<!--
Source: https://scrutica.com/capacity
Generated: 2026-09-04T08:28:06.802Z
Format: Markdown extraction of the rendered HTML at the source URL.
For the full agent guide see: https://scrutica.com/llms-full.txt
For the MCP server see: https://scrutica.com/api/mcp
-->

# Capacity Estimator
## On 279 of 314 sites, the chip count reads higher than the power

314 facilities disclose both a chip count and a nameplate power figure: two independent routes to one number, the compute that site can run in a day. Run both and they agree within a factor of two on 294 of them. The residue is asymmetric. On 89 per cent of sites the chips imply more compute than the power figure will carry, and noise does not point one way. Either the disclosed chip counts run ahead of the power built to serve them (announced fleets, phased sites, full build-outs quoted against a first-phase substation), or the constants below understate how much of a site’s load reaches the accelerators. The estimator below settles that for a site you know.

Plate ITwo readings of one quantity314 sites · complete

1020102110221023102410251026daily flop, log scale1.5×2×3×estimates agreeestimates diverge314 facilities, ordered by cross-path divergence

from the chip countfrom the power figurethe disagreement

279within 1.5×

151.5–2×

142–3×

6beyond 3×

_Plate I._ One upright per facility, drawn between two estimates of that site’s daily FLOP budget: a filled mark for the chip-count path, an open mark for the power path.

Vertical position is the estimate itself, on a log scale; the upright’s length is the disagreement between the two. Horizontal position is RANK by that disagreement, tightest at the left, so the distance between neighbouring uprights means nothing.

Every facility disclosing both inputs is drawn: 314 of the 512 that disclose either, across 32 countries, none sampled out. That wider figure counts every facility type, chip-manufacturing sites included: a fab’s chip count and power figure can be checked against each other like any other site’s. [Thresholds](/capacity/thresholds) excludes them and states a correspondingly smaller population, because a fab’s power draw is production load and the power path would read it as training compute.

The two estimates are 1.15× apart at the median and 9.3× at the widest; the tightest pair whose accelerator the record names is NVIDIA SATURN V Volta, whose 264 chips and 0.2 MW reconcile to within 0.0 per cent on its own V100 SXM2 spec.

Each site runs the chip its record discloses, on both paths — 285 of 314 do; the rest fall back to the H100 SXM5 default. Because the two paths always share one chip per site, an upright measures how that site’s disclosed pair (chip count and megawatts) reconciles against the constants; it is not a hardware-mix artefact, and it is not a claim that either disclosure is wrong.

A skew as one-directional as this one is therefore evidence about the constants and about disclosure practice at least as much as about any individual site. Every value here is an estimate. Facility records as of 2026-08-30.

Every figure here is a back-solve from something an operator disclosed: a chip count, a nameplate power figure, a construction budget. Run more than one against the same site and the spread between them shows which assumptions carry the uncertainty. [How this is derived↗](/methodology#flop-estimation "How this is derived: facility FLOP capacity. Three independent estimation paths (hardware, power, capital cost); every assumption adjustable in-browser.")

Cite

CSVJSONJSON+Prov

PrintScreenshot

The estimator

### Estimate a site you know

Give it a chip count, a nameplate power figure, or a construction budget — whichever the operator actually published — and each path returns a daily FLOP budget with the sensitivity range its parameters imply. Every default below is adjustable, and the state of every slider is in the URL, so an estimate under a particular set of assumptions is a link you can put in a footnote.

Default-parameters derivation (Hardware path)

10,000 H100 SXM5 GPUs at the documented defaults (interconnect efficiency 0.85, MFU 0.4). The interactive calculator below hydrates on JS and exposes every parameter as an adjustable slider.

Hardware-path FLOP derivation steps at default parameters (10,000 H100 SXM5)

Step

Formula

Result

Peak FP16

gpu\_count × per\_gpu\_fp16\_tflops

9.89 × 1018 FLOP/s

Effective FP16

peak × interconnect\_efficiency (0.85)

8.41 × 1018 FLOP/s

Training Throughput

effective × mfu\_assumption (0.4)

3.36 × 1018 FLOP/s

Daily FLOP Budget

training × 86,400

2.91 × 1023 FLOP/day

At this daily budget, the EU AI Act 1025 FLOP threshold is reached in approximately 34.4 days. Adjust GPU count, model, interconnect efficiency, and MFU in the interactive calculator below.

