Anthropic Reportedly Expanding to xAI's Colossus2, Tapping Nvidia GB200 Hardware

A social media post claims Anthropic is extending its compute footprint to Colossus2, the next-generation supercomputer cluster associated with xAI, and will run workloads on Nvidia’s GB200 Grace Blackwell systems. If accurate, the move would mark an unusual cross-pollination between competing AI labs at the infrastructure layer, with Anthropic joining the small group of frontier developers committing to GB200-class hardware for training and inference at scale.

The linked source is an X post that could not be retrieved without JavaScript, so the underlying claim cannot be independently verified from the page content itself. Anthropic has historically leaned on Amazon Web Services and Google Cloud for compute, making a Colossus2 arrangement a notable diversification if confirmed.

The significance, assuming the report holds up, is twofold: it would signal that demand for top-tier GB200 capacity is outrunning what any single hyperscaler partner can supply on Anthropic’s timeline, and it would further blur the lines between rival labs as they compete for the same scarce silicon and power.