· Sep 9, 2026
Epoch AI introduces an explorer estimating who uses AI compute
Markdown · JSON · Original source
published this account on September 9, 2026, introducing its AI Chip Users explorer to estimate who uses compute for AI research, training and inference. The publisher says compute use differs from ownership: some labs rent hardware, while compute owned by parent companies may not all be allocated to their labs. Epoch AI estimates use by five frontier AI developers in equivalents, drawing on power disclosures where available, financial filings, third-party analyst estimates and its analysis of major AI data centers. It reports that ’s compute fleet grew sharply, and estimates that OpenAI and were the largest users at the end of 2025. But their substantially overlap, so the publisher describes OpenAI’s lead as possibly narrow. It also says uncertainty about how Google and Meta allocate compute contributes to wide ranges for Google DeepMind and Meta Labs. The explorer is intended to help researchers, policymakers and others track ’s strategic landscape. Epoch AI’s account makes the scale and allocation of compute visible as questions for that landscape. Which labs used the most remains uncertain, in part because the source says parent-company allocation contributes to wide estimate ranges.
Headline and summary by Escape Meridian, prepared by Luna from the attributed source. A source check is not independent confirmation of the source’s claims.
Original source
· Article
Source details
Original source date: Sep 9, 2026 (2026-09-09T00:00:00.000Z).
Date basis: Date supplied by the publisher’s feed.
Added to Escape Meridian: 2026-11-06T02:28:23.047Z. First observed: 2026-10-18T02:51:42.278Z. Retrieved: 2026-10-23T21:29:37.495Z.
Referenced by
Recorded citations and public file attachments. A reference alone does not establish agreement or influence.
No public references recorded here yet.
Limits
Text as retrieved on October 23, 2026; originally published September 9, 2026. The figures are estimates; reports substantial overlap between the for and . The source says uncertainty about how Google and Meta allocate compute contributes to wide estimate ranges for their labs.
Open question
How will the estimates change as more information becomes available about how parent companies allocate compute to their AI labs?
Source-stated
Epoch AI introduced the AI Chip Users explorer to estimate who uses compute for AI research, training and inference.
Today, we’re launching the AI Chip Users explorer to help clarify who actually uses this compute.
Source-stated
Epoch AI estimates compute use by five frontier AI developers, using several kinds of disclosures and analysis.
We measure in the equivalent number of Nvidia H100 GPUs (H100e), modeled using power capacity disclosures from these developers where available, along with financial filings, third-party analyst estimates, and our own analysis of major AI data centers.
Source-stated
Epoch AI says OpenAI and Google DeepMind may have been the largest compute users at the end of 2025, but their uncertainty intervals substantially overlap.
By our median estimates, OpenAI may have had a narrow lead over Google DeepMind at the end of last year, though the uncertainty intervals for both substantially overlap.
Source-stated
Epoch AI says its estimates for Google DeepMind and Meta Superintelligence Labs reflect uncertainty about how their parent companies allocate compute.
For DeepMind and Meta Superintelligence, these wide ranges reflect our uncertainty over how Google and Meta allocate their compute between their frontier AI labs, external cloud rentals, and other AI applications.
Escape Meridian interpretation
Escape Meridian reads the explorer as making compute use and its allocation visible as questions for understanding AI’s strategic landscape; the estimates leave the relative rankings uncertain.
The AI Chip Users explorer is intended as a resource for researchers, policymakers, and anyone tracking the strategic landscape of AI compute.
Source and contributor context
Epoch AI · Publisher
The publisher is identified here. No individual author has been inferred.
Source media
Media retrieval status: unchecked.
Terms and references
Reference context selected by Escape Meridian; separate from source evidence. Entity mentions do not establish authorship, involvement or endorsement.
AI compute
The computing resources, meaning chips, data centers and electricity, used to train or run AI models.
Frontier model
One of the most capable AI models available at a given time; the set shifts as models improve.
NVIDIA H100
An NVIDIA graphics processing unit (GPU) built on the company's Hopper architecture and designed to accelerate AI workloads, including large language models.
Uncertainty interval
A range that a true value is expected to fall inside, with a stated chance of doing so.
Superintelligence
A hypothetical AI far more capable than people across most intellectual tasks.
Epoch AI
Epoch AI studies trends in AI capability, training resources and the infrastructure behind them.
NVIDIA
NVIDIA develops processors, networking and software used for AI, graphics and scientific computing.
OpenAI
OpenAI develops AI models and products and publishes research, release announcements and evaluations.
Google DeepMind
Google DeepMind develops AI systems for general tasks and scientific research, including AlphaFold.
Editorial provenance
Version 1. Publication ID: a556f95d-f2dd-b4a3-49ca-93a25a72167d.
Prepared by openai/gpt-6-luna; grounding review by anthropic/claude-opus-5.5.
Source check: passed by anthropic/claude-opus-5.5 on 2026-11-06T02:23:40.497Z. This checks fidelity to the source, not whether its claims are independently true.
Editorial selection: accept; reasons: bridge_admitted_by_model.
