Meta details Prometheus AI supercluster, scaling plans and training reliability

Meta published an account of Prometheus, its gigawatt-scale AI supercluster in New Albany, Ohio, whose first phase is complete and has trained the Muse family of models, including Muse Wonder. The company says it is scaling the site toward roughly 1 million GPUs and describes its hardware, networking and closed-loop liquid cooling. Meta reports more than 85% goodput for Muse Wonder pre-training and more than 90% for reinforcement learning, with interruptions averaging about 47,000 GPU hours apart; firmware fixes have raised that figure past 54,000 GPU hours.

AI-written summary, checked against the cited sources. How summaries work

Image: Meta Engineering

Why it matters

AI infrastructure operators can use Meta’s reported goodput and interruption figures as concrete reliability targets when planning large-scale training clusters.

Primary source

Inside Prometheus: Meta’s Journey to a Gigawatt-Scale Supercluster Meta Engineering

What we know

Confirmed by several sources

  • Meta says Prometheus is a gigawatt-scale AI supercluster in New Albany, Ohio, and its first phase is complete. — StorageReview, Hacker News
  • The first phase was used to train Meta’s Muse models, including Muse Wonder, and the project is scaling toward roughly 1 million GPUs. — StorageReview, Hacker News

Reported by one source

  • Meta reports more than 85% goodput on Muse Wonder pre-training and more than 90% on reinforcement learning, with mean time between interruptions of about 47,000 GPU hours, later exceeding 54,000 after firmware fixes. — StorageReview
  • The facility uses closed-loop liquid cooling. — Hacker News
  • Meta’s target is two interruptions per 8,000 GPUs per day, or roughly 100,000 GPU hours between interruptions. — StorageReview

Coverage 4 publishers

  1. Meta EngineeringFirst reportPrimary source

    Inside Prometheus: Meta’s Journey to a Gigawatt-Scale Supercluster

    With the recent release of the Muse family of foundational models, we want to take you inside the infrastructure that has made it all possible. The first phase of Prometheus, our gigawatt-scale supercluster, is already complete and we’re…

  2. Techmeme

    Inside Prometheus: Meta’s Journey to a Gigawatt-Scale Supercluster

    With the recent release of the Muse family of foundational models, we want to take you inside the infrastructure that has made it all possible. The first phase of Prometheus, our gigawatt-scale supercluster, is already complete and we’re…

  3. Hacker News

    Inside Prometheus: Meta’s Journey to a Gigawatt-Scale Supercluster

    With the recent release of the Muse family of foundational models, we want to take you inside the infrastructure that has made it all possible. The first phase of Prometheus, our gigawatt-scale supercluster, is already complete and we’re…

  4. StorageReview

    Meta Prometheus: Inside the Gigawatt Supercluster Scaling to 1 Million GPUs on GB200 and GB300 NVL72

    Meta published a detailed look inside Prometheus today, the gigawatt-scale AI supercluster it is building in New Albany, Ohio. The first phase is complete and has trained the Muse family of foundation models, including the Muse Wonder…

Earliest report first. Articles stay on their publishers’ sites; each link opens the original.