Anthropic Secures 220,000 GPUs at SpaceX Colossus 1 Data Centre

The battle for artificial intelligence supremacy isn't just about algorithms anymore; it's fundamentally about raw compute power. Anthropic has just made a massive play in this arena by securing an agreement with SpaceX.

The deal grants Anthropic access to the entire compute capacity of xAI's Colossus 1 data centre in Memphis, Tennessee. We're talking about more than 300 megawatts of new capacity coming online within a month. That translates to over 220,000 NVIDIA GPUs, specifically dense deployments of H100, H200, and the next-generation GB200 accelerators. This is more than just an upgrade; it's a structural shift in Anthropic's operational capacity.

To understand the scale of this, consider the capital involved. A single NVIDIA H100 GPU costs tens of thousands of pounds. Multiplying that by 220,000, and adding the necessary infrastructure for cooling, power distribution, and high-speed networking, represents a multi-billion-pound investment. By leasing this capacity from SpaceX rather than building it from scratch, Anthropic gains immediate access to top-tier hardware without the crushing upfront capital expenditure.

It's a strategic move that allows them to scale rapidly to meet user demand while maintaining financial agility in a fast-moving hardware market.