· Jun 11, 2026
Epoch AI describes the frontier of computing capacity in a single data center
Markdown · JSON · Original source
’s page, published June 11, 2026, describes a graph of computing capacity estimates for AI data centers over . It defines computing capacity as the total peak per second (FLOP/s) of AI chips in a data center, expressed as the equivalent number of chips required. Epoch AI says it focuses on computing capacity rather than power capacity because it considers computing capacity more relevant to AI capabilities. The page presents the frontier trend as an indication of the limits of what the AI industry is willing and able to build in a single facility. It says that this frontier is dominated by Google, Microsoft, Amazon, Meta and . The data comes from Epoch AI’s Frontier Data Centers dataset, which contains computing capacity estimates for each data center over time. The page links to code for the analysis and names assumptions and limitations as a section, but the retrieved text does not give their details. The estimates describe capacity, not demonstrated AI capabilities; how the estimates and assumptions affect the frontier trend is not specified in the retrieved text.
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: Jun 11, 2026 (2026-06-11T00:00:00.000Z).
Date basis: Date supplied by the publisher’s feed.
Added to Escape Meridian: 2026-11-06T16:08:23.623Z. First observed: 2026-10-18T03:17:22.922Z. Retrieved: 2026-10-23T23:53:37.553Z.
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 June 11, 2026. The retrieved text does not provide the graph’s values or details of the assumptions and limitations it names. The page describes estimated computing capacity; the retrieved text does not establish resulting AI capabilities.
Open question
What assumptions and limitations shape the data center capacity estimates and the frontier trend?
Source-stated
Epoch AI says the dataset includes estimates of computing capacity for each data center over time.
Data comes from our Frontier Data Centers dataset, which includes computing capacity estimates for each data center over time.
Source-stated
Epoch AI defines computing capacity as the total peak floating point operations per second of AI chips in a data center, expressed as the equivalent number of NVIDIA H100 chips required.
Computing capacity is the total peak floating point operations per second (FLOP/s) of AI chips in the data center. We express this as the equivalent number of NVIDIA H100 chips required.
Source-stated
Epoch AI says computing capacity is more relevant to AI capabilities than power capacity.
We focus on computing capacity rather than power capacity because it is more relevant to AI capabilities.
Source-stated
Epoch AI says the frontier trend indicates the limits of what the AI industry is willing and able to build in a single facility.
The frontier trend in computing capacity indicates the limits of what the AI industry is willing and able to build in a single facility.
Source-stated
Epoch AI says the frontier is dominated by Google, Microsoft, Amazon, Meta and SpaceXAI.
The data shows that this frontier is dominated by hyperscalers (Google, Microsoft, Amazon, Meta) along with SpaceXAI.
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.
Floating-point operations
A count of basic arithmetic steps; used to measure how much computation went into training a model.
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.
Epoch AI
Epoch AI studies trends in AI capability, training resources and the infrastructure behind them.
TIME
TIME is a news magazine publishing reporting on world affairs, business, technology and the workers behind digital services.
xAI
xAI develops the Grok family of AI models and publishes its own release claims and evaluations.
Editorial provenance
Version 1. Publication ID: 5215c380-cdf0-d83f-3f21-d9c9b3ce0d01.
Prepared by openai/gpt-6-luna; grounding review by anthropic/claude-opus-5.5.
Source check: passed by anthropic/claude-opus-5.5 on 2026-11-06T16:03:25.042Z. This checks fidelity to the source, not whether its claims are independently true.
Editorial selection: accept; reasons: bridge_admitted_by_model.
Observation: 07798946aecbcca8786a2e915f448f6ab884dce87f9d89d59ff86c912fa95f59. Source text hash: c5e9efba263568b4e02642304c8cd008ce6192f0181ba603292919b0cdb1a607.
Revision and source-check history
Version 1 · 2026-11-06T16:08:23.623Z
Epoch AI describes the frontier of computing capacity in a single data center
’s page, published June 11, 2026, describes a graph of computing capacity estimates for AI data centers over . It defines computing capacity as the total peak per second (FLOP/s) of AI chips in a data center, expressed as the equivalent number of chips required. Epoch AI says it focuses on computing capacity rather than power capacity because it considers computing capacity more relevant to AI capabilities. The page presents the frontier trend as an indication of the limits of what the AI industry is willing and able to build in a single facility. It says that this frontier is dominated by Google, Microsoft, Amazon, Meta and . The data comes from Epoch AI’s Frontier Data Centers dataset, which contains computing capacity estimates for each data center over time. The page links to code for the analysis and names assumptions and limitations as a section, but the retrieved text does not give their details. The estimates describe capacity, not demonstrated AI capabilities; how the estimates and assumptions affect the frontier trend is not specified in the retrieved text.
Original source
· Article
Source details
Original source date: Jun 11, 2026 (2026-06-11T00:00:00.000Z).
Date basis: Date supplied by the publisher’s feed.
Added to Escape Meridian: 2026-11-06T16:08:23.623Z. First observed: 2026-10-18T03:17:22.922Z. Retrieved: 2026-10-23T23:53:37.553Z.
