xAI Built a 150,000-GPU Supercomputer in Four Months, Not 24

xAI completed its Colossus supercomputer in four months—six times faster than the 24-month industry baseline for infrastructure of this scale. The acceleration came through leveraging existing data center facilities and established supply chain relationships, sidesteping the facility prep and procurement delays that typically dominate large-scale supercomputer deployments. Colossus currently runs jobs across 150,000 GPUs at 99% uptime, with plans to scale to 1 million.