Check

### Against runs whose compute was disclosed

Plate I reconciles two estimates against each other, which cannot tell you whether both are wrong together. These 10 frontier-class training runs (≥ 10²⁵ FLOP) are the external check: for each, a lab disclosed a hardware count or Epoch derived a central FLOP estimate, and the same three paths run against it. Where both exist they appear in one row, because the column-level disagreement is the auditable thing.

Reading the table

Lab-disclosed hardware leads when present; the hardware-path of the calculator above can be reproduced against the disclosed GPU model and count. Rows lacking lab hardware use the Epoch T2 estimate alone. Where the two sources disagree (Pangu Ultra: Huawei’s 8,192 Ascend NPU disclosure against Epoch’s 1.1e25 derived via Ascend-MFU calibration off the NVIDIA path), both rows are retained and the calculator preserves the spread. The calculator is a methodology; it measures no live run. Epoch’s “Speculative” / “Likely” / “Confident” bands express that residual uncertainty.

Model

Hardware (lab disclosure)

Epoch FLOP

Epoch confidence

Source provenance

Grok 4

xAI · Jul 2025

200,000 GPUs (class undisclosed)

via Epoch derivation

5.0e+26 FLOP

T2

Speculative

[Epoch (T2)](https://epoch.ai/data/notable-ai-models)

Grok 3

xAI · Feb 2025

NVIDIA H100 SXM5 80GB × 80,000

via Epoch derivation

3.5e+26 FLOP

T2

Likely

[Epoch (T2)](https://epoch.ai/data/notable-ai-models)

Llama 4 Behemoth (preview)

Meta Platforms, Inc. · Apr 2025

32,000 GPUs (class undisclosed)

[Lab paper / announcement](https://ai.meta.com/blog/llama-4-multimodal-intelligence/)

5.2e+25 FLOP

T2

Likely

[Lab disclosure (T4)](https://ai.meta.com/blog/llama-4-multimodal-intelligence/)[Epoch (T2)](https://epoch.ai/data/notable-ai-models)

Gemini 1.0 Ultra

Google DeepMind · Dec 2023

Google TPUv4

[Lab paper / announcement](https://arxiv.org/abs/2312.11805)

5.0e+25 FLOP

T2

Speculative

[Lab disclosure (T4)](https://arxiv.org/abs/2312.11805)[Epoch (T2)](https://epoch.ai/data/notable-ai-models)

Llama 3.1-405B

Meta Platforms, Inc. · Jul 2024

NVIDIA H100 80GB × 16,384

[Lab paper / announcement](https://arxiv.org/abs/2407.21783)

3.8e+25 FLOP

T2

Confident

[Lab disclosure (T1)](https://arxiv.org/abs/2407.21783)[Epoch (T2)](https://epoch.ai/data/notable-ai-models)

GPT-4 (Jun 2023)

OpenAI · Jun 2023

NVIDIA A100 SXM4 40 GB × 25,000

via Epoch derivation

2.1e+25 FLOP

T2

Likely

[Epoch (T2)](https://epoch.ai/data/notable-ai-models)

GPT-4 (Mar 2023)

OpenAI · Mar 2023

NVIDIA A100 SXM4 40 GB × 25,000

via Epoch derivation

2.1e+25 FLOP

T2

Likely

[Lab disclosure (T4)](https://arxiv.org/abs/2303.08774)[Epoch (T2)](https://epoch.ai/data/notable-ai-models)

Nemotron-4 340B

Nvidia · Jun 2024

NVIDIA H100 80GB SXM5 × 6,144

[Lab paper / announcement](https://arxiv.org/abs/2406.11704)

1.8e+25 FLOP

T2

Confident

[Lab disclosure (T4)](https://arxiv.org/abs/2406.11704)[Epoch (T2)](https://epoch.ai/data/notable-ai-models)

Pangu Ultra

Huawei Technologies · Apr 2025

Huawei Ascend NPUs × 8,192

[Lab paper / announcement](https://arxiv.org/abs/2504.07866)

1.1e+25 FLOP

T2

Confident

[Lab disclosure (T4)](https://arxiv.org/abs/2504.07866)[Epoch (T2)](https://epoch.ai/data/notable-ai-models)

Inflection-2

Inflection AI · Nov 2023

NVIDIA H100 × 5,000

[Lab paper / announcement](https://inflection.ai/inflection-2)

1.0e+25 FLOP

T2

Confident

[Lab disclosure (T1)](https://inflection.ai/inflection-2)[Epoch (T2)](https://epoch.ai/data/notable-ai-models)

[Source: Epoch AI](https://epoch.ai/data), CC BY 4.0 · lab-primary rows from each lab’s canonical paper / system card / announcement (URLs above). Table: `frontier_model_runs`.

