreement adds to Anthropic’s broader infrastructure strategy. The company already has compute partnerships with Amazon, Alphabet, Broadcom, Microsoft, and NVIDIA, alongside a reported $50 billion investment in U.S.-based AI infrastructure with Fluidstack.
The developments highlight intensifying competition in the AI sector, where companies are scaling infrastructure to meet rising demand for advanced models and high-performance computing.