{
    "format": "anawords-story/1",
    "docs": "https://anawords.com/data",
    "url": "https://anawords.com/story/meta-prometheus-supercluster-ohio-1-million-gpus",
    "title": "Meta details Prometheus, its Ohio AI supercluster heading for about 1 million GPUs",
    "summary": "Meta’s engineers say the New Albany site trained its Muse models, now runs hundreds of thousands of GPUs, and cut training interruptions by 99% from a 2025 peak.",
    "section": "Chips & Compute",
    "type": "report",
    "published": "2026-10-20T13:37:16+05:30",
    "updated": "2026-10-20T13:37:16+05:30",
    "version": 1,
    "author": {
        "name": "Mayank Sahu",
        "url": "https://anawords.com/author/mayank-sahu"
    },
    "how_checked": "Sources checked by AI; published under the editor’s standing approval. Not read line by line by a person before publication.",
    "ai_assisted": true,
    "research_note": "https://anawords.com/research/meta-prometheus-supercluster-ohio-1-million-gpus",
    "labels": {
        "confirmed": {
            "name": "Confirmed",
            "meaning": "An official source says it: lab docs, pricing page, notice."
        },
        "reported": {
            "name": "Reported",
            "meaning": "Two or more independent, named outlets say it."
        },
        "claim": {
            "name": "Company claim",
            "meaning": "Only the company says it. Not checked by anyone else yet."
        },
        "unconfirmed": {
            "name": "Unconfirmed",
            "meaning": "One report only, shown with its source."
        },
        "unknown": {
            "name": "Not known",
            "meaning": "Not announced or not answered yet."
        }
    },
    "facts": [
        {
            "id": "f1",
            "text": "Prometheus is in New Albany, Ohio, and trained Meta’s Muse models.",
            "label": "confirmed",
            "label_name": "Confirmed",
            "key_fact": true,
            "sources": [
                1
            ],
            "outcome": null
        },
        {
            "id": "f2",
            "text": "It has hundreds of thousands of GPUs and is being scaled toward roughly 1 million.",
            "label": "confirmed",
            "label_name": "Confirmed",
            "key_fact": true,
            "sources": [
                1
            ],
            "outcome": null
        },
        {
            "id": "f3",
            "text": "It uses Nvidia GB200 and GB300 NVL72 racks, with Vera Rubin NVL72 coming.",
            "label": "confirmed",
            "label_name": "Confirmed",
            "key_fact": false,
            "sources": [
                1
            ],
            "outcome": null
        },
        {
            "id": "f4",
            "text": "Meta reports goodput above 85% for pre-training and above 90% for RL runs.",
            "label": "claim",
            "label_name": "Company claim",
            "key_fact": false,
            "sources": [
                1,
                2
            ],
            "outcome": null
        },
        {
            "id": "f5",
            "text": "Interruptions fell 99% from a mid-2025 peak, per Meta.",
            "label": "claim",
            "label_name": "Company claim",
            "key_fact": true,
            "sources": [
                1,
                2
            ],
            "outcome": null
        },
        {
            "id": "f6",
            "text": "The site’s exact power capacity.",
            "label": "unknown",
            "label_name": "Not known",
            "key_fact": false,
            "sources": [],
            "outcome": null
        }
    ],
    "sources": [
        {
            "n": 1,
            "url": "https://engineering.fb.com/2026/10/19/data-center-engineering/inside-prometheus-meta-gigawatt-scale-supercluster/",
            "title": "Engineering at Meta: Inside Prometheus, Meta’s journey to a gigawatt-scale supercluster",
            "type": "official",
            "checked": "20 Oct 2026",
            "archive": {
                "url": "https://web.archive.org/web/20261020042210/https://engineering.fb.com/2026/10/19/data-center-engineering/inside-prometheus-meta-gigawatt-scale-supercluster/",
                "archived": "2026-10-20T09:52:10+05:30"
            },
            "original_gone": false
        },
        {
            "n": 2,
            "url": "https://www.storagereview.com/news/meta-prometheus-gigawatt-supercluster-1-million-gpus-gb200-gb300-nvl72",
            "title": "StorageReview: Meta Prometheus, inside the gigawatt supercluster scaling to 1 million GPUs",
            "type": "outlet",
            "checked": "20 Oct 2026",
            "archive": {
                "url": "https://web.archive.org/web/20261020081048/https://www.storagereview.com/news/meta-prometheus-gigawatt-supercluster-1-million-gpus-gb200-gb300-nvl72",
                "archived": "2026-10-20T13:40:48+05:30"
            },
            "original_gone": false
        }
    ],
    "promises": [],
    "prices": [],
    "incidents": [],
    "image": {
        "url": "https://anawords.com/uploads/2026/10/meta-prometheus-supercluster-ohio-1-million-gpus-16x9.webp",
        "kind": "illustration",
        "note": "Anawords illustration made by software from facts in the story. Not a photograph."
    },
    "changes": [
        {
            "version": 1,
            "date": "2026-10-20T13:37:16+05:30",
            "correction": false,
            "note": "First published. Research note published first. Sources checked by AI; approved for publication by the editor."
        }
    ],
    "license": {
        "facts_and_data": "Creative Commons Attribution 4.0 (CC BY 4.0)",
        "url": "https://creativecommons.org/licenses/by/4.0/",
        "article_text": "© Anawords. Quote with a link.",
        "attribution": "Anawords, https://anawords.com/story/meta-prometheus-supercluster-ohio-1-million-gpus"
    }
}