Parameters

### The five constants both views run on

The sliders above move these; the Thresholds view reads the same registry, so the two views cannot drift apart on the constants. Each card shows the default, its adjustable range, the paths it touches and the source it was calibrated against. Methodology version 1.3.6, dated 2026-07-22.

Shared assumptions5 parameters · methodology v1.3.6

Interconnect EfficiencyDefault 0.85

range 0.6–0.95·Hardware / Power / Cost paths

[Epoch AI (Sevilla et al, 2022)](https://epoch.ai/blog/estimating-training-compute)

Model FLOP Utilization (MFU)Default 0.4

range 0.2–0.5·Hardware / Power / Cost paths

[Calibrated to PaLM 540B Table 3 (MFU 0.462) + LLaMA 3 405B (MFU 0.384) + SemiAnalysis 100K-H100 reference (FP16 MFU 0.40 on trillion-parameter training, 2024-06-17); recalibrated above Epoch 2022 LLM default of 0.30](https://newsletter.semianalysis.com/p/100000-h100-clusters-power-network)

Power Usage Effectiveness (PUE)Default 1.15

range 1.05–1.45·Power path

[Operator fleet disclosures (Microsoft FY25 1.16, AWS 2024 1.15, Google fleet 1.09, Meta FY23 1.08); xAI Colossus Memphis 1.18; Stargate Abilene 1.12.](https://datacenters.microsoft.com/sustainability/efficiency/)

GPU Fraction of IT LoadDefault 0.49

range 0.35–0.65·Power path

[Scrutica-editorial midpoint, calibrated against SemiAnalysis's 100K H100 per-GPU server power decomposition (700 W GPU / 1,275 W per-GPU server ≈ 0.55 in-server share; cluster-level ≈ 0.47 after a Scrutica-derived ~10% allowance for external networking/storage on top of the per-server total).](https://newsletter.semianalysis.com/p/100000-h100-clusters-power-network)

GPU Fraction of CapexDefault 0.45

range 0.3–0.6·Cost path

SemiAnalysis

Every value here is adjustable on the Capacity Estimator sliders and traces to `src/lib/methodology/versions.ts`. The full derivation chain, with each formula and its calibration, is documented at [/methodology](/methodology#flop-estimation). Estimation bounds are parameter-sensitivity ranges (not statistical confidence intervals).

The register behind Plate I

### Every reconciled site, named

The plate draws all 314 pairs at once and names one. This is the same population as rows (both estimates, the disclosed inputs behind them, and the divergence), filterable by band and country, sortable, addressable by URL. A tight row means two independent disclosures reconcile under the defaults. A loose row is a lead, usually about the record itself: an unknown chip model behind a known power figure, or a first-phase chip count quoted against a full-build substation. The cost path is excluded at facility level; the note at the foot of the table says why.

Facility

Country

Hardware-path FLOP/day

Power-path FLOP/day

Divergence

Inputs

[NVIDIA SATURN V Volta](/facilities/fac-epoch-gc-nvidia-saturn-v-volta)

US

9.69 × 10²⁰ — view full provenance

9.70 × 10²⁰ — view full provenance

1.00×

0% normalized

264 GPUs

0 MW

[Meta 2017 V100 Cluster](/facilities/fac-epoch-gc-meta-2017-v100-cluster)

US

8.08 × 10²² — view full provenance

8.08 × 10²² — view full provenance

1.00×

0% normalized

22,000 GPUs

15 MW

[Sakura's B200s Phase 2](/facilities/fac-epoch-gc-sakura-s-b200s-phase-2)

JP

2.91 × 10²³ — view full provenance

2.90 × 10²³ — view full provenance

1.00×

0% normalized

10,000 GPUs

16 MW

[OpenAI/Microsoft Atlanta](/facilities/fac-epoch-gc-openai-microsoft-atlanta)