Limits
Text as retrieved on October 23, 2026; originally published June 11, 2026. The retrieved text does not provide the graph’s values or details of the assumptions and limitations it names. The page describes estimated computing capacity; the retrieved text does not establish resulting AI capabilities.
Open question
What assumptions and limitations shape the data center capacity estimates and the frontier trend?
Source-stated
Epoch AI says the dataset includes estimates of computing capacity for each data center over time.
Data comes from our Frontier Data Centers dataset, which includes computing capacity estimates for each data center over time.
Source-stated
Epoch AI defines computing capacity as the total peak floating point operations per second of AI chips in a data center, expressed as the equivalent number of NVIDIA H100 chips required.
Computing capacity is the total peak floating point operations per second (FLOP/s) of AI chips in the data center. We express this as the equivalent number of NVIDIA H100 chips required.
Source-stated
Epoch AI says computing capacity is more relevant to AI capabilities than power capacity.
We focus on computing capacity rather than power capacity because it is more relevant to AI capabilities.
Source-stated
Epoch AI says the frontier trend indicates the limits of what the AI industry is willing and able to build in a single facility.
The frontier trend in computing capacity indicates the limits of what the AI industry is willing and able to build in a single facility.
Source-stated
Epoch AI says the frontier is dominated by Google, Microsoft, Amazon, Meta and SpaceXAI.
The data shows that this frontier is dominated by hyperscalers (Google, Microsoft, Amazon, Meta) along with SpaceXAI.
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.
Floating-point operations
A count of basic arithmetic steps; used to measure how much computation went into training a model.
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.
Epoch AI
Epoch AI studies trends in AI capability, training resources and the infrastructure behind them.
TIME
TIME is a news magazine publishing reporting on world affairs, business, technology and the workers behind digital services.
xAI
xAI develops the Grok family of AI models and publishes its own release claims and evaluations.
Editorial provenance
Version 1. Publication ID: 5215c380-cdf0-d83f-3f21-d9c9b3ce0d01.
Prepared by openai/gpt-6-luna; grounding review by anthropic/claude-opus-5.5.
Source check: passed by anthropic/claude-opus-5.5 on 2026-11-06T16:03:25.042Z. This checks fidelity to the source, not whether its claims are independently true.
Editorial selection: accept; reasons: bridge_admitted_by_model.
Observation: 07798946aecbcca8786a2e915f448f6ab884dce87f9d89d59ff86c912fa95f59. Source text hash: c5e9efba263568b4e02642304c8cd008ce6192f0181ba603292919b0cdb1a607.
passed · 2026-11-06T16:03:25.042Z · anthropic/claude-opus-5.5
Publication: 5215c380-cdf0-d83f-3f21-d9c9b3ce0d01
attribution — pass: The reading attributes the definitions, the focus on computing capacity, the frontier interpretation and the list of dominant firms to Epoch AI. This is one data-insight page with its own explanation, not a listing. Dropping the word 'hyperscalers' does not change who Epoch says dominates the frontier.
The data shows that this frontier is dominated by hyperscalers (Google, Microsoft, Amazon, Meta) along with SpaceXAI.
summary — pass: The reading explains in plain terms what computing capacity means, why Epoch prefers it to power capacity, and what the frontier trend indicates. FLOP/s and H100 are covered by the definitions. The page title's doubling-every-7-months figure does not appear in the retrieved text, so the reading correctly does not assert it. The question follows from the unspecified assumptions section.
The frontier trend in computing capacity indicates the limits of what the AI industry is willing and able to build in a single facility.
date — pass: originalPublished is the instant 2026-06-11. The reading states June 11, 2026, the matching UTC day, which is consistent with the CSV update date in the text. The evidence is a publisher page, not an archive copy. There is no hindsight framing, and no forecast is presented as an outcome.
CSV, Updated Jun. 11, 2026
support — pass: Each claim quotes the source verbatim. The caution that capacity estimates are not demonstrated AI capabilities is framed as a limit of the evidence, not a new fact. The note that the assumptions-and-limitations details are not retrieved is accurate, since only the section heading appears in the text. No benefits or costs are invented.
We focus on computing capacity rather than power capacity because it is more relevant to AI capabilities.
media — pass: No media was selected, and no verified media candidates were supplied, so no media identity or relevance issue arises.
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 AITIME is a news magazine publishing reporting on world affairs, business, technology and the workers behind digital services.
Explore TIMEA count of basic arithmetic steps; used to measure how much computation went into training a model.
Definition referenceAn NVIDIA graphics processing unit (GPU) built on the company's Hopper architecture and designed to accelerate AI workloads, including large language models.
Definition referencexAI develops the Grok family of AI models and publishes its own release claims and evaluations.
Explore xAIEpoch 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 AITIME is a news magazine publishing reporting on world affairs, business, technology and the workers behind digital services.
Explore TIMEA count of basic arithmetic steps; used to measure how much computation went into training a model.
Definition referenceAn NVIDIA graphics processing unit (GPU) built on the company's Hopper architecture and designed to accelerate AI workloads, including large language models.
Definition referencexAI develops the Grok family of AI models and publishes its own release claims and evaluations.
Explore xAIEpoch AI studies trends in AI capability, training resources and the infrastructure behind them.
Explore Epoch AI