Research noteResearched 20 Oct 2026 · published 20 Oct 2026, 13:27 IST

Research note: Meta details Prometheus, its Ohio AI supercluster heading for about 1 million GPUs

How we checked Meta’s Prometheus figures and which ones are Meta’s own measurements.

Research led by Mayank Sahu

Compiled with AI assistance (Claude by Anthropic)

Read the story →

At a glance

2searches run

3sources read

2sources used

6facts labelled

Mixedevidence strength · 3 of 5 facts (60%)

How we rate evidence: Some facts are confirmed, others rest on one report or on the company’s own claims. We count the share of known facts labelled Confirmed or Reported: 70% or more is Strong, 40–69% Mixed, below 40% Limited.

From sources to story

How many sources we read, how many we used, and how many are official or archived.

Read3Every source we opened

Used2Relied on for the story

Cited2Linked in the story

Official1From the company or a public body

Archived2Saved on the Internet Archive

How sure we are, fact by fact

321

Confirmed · 3 — An official source says it: lab docs, pricing page, notice.
Company claim · 2 — Only the company says it. Not checked by anyone else yet.
Not known · 1 — Not announced or not answered yet.

Evidence map

Each row is a fact from the story; each column is a source. A filled dot means that source supports the fact.

Fact	Label	1OFF	2OUT
Prometheus is in New Albany, Ohio, and trained Meta’s Muse models.	Confirmed
It has hundreds of thousands of GPUs and is being scaled toward roughly 1 million.	Confirmed
It uses Nvidia GB200 and GB300 NVL72 racks, with Vera Rubin NVL72 coming.	Confirmed
Meta reports goodput above 85% for pre-training and above 90% for RL runs.	Company claim
Interruptions fell 99% from a mid-2025 peak, per Meta.	Company claim
The site’s exact power capacity.	Not known

Question

What we set out to check

Where Prometheus is, how big it is, what hardware it uses and how reliable Meta says it is.

Findings

What we found

Meta’s post gives the location, GPU scale, rack types, network and reliability numbers.
StorageReview repeats the same figures and adds that one 32,768-GPU run used 523 racks including spares.
Meta does not give a power figure in watts, only “gigawatt-scale”.

Discrepancies

Numbers that differ

Meta first reports a mean time between interruptions of about 47,000 GPU hours, later above 54,000. Both are in the post; we use the later figure.

Method

Labels we chose and why

Confirmed for location, scale and hardware from Meta’s post.
Claim for goodput and interruption figures, which are Meta’s own measurements.

Open questions

What we could not verify

The site’s exact power capacity.
Independent checks of the reliability figures.

Can you help confirm one of these? Send us a source. When we find an answer, we update this note and the story, with the date.

Next steps

How we could improve this

Look for local utility or planning filings that give the site’s power capacity.

Research timeline

Research20 Oct 2026
Research note published20 Oct 2026, 13:27 IST
Story published20 Oct 2026, 13:37 IST

What we searched

Meta Prometheus supercluster New Albany Ohio gigawatt October 2026
Meta "Prometheus" supercluster engineering blog 1 million GPUs Hyperion

Every source we read

Source	Used?	Why
Engineering at Meta Archived copy	USED	Primary source.
StorageReview Archived copy	USED	Independent write-up that matches the figures.
Converge Digest	NOT USED	Similar summary.

Sources cited in the story

Research notes show our working. They are not peer-reviewed papers. Data: facts as JSON · How we report