US

1.98 × 10²⁵ — view full provenance

1.97 × 10²⁵ — view full provenance

1.01×

1% normalized

300,000 GPUs

700 MW

[EuroHPC Karolina](/facilities/fac-epoch-gc-eurohpc-karolina)

CZ

5.28 × 10²¹ — view full provenance

5.35 × 10²¹ — view full provenance

1.01×

1% normalized

576 GPUs

1 MW

[University of Edinburgh DiRAC Tursa](/facilities/fac-epoch-gc-university-of-edinburgh-dirac-tursa)

GB

4.11 × 10²¹ — view full provenance

4.19 × 10²¹ — view full provenance

1.02×

2% normalized

448 GPUs

0 MW

[xAI Fulton Georgia](/facilities/fac-epoch-gc-xai-fulton-georgia)

US

3.62 × 10²³ — view full provenance

3.54 × 10²³ — view full provenance

1.02×

2% normalized

12,448 GPUs

20 MW

[NVIDIA Circe](/facilities/fac-epoch-gc-nvidia-circe)

US

2.12 × 10²¹ — view full provenance

2.07 × 10²¹ — view full provenance

1.02×

2% normalized

576 GPUs

0 MW

[University of Florida HiPerGator 3.0 Superpod](/facilities/fac-epoch-gc-university-of-florida-hipergator-3-0-superpod)

US

1.03 × 10²² — view full provenance

1.00 × 10²² — view full provenance

1.02×

2% normalized

1,120 GPUs

1 MW

[Lawrence Berkeley NL NERSC Perlmutter](/facilities/fac-epoch-gc-lawrence-berkeley-nl-nersc-perlmutter)

US

6.57 × 10²² — view full provenance

6.74 × 10²² — view full provenance

1.03×

3% normalized

7,168 GPUs

7 MW

[Simon Fraser University/Compute Canada Cedar](/facilities/fac-epoch-gc-simon-fraser-university-compute-canada-cedar)

CA

2.82 × 10²¹ — view full provenance

2.90 × 10²¹ — view full provenance

1.03×

3% normalized

768 GPUs

1 MW

[EuroHPC Leonardo](/facilities/fac-epoch-gc-eurohpc-leonardo)

IT

1.27 × 10²³ — view full provenance

1.23 × 10²³ — view full provenance

1.03×

3% normalized

13,824 GPUs

13 MW

[University of Tokyo Wisteria/BDEC-01 (Aquarius)](/facilities/fac-epoch-gc-university-of-tokyo-wisteria-bdec-01-aquarius)

JP

3.30 × 10²¹ — view full provenance

3.16 × 10²¹ — view full provenance

1.04×

4% normalized

360 GPUs

0 MW

[TotalEnergies Pangea III](/facilities/fac-epoch-gc-totalenergies-pangea-iii)

FR

1.24 × 10²² — view full provenance

1.30 × 10²² — view full provenance

1.04×

4% normalized

3,384 GPUs

2 MW

[Petrobras Pegasus (Pégaso)](/facilities/fac-epoch-gc-petrobras-pegasus-p-gaso)

BR

1.85 × 10²² — view full provenance

1.76 × 10²² — view full provenance

1.05×

5% normalized

2,016 GPUs

2 MW

[Atos Spartan2](/facilities/fac-epoch-gc-atos-spartan2)

FR

1.76 × 10²¹ — view full provenance

1.85 × 10²¹ — view full provenance

1.05×

5% normalized

192 GPUs

0 MW

[Cineca Marconi-100](/facilities/fac-epoch-gc-cineca-marconi-100)

IT

1.45 × 10²² — view full provenance

1.38 × 10²² — view full provenance

1.05×

5% normalized

3,952 GPUs

3 MW

[Taiwania 2](/facilities/fac-epoch-gc-taiwania-2)

TW

7.40 × 10²¹ — view full provenance

7.82 × 10²¹ — view full provenance

1.06×

6% normalized

2,016 GPUs

1 MW

[NVIDIA SATURN V Phase 3](/facilities/fac-epoch-gc-nvidia-saturn-v-phase-3)

US

2.04 × 10²² — view full provenance

1.93 × 10²² — view full provenance

1.06×

6% normalized

5,560 GPUs

4 MW

[IBM DYEUS](/facilities/fac-epoch-gc-ibm-dyeus)

