xAI’s Colossus 2: The First Gigawatt-Scale AI Datacenter—Power Strategy, RL Bets, Funding Paths, and Product Traction

Much has been written about xAI’s Colossus 1. The Memphis build belongs in the history books: the largest AI-training cluster, assembled from scratch in 122 days. With roughly 200,000 H100/H200s and ~30,000 GB200 NVL72, it remains, as of today, the largest fully operational single-coherent cluster (setting aside Google’s multi-datacenter training approach). Even so, Colossus 1’s ~300 MW looks modest next...