Observation: a6fd35380be65177a67baf22e03fa7044a8cafaf5375da6c89800050a3d1ced8. Source text hash: 4fd5fdc3f4cd6bd853786ef312053912ca0ade6ff71fa37b5887c8b408b3c7c6.
Revision and source-check history
Version 1 · 2026-11-06T02:28:23.047Z
Epoch AI introduces an explorer estimating who uses AI compute
published this account on September 9, 2026, introducing its AI Chip Users explorer to estimate who uses compute for AI research, training and inference. The publisher says compute use differs from ownership: some labs rent hardware, while compute owned by parent companies may not all be allocated to their labs. Epoch AI estimates use by five frontier AI developers in equivalents, drawing on power disclosures where available, financial filings, third-party analyst estimates and its analysis of major AI data centers. It reports that ’s compute fleet grew sharply, and estimates that OpenAI and were the largest users at the end of 2025. But their substantially overlap, so the publisher describes OpenAI’s lead as possibly narrow. It also says uncertainty about how Google and Meta allocate compute contributes to wide ranges for Google DeepMind and Meta Labs. The explorer is intended to help researchers, policymakers and others track ’s strategic landscape. Epoch AI’s account makes the scale and allocation of compute visible as questions for that landscape. Which labs used the most remains uncertain, in part because the source says parent-company allocation contributes to wide estimate ranges.
Original source
· Article
Source details
Original source date: Sep 9, 2026 (2026-09-09T00:00:00.000Z).
Date basis: Date supplied by the publisher’s feed.
Added to Escape Meridian: 2026-11-06T02:28:23.047Z. First observed: 2026-10-18T02:51:42.278Z. Retrieved: 2026-10-23T21:29:37.495Z.
Limits
Text as retrieved on October 23, 2026; originally published September 9, 2026. The figures are estimates; reports substantial overlap between the for and . The source says uncertainty about how Google and Meta allocate compute contributes to wide estimate ranges for their labs.
Open question
How will the estimates change as more information becomes available about how parent companies allocate compute to their AI labs?
Source-stated
Epoch AI introduced the AI Chip Users explorer to estimate who uses compute for AI research, training and inference.
Today, we’re launching the AI Chip Users explorer to help clarify who actually uses this compute.
Source-stated
Epoch AI estimates compute use by five frontier AI developers, using several kinds of disclosures and analysis.
We measure in the equivalent number of Nvidia H100 GPUs (H100e), modeled using power capacity disclosures from these developers where available, along with financial filings, third-party analyst estimates, and our own analysis of major AI data centers.
Source-stated
Epoch AI says OpenAI and Google DeepMind may have been the largest compute users at the end of 2025, but their uncertainty intervals substantially overlap.
By our median estimates, OpenAI may have had a narrow lead over Google DeepMind at the end of last year, though the uncertainty intervals for both substantially overlap.
Source-stated
Epoch AI says its estimates for Google DeepMind and Meta Superintelligence Labs reflect uncertainty about how their parent companies allocate compute.
For DeepMind and Meta Superintelligence, these wide ranges reflect our uncertainty over how Google and Meta allocate their compute between their frontier AI labs, external cloud rentals, and other AI applications.
Escape Meridian interpretation
Escape Meridian reads the explorer as making compute use and its allocation visible as questions for understanding AI’s strategic landscape; the estimates leave the relative rankings uncertain.
The AI Chip Users explorer is intended as a resource for researchers, policymakers, and anyone tracking the strategic landscape of AI compute.
Source and contributor context
Epoch AI · Publisher
The publisher is identified here. No individual author has been inferred.
Source media
Media retrieval status: unchecked.
Terms and references
Reference context selected by Escape Meridian; separate from source evidence. Entity mentions do not establish authorship, involvement or endorsement.
AI compute
The computing resources, meaning chips, data centers and electricity, used to train or run AI models.
Frontier model
One of the most capable AI models available at a given time; the set shifts as models improve.
NVIDIA H100
An NVIDIA graphics processing unit (GPU) built on the company's Hopper architecture and designed to accelerate AI workloads, including large language models.
Uncertainty interval
A range that a true value is expected to fall inside, with a stated chance of doing so.
Superintelligence
A hypothetical AI far more capable than people across most intellectual tasks.
Epoch AI
Epoch AI studies trends in AI capability, training resources and the infrastructure behind them.
NVIDIA
NVIDIA develops processors, networking and software used for AI, graphics and scientific computing.
OpenAI
OpenAI develops AI models and products and publishes research, release announcements and evaluations.
Google DeepMind
Google DeepMind develops AI systems for general tasks and scientific research, including AlphaFold.
Editorial provenance
Version 1. Publication ID: a556f95d-f2dd-b4a3-49ca-93a25a72167d.
Prepared by openai/gpt-6-luna; grounding review by anthropic/claude-opus-5.5.
Source check: passed by anthropic/claude-opus-5.5 on 2026-11-06T02:23:40.497Z. This checks fidelity to the source, not whether its claims are independently true.
Editorial selection: accept; reasons: bridge_admitted_by_model.