US

1.09 × 10²¹ — view full provenance

1.03 × 10²¹ — view full provenance

1.06×

6% normalized

296 GPUs

0 MW

[Laboratório Nacional de Computação Científica Santos Dumont](/facilities/fac-epoch-gc-laborat-rio-nacional-de-computa-o-cient-fica-santos-dumont)

BR

1.35 × 10²¹ — view full provenance

1.28 × 10²¹ — view full provenance

1.06×

6% normalized

368 GPUs

0 MW

[Microsoft Azure ND v2 Largest Stated](/facilities/fac-epoch-gc-microsoft-azure-nd-v2-largest-stated)

US

2.94 × 10²¹ — view full provenance

2.78 × 10²¹ — view full provenance

1.06×

6% normalized

800 GPUs

1 MW

[SberCloud Christofari](/facilities/fac-epoch-gc-sbercloud-christofari)

RU

4.41 × 10²¹ — view full provenance

4.17 × 10²¹ — view full provenance

1.06×

6% normalized

1,200 GPUs

1 MW

[SENAI CIMATEC Ogbon Cimatec/Petrobras](/facilities/fac-epoch-gc-senai-cimatec-ogbon-cimatec-petrobras)

BR

1.15 × 10²¹ — view full provenance

1.08 × 10²¹ — view full provenance

1.06×

6% normalized

312 GPUs

0 MW

[AWS Fast BERT Training](/facilities/fac-epoch-gc-aws-fast-bert-training)

US

7.52 × 10²¹ — view full provenance

7.11 × 10²¹ — view full provenance

1.06×

6% normalized

2,048 GPUs

1 MW

[Preferred Networks MN-2](/facilities/fac-epoch-gc-preferred-networks-mn-2)

JP

3.76 × 10²¹ — view full provenance

3.56 × 10²¹ — view full provenance

1.06×

6% normalized

1,024 GPUs

1 MW

[SAKURA Internet 2019 V100 Supercomputer](/facilities/fac-epoch-gc-sakura-internet-2019-v100-supercomputer)

JP

2.82 × 10²¹ — view full provenance

2.67 × 10²¹ — view full provenance

1.06×

6% normalized

768 GPUs

1 MW

[TACC Longhorn](/facilities/fac-epoch-gc-tacc-longhorn)

US

1.65 × 10²¹ — view full provenance

1.56 × 10²¹ — view full provenance

1.06×

6% normalized

448 GPUs

0 MW

[CSC Puhti](/facilities/fac-epoch-gc-csc-puhti)

FI

1.18 × 10²¹ — view full provenance

1.11 × 10²¹ — view full provenance

1.06×

6% normalized

320 GPUs

0 MW

[Microsoft Research Hypercluster](/facilities/fac-epoch-gc-microsoft-research-hypercluster)

US

2.12 × 10²¹ — view full provenance

2.00 × 10²¹ — view full provenance

1.06×

6% normalized

576 GPUs

0 MW

[University of Tsukuba Cygnus](/facilities/fac-epoch-gc-university-of-tsukuba-cygnus)

JP

1.18 × 10²¹ — view full provenance

1.11 × 10²¹ — view full provenance

1.06×

6% normalized

320 GPUs

0 MW

[Lawrence Livermore NL Sierra](/facilities/fac-epoch-gc-lawrence-livermore-nl-sierra)

US

6.35 × 10²² — view full provenance

6.00 × 10²² — view full provenance

1.06×

6% normalized

17,280 GPUs

12 MW

[Ubilink.AI Supercomputer](/facilities/fac-epoch-gc-ubilink-ai-supercomputer)

TW

2.98 × 10²² — view full provenance

3.15 × 10²² — view full provenance

1.06×

6% normalized

1,024 GPUs

2 MW

[Lawrence Livermore NL RZAdams](/facilities/fac-epoch-gc-lawrence-livermore-nl-rzadams)

US

1.49 × 10²² — view full provenance

1.40 × 10²² — view full provenance

1.06×

6% normalized

512 GPUs

1 MW

[Eni HPC4 Phase 2](/facilities/fac-epoch-gc-eni-hpc4-phase-2)

IT

1.10 × 10²² — view full provenance

1.04 × 10²² — view full provenance

1.06×

6% normalized

3,000 GPUs

2 MW

[FZJ JURECA](/facilities/fac-epoch-gc-fzj-jureca)

