Inside Prometheus: Meta's Journey to a Gigawatt-Scale Supercluster
Meta turned its New Albany, Ohio data center site from web traffic into Prometheus, a gigawatt-scale AI supercluster spanning multiple interconnected buildings that behaves as one machine. It runs NVL72 racks (Catalina on GB200 with 800G of scale-out per GPU, Clemente on GB300 in a single rack) on a dedicated Ethernet back-end network for GPU-to-GPU traffic. Reliability work pushed pre-training goodput above 85% and mean time between interruptions past 54K GPU hours, and the cluster is growing toward roughly 1 million GPUs.
- Published
- 2026-10-19
- Scale
- gigawatt-scale supercluster, hundreds of thousands of GPUs today, ~1 million planned, 85%+ pre-training goodput, 90%+ RL goodput, MTBI raised from ~47K to 54K+ GPU hours
- Techniques
- GPU clusters distributed training hardware reliability networking liquid cooling
- Stack
- NVIDIA GB200 NVL72 NVIDIA GB300 NVL72 Ethernet back-end network liquid cooling