AI training infrastructure

Meta has detailed Prometheus, its gigawatt-scale AI supercluster in New Albany, Ohio, which trained the Muse Wonder frontier models on roughly 1 million GPUs. The infrastructure combines NVIDIA Blackwell and Blackwell Ultra platforms with a novel fault recovery architecture that achieved a 99% reduction in training interruptions. For enterprise buyers and developers, Prometheus defines the new upper bound of AI compute concentration and shapes what frontier models will be capable of doing.