Independent newsroom

Tuesday, October 20, 2026

The File

Every claim traces to a primary source. Uncertainty is labeled. Corrections are public.

AI · Tuesday, October 20, 2026

Meta describes the giant Ohio computer training its newest AI

The File desk · Oct. 20, 2026, 6:44 a.m. ET

Partially confirmed1 primary sources

Status

Partially confirmed. Meta posted the account on Monday, and this desk re-opened the page. The size and reliability figures are the company's claims. No outside measurement was found.
  • Confirmed
    Meta published a description of Prometheus, at its data center in New Albany, Ohio, and said the first phase is complete.
  • Partially confirmed
    The chip counts and the reliability percentages are Meta's. This desk did not measure them.
Meta on Monday described Prometheus, the computer cluster at its data center in New Albany, Ohio. The company says the first phase is finished and was used to train its Muse models, including Muse Wonder. The size figures are Meta's.1
Meta wrote, “The first phase of Prometheus, our gigawatt-scale supercluster, is already complete and we're working toward scaling it to comprise ~1 million GPUs.” It also wrote, “Today, Prometheus comprises an array of hundreds of thousands of GPUs across multiple interconnected buildings.” Those counts are the company's. The page gives no dollar cost and no energy total, and this story doesn't compute one.1
Meta says the cluster ran at “more than 85% goodput for pre-training runs and more than 90% goodput for reinforcement learning (RL) runs of Muse Wonder.” That, too, is a company claim.
Why it matters (Estimate). Training the newest models takes buildings full of chips and a city's worth of electricity. Meta is telling the public how big this one machine is and that it already trained Muse Wonder on it. The numbers are the company's, and the power bill isn't in the post.1
“Two years ago, New Albany was a conventional data center site running web traffic.” The engineering post is timed at 9:00 a.m. ET.
How we know

Notes

  1. 1.Engineering at Meta, “Inside Prometheus: Meta's Journey to a Gigawatt-Scale Supercluster,” Oct. 19, 2026, re-opened for this edition. https://engineering.fb.com/2026/10/19/data-center-engineering/inside-prometheus-meta-gigawatt-scale-supercluster/ datePublished 13:00 UTC, which is 9:00 a.m. ET. “The first phase of Prometheus, our gigawatt-scale supercluster, is already complete and we're working toward scaling it to comprise ~1 million GPUs.” “Today, Prometheus comprises an array of hundreds of thousands of GPUs across multiple interconnected buildings.” “more than 85% goodput for pre-training runs and more than 90% goodput for reinforcement learning (RL) runs of Muse Wonder.” “Two years ago, New Albany was a conventional data center site running web traffic.” Company post: https://x.com/AIatMeta/status/2112245968750145561 at 2:14 p.m. ET. No dollar figure and no megawatt-hour figure are on the page. A separate Meta technical report on an on-device model isn't a launch and isn't covered here.

What is still unknown or disputed

  • Power use and cost are not stated.
  • No date is given for reaching about 1 million chips.
  • No outside audit of the goodput figures was found.

Primary sources

Every claim in this story is drawn from the documents below. If a fetch failed, that is recorded on the card.

  1. Source 1

    Inside Prometheus: Meta's journey to a gigawatt-scale supercluster

    Engineering at Meta · October 19, 2026

    The first phase of Prometheus, our gigawatt-scale supercluster, is already complete and we're working toward scaling it to comprise ~1 million GPUs.

    https://engineering.fb.com/2026/10/19/data-center-engineering/inside-prometheus-meta-gigawatt-scale-supercluster/

Ask this story

Answers come only from this story and its sources. Free test: limited questions per day.

0/300