DE

7.04 × 10²¹ — view full provenance

6.60 × 10²¹ — view full provenance

1.07×

6% normalized

768 GPUs

1 MW

[Nebius Kansas City Phase 2](/facilities/fac-epoch-gc-nebius-kansas-city-phase-2)

US

7.56 × 10²³ — view full provenance

7.08 × 10²³ — view full provenance

1.07×

7% normalized

26,000 GPUs

40 MW

[AWS EC2 P4d](/facilities/fac-epoch-gc-aws-ec2-p4d)

US

3.67 × 10²² — view full provenance

3.41 × 10²² — view full provenance

1.07×

7% normalized

4,000 GPUs

3 MW

[Continental DGX Supercomputer](/facilities/fac-epoch-gc-continental-dgx-supercomputer)

DE

1.65 × 10²¹ — view full provenance

1.53 × 10²¹ — view full provenance

1.07×

7% normalized

448 GPUs

0 MW

[Immunity Bio Cluster](/facilities/fac-epoch-gc-immunity-bio-cluster)

US

1.18 × 10²¹ — view full provenance

1.09 × 10²¹ — view full provenance

1.07×

7% normalized

320 GPUs

0 MW

[Nagoya University Flow Type II (Furo)](/facilities/fac-epoch-gc-nagoya-university-flow-type-ii-furo)

JP

3.25 × 10²¹ — view full provenance

3.02 × 10²¹ — view full provenance

1.07×

7% normalized

884 GPUs

1 MW

[NCI Australia Gadi](/facilities/fac-epoch-gc-nci-australia-gadi)

AU

2.35 × 10²¹ — view full provenance

2.19 × 10²¹ — view full provenance

1.07×

7% normalized

640 GPUs

0 MW

[Argonne NL Sophia (Formerly Theta)](/facilities/fac-epoch-gc-argonne-nl-sophia-formerly-theta)

US

1.76 × 10²¹ — view full provenance

1.64 × 10²¹ — view full provenance

1.07×

7% normalized

192 GPUs

0 MW

[Azure OpenAI GPT-3 Cluster](/facilities/fac-epoch-gc-azure-openai-gpt-3-cluster)

US

3.67 × 10²² — view full provenance

3.42 × 10²² — view full provenance

1.07×

7% normalized

10,000 GPUs

7 MW

[Japan Atomic Energy Agency and Quantum and Radiological Science and Technology HPE SGI8600](/facilities/fac-epoch-gc-japan-atomic-energy-agency-and-quantum-and-radiological-science-and-technology-h)

JP

4.00 × 10²¹ — view full provenance

3.72 × 10²¹ — view full provenance

1.07×

7% normalized

1,088 GPUs

1 MW

[Leonardo SpA davinci-1](/facilities/fac-epoch-gc-leonardo-spa-davinci-1)

IT

2.93 × 10²¹ — view full provenance

2.73 × 10²¹ — view full provenance

1.07×

7% normalized

320 GPUs

0 MW

[Microsoft Azure Immunity Bio](/facilities/fac-epoch-gc-microsoft-azure-immunity-bio)

US

4.59 × 10²¹ — view full provenance

4.27 × 10²¹ — view full provenance

1.07×

7% normalized

1,250 GPUs

1 MW

[MIT Supercloud](/facilities/fac-epoch-gc-mit-supercloud)

US

1.79 × 10²¹ — view full provenance

1.67 × 10²¹ — view full provenance

1.07×

7% normalized

488 GPUs

0 MW

[Naver DGX Superpod](/facilities/fac-epoch-gc-naver-dgx-superpod)

KR

1.03 × 10²² — view full provenance

9.55 × 10²¹ — view full provenance

1.07×

7% normalized

1,120 GPUs

1 MW

[Oracle 2020 A100 Cluster](/facilities/fac-epoch-gc-oracle-2020-a100-cluster)

US

4.69 × 10²¹ — view full provenance

4.37 × 10²¹ — view full provenance

1.07×

7% normalized

512 GPUs

0 MW

[NVIDIA In-house DGX A100 Cluster](/facilities/fac-epoch-gc-nvidia-in-house-dgx-a100-cluster)

US

3.67 × 10²¹ — view full provenance

3.41 × 10²¹ — view full provenance

1.07×

7% normalized

400 GPUs

0 MW

[RPI Supercomputer 2](/facilities/fac-epoch-gc-rpi-supercomputer-2)

