# Meta brings Prometheus supercluster online in Ohio to train Muse models

Digest AI · Hardware & Compute · published 2026-10-19T11:18:00Z · updated 2026-10-19T18:49:06Z

Canonical: https://digestai.news/story/meta-brings-prometheus-supercluster-online-in-ohio-to-train-muse-model

## Summary

Meta has activated the first phase of its Prometheus AI supercluster at the New Albany, Ohio data‑center campus. The facility, which went live in October 2026, is built to operate at gigawatt scale and currently contains hundreds of thousands of GPUs, with a long‑term goal of roughly one million GPUs and a total capacity of 1 GW. Prometheus will primarily train Meta’s Muse family of foundational models, including the Muse Wonder series that the company says deliver industry‑leading performance on personal and professional tasks.

Meta’s engineering blog notes more than 85% goodput during pre‑training runs and over 90% during reinforcement‑learning runs, with a mean time between interruptions of about 47,000 GPU hours. The cluster relies on a Backend Aggregation network (BAG) linking GPUs across multiple buildings via high‑capacity Ethernet and a closed‑loop liquid‑cooling system to improve efficiency and reliability. Meta has invested over $7.8 million in local grants and sponsorships and supports hundreds of jobs in the region. Lessons from Prometheus are expected to guide Hyperion, a follow‑on supercluster planned for Louisiana that could reach up to 5 GW of capacity.

## Key points

- Meta's Prometheus supercluster in Ohio is operational, housing hundreds of thousands of GPUs with a target of one million GPUs and 1 GW capacity.
- Meta reports over 85% goodput in pre‑training runs and over 90% in reinforcement‑learning runs, with a mean time between interruptions of about 47,000 GPU hours.
- Prometheus will inform the design of Hyperion, a planned Louisiana supercluster aiming for up to 5 GW capacity.

## Why it matters

Large‑scale compute infrastructure like Prometheus accelerates training of foundational models, influencing AI performance and competition, while its open‑hardware approach could shape industry standards for future superclusters.

## Sources

1. [Meta brings Prometheus supercluster online in Ohio to train its Muse AI models](https://cryptobriefing.com/meta-prometheus-supercluster-ohio-ai-training) (cryptobriefing.com, 2026-10-19)
2. [Meta Details Prometheus Supercluster Scaling Toward 1 Million GPUs](https://unite.ai/meta-details-prometheus-supercluster-scaling-toward-1-million-gpus) (Unite.AI, 2026-10-19)

Part of the developing story: [Meta's Massive AI Gamble and Tax Strategy](https://digestai.news/thread/meta-aggressively-claims-tax-credit-for-ai-data-center-build-outs) (5 stories)

## Cite

Digest AI, "Meta brings Prometheus supercluster online in Ohio to train Muse models", 19 October 2026, https://digestai.news/story/meta-brings-prometheus-supercluster-online-in-ohio-to-train-muse-model

---

Written by Digest AI's editorial model from the linked sources; the sources are the record. Headlines, digests and key points are written by Digest AI and may be quoted with a link to the story page. Linked articles belong to their publishers. Terms: https://digestai.news/terms#reuse
JSON: https://digestai.news/story/meta-brings-prometheus-supercluster-online-in-ohio-to-train-muse-model.json