DevelopingHardware

Meta details Prometheus: hundreds of thousands of GPUs in Ohio, scaling toward 1 million

Meta's engineering blog says phase one of its Prometheus supercluster in New Albany, Ohio is complete, with hundreds of thousands of GPUs across linked buildings and a path to roughly 1 million. It trained the Muse models. Meta reports over 85% pre-training goodput and a 99% cut in interruptions in about a year. Vera Rubin racks are next.

HEAT
OUTLETS: 5
NEW IN 6H: +5
FIRST SEEN: Oct 19, 2026, 14:40 ET
LAST SIGNAL: 1h ago

Primary source

Engineering at Metaengineering.fb.com/2026/10/19/data-center-engineering/inside-prometheus-meta-gigawatt-scale-supercluster/

Coverage · 5 articles, oldest first

Monday, October 19, 2026

14:40
Engineering at Meta
Meta details Prometheus, its gigawatt-scale supercluster in Ohio, which will have ~1M GPUs at full scale and was used to train its Muse foundational models
14:40
Crypto Briefing
cryptobriefing.com/meta-prometheus-supercluster-ohio-ai-training/
14:40
StorageReview.com
storagereview.com/news/meta-prometheus-gigawatt-supercluster-1-million-gpus-gb200-gb300-nvl72
15:10
Unite.AI
unite.ai/meta-details-prometheus-supercluster-scaling-toward-1-million-gpus/
16:10
NeoTeo
neoteo.com/en/meta-detailed-prometheuss-network-and-tent-construction

ShareX LinkedIn Hacker News Email

← The whole wire