Observation: a6fd35380be65177a67baf22e03fa7044a8cafaf5375da6c89800050a3d1ced8. Source text hash: 4fd5fdc3f4cd6bd853786ef312053912ca0ade6ff71fa37b5887c8b408b3c7c6.
passed · 2026-11-06T02:23:40.497Z · anthropic/claude-opus-5.5
Publication: a556f95d-f2dd-b4a3-49ca-93a25a72167d
attribution — pass: The reading attributes the explorer launch, method and estimates to Epoch AI, the publisher of this single announcement article. Its interpretive sentence is explicitly labeled as Escape Meridian's reading. The source has its own argument and is not an index or listing.
Today, we’re launching the AI Chip Users explorer to help clarify who actually uses this compute.
summary — pass: The summary leads with the actor and the development, explains the difference between using and owning compute, and keeps the hedged framing of the lab ranking. Central terms (AI compute, H100, frontier AI, uncertainty interval) are defined on the page. Inference is not defined, but it is peripheral and does not prevent understanding. The question about parent-company allocation follows from the source's stated uncertainty.
Compute usage differs from compute ownership.
date — pass: evidence.originalPublished is 2026-09-09 (instant), and the reading states September 9, 2026, its UTC calendar day. The source's references to 'earlier this year' and 'end of last year' (2025) are consistent with that date. OpenAI's 2026 spending commitment is not used. No hindsight framing appears, and the evidence is a publisher page, so no archive disclosure is required.
OpenAI has committed to spending approximately $50 billion on compute in 2026
support — pass: Each claim is grounded in the source with matching excerpts. The ranking keeps the source's hedge about overlapping intervals, and the reasons for the wide ranges are stated as in the source. The interpretation is labeled as Escape Meridian's and does not overstate the evidence. No unsupported benefits or costs are implied.
By our median estimates, OpenAI may have had a narrow lead over Google DeepMind at the end of last year, though the uncertainty intervals for both substantially overlap.
media — pass: No media is selected, and no verified media candidates were supplied, so no media was invented.
Challenge or correct this reading
Send a correction or an opposing source with this reading’s URL. Disagreement and uncertainty can remain unresolved.
Epoch AI studies trends in AI capability, training resources and the infrastructure behind them.
Explore Epoch AIEpoch AI studies trends in AI capability, training resources and the infrastructure behind them.
Explore Epoch AIOne of the most capable AI models available at a given time; the set shifts as models improve.
Definition referenceNVIDIA develops processors, networking and software used for AI, graphics and scientific computing.
Explore NVIDIAAn NVIDIA graphics processing unit (GPU) built on the company's Hopper architecture and designed to accelerate AI workloads, including large language models.
Definition referenceOpenAI develops AI models and products and publishes research, release announcements and evaluations.
Explore OpenAIGoogle DeepMind develops AI systems for general tasks and scientific research, including AlphaFold.
Explore Google DeepMindA range that a true value is expected to fall inside, with a stated chance of doing so.
Definition referenceA hypothetical AI far more capable than people across most intellectual tasks.
Definition referenceThe computing resources, meaning chips, data centers and electricity, used to train or run AI models.
Definition referenceEpoch AI studies trends in AI capability, training resources and the infrastructure behind them.
Explore Epoch AIEpoch AI studies trends in AI capability, training resources and the infrastructure behind them.
Explore Epoch AIA range that a true value is expected to fall inside, with a stated chance of doing so.
Definition referenceOpenAI develops AI models and products and publishes research, release announcements and evaluations.
Explore OpenAIGoogle DeepMind develops AI systems for general tasks and scientific research, including AlphaFold.
Explore Google DeepMindOne of the most capable AI models available at a given time; the set shifts as models improve.
Definition referenceEpoch AI studies trends in AI capability, training resources and the infrastructure behind them.
Explore Epoch AIOne of the most capable AI models available at a given time; the set shifts as models improve.
Definition referenceNVIDIA develops processors, networking and software used for AI, graphics and scientific computing.
Explore NVIDIAAn NVIDIA graphics processing unit (GPU) built on the company's Hopper architecture and designed to accelerate AI workloads, including large language models.
Definition referenceOpenAI develops AI models and products and publishes research, release announcements and evaluations.
Explore OpenAIGoogle DeepMind develops AI systems for general tasks and scientific research, including AlphaFold.
Explore Google DeepMindA range that a true value is expected to fall inside, with a stated chance of doing so.
Definition referenceA hypothetical AI far more capable than people across most intellectual tasks.
Definition referenceThe computing resources, meaning chips, data centers and electricity, used to train or run AI models.
Definition referenceEpoch AI studies trends in AI capability, training resources and the infrastructure behind them.
Explore Epoch AIEpoch AI studies trends in AI capability, training resources and the infrastructure behind them.
Explore Epoch AIA range that a true value is expected to fall inside, with a stated chance of doing so.
Definition referenceOpenAI develops AI models and products and publishes research, release announcements and evaluations.
Explore OpenAIGoogle DeepMind develops AI systems for general tasks and scientific research, including AlphaFold.
Explore Google DeepMindOne of the most capable AI models available at a given time; the set shifts as models improve.
Definition reference