Colossus — a supercomputer built in 122 days

To train frontier AI, you need staggering amounts of computing power — and xAI built it at a pace the industry didn't think possible. In a converted factory in Memphis, xAI stood up "Colossus," installing 100,000 Nvidia H100 GPUs in just 122 days. Nvidia's own leadership called the timeline unprecedented; such buildouts normally take years.

xAI then kept expanding it aggressively, more than doubling the cluster and pushing toward a stated goal of a million GPUs — among the largest concentrations of AI compute anywhere on Earth, drawing on the order of gigawatts of power.

That raw execution speed is the engine behind Grok's rapid progress. Building world-class AI infrastructure this fast — power, cooling, networking and hardware integration at massive scale — is itself a formidable engineering and logistics achievement, and a big part of why a company founded in 2023 could reach the AI frontier so quickly.