US

2.23 × 10²¹ — view full provenance

2.08 × 10²¹ — view full provenance

1.07×

7% normalized

608 GPUs

0 MW

[Lawrence Livermore NL Lassen Phase 2](/facilities/fac-epoch-gc-lawrence-livermore-nl-lassen-phase-2)

US

1.16 × 10²² — view full provenance

1.08 × 10²² — view full provenance

1.07×

7% normalized

3,169 GPUs

2 MW

[Petrobras Dragão](/facilities/fac-epoch-gc-petrobras-drag-o)

BR

7.99 × 10²¹ — view full provenance

8.66 × 10²¹ — view full provenance

1.08×

8% normalized

2,176 GPUs

2 MW

[Max-Planck-Gesellschaft Raven](/facilities/fac-epoch-gc-max-planck-gesellschaft-raven)

DE

7.04 × 10²¹ — view full provenance

6.48 × 10²¹ — view full provenance

1.09×

8% normalized

768 GPUs

1 MW

[Colossus 2](/facilities/fac-epoch-xai-colossus-2)

US

1.54 × 10²⁵ — view full provenance

1.67 × 10²⁵ — view full provenance

1.09×

8% normalized

530,000 GPUs

946 MW

[Calcul Québec Narval](/facilities/fac-epoch-gc-calcul-qu-bec-narval)

CA

5.83 × 10²¹ — view full provenance

5.35 × 10²¹ — view full provenance

1.09×

9% normalized

636 GPUs

1 MW

[EuroHPC MeluXina](/facilities/fac-epoch-gc-eurohpc-meluxina)

LU

7.33 × 10²¹ — view full provenance

6.71 × 10²¹ — view full provenance

1.09×

9% normalized

800 GPUs

1 MW

[EuroHPC Vega](/facilities/fac-epoch-gc-eurohpc-vega)

SI

2.20 × 10²¹ — view full provenance

2.01 × 10²¹ — view full provenance

1.09×

9% normalized

240 GPUs

0 MW

[ND A100 v4](/facilities/fac-epoch-gc-nd-a100-v4)

US

1.83 × 10²² — view full provenance

1.68 × 10²² — view full provenance

1.09×

9% normalized

2,000 GPUs

2 MW

[Argonne NL Polaris](/facilities/fac-epoch-gc-argonne-nl-polaris)

US

2.05 × 10²² — view full provenance

1.88 × 10²² — view full provenance

1.09×

9% normalized

2,240 GPUs

2 MW

[Commissariat a l'Energie Atomique Topaze](/facilities/fac-epoch-gc-commissariat-a-l-energie-atomique-topaze)

FR

1.76 × 10²¹ — view full provenance

1.61 × 10²¹ — view full provenance

1.09×

9% normalized

192 GPUs

0 MW

[Dell Rattler](/facilities/fac-epoch-gc-dell-rattler)

US

1.76 × 10²¹ — view full provenance

1.61 × 10²¹ — view full provenance

1.09×

9% normalized

192 GPUs

0 MW

[High-Performance Computing Center Stuttgart Hawk](/facilities/fac-epoch-gc-high-performance-computing-center-stuttgart-hawk)

DE

1.76 × 10²¹ — view full provenance

1.61 × 10²¹ — view full provenance

1.09×

9% normalized

192 GPUs

0 MW

[Indiana University Big Red 200](/facilities/fac-epoch-gc-indiana-university-big-red-200)

US

2.35 × 10²¹ — view full provenance

2.15 × 10²¹ — view full provenance

1.09×

9% normalized

256 GPUs

0 MW

[Los Alamos NL Chicoma](/facilities/fac-epoch-gc-los-alamos-nl-chicoma)

US

4.69 × 10²¹ — view full provenance

4.29 × 10²¹ — view full provenance

1.09×

9% normalized

512 GPUs

0 MW

[Osaka University SQUID](/facilities/fac-epoch-gc-osaka-university-squid)

JP

3.08 × 10²¹ — view full provenance

2.82 × 10²¹ — view full provenance

1.09×

9% normalized

336 GPUs

0 MW

[Samsung SSC-21](/facilities/fac-epoch-gc-samsung-ssc-21)

KR

1.61 × 10²² — view full provenance

1.48 × 10²² — view full provenance

1.09×

9% normalized

1,760 GPUs

2 MW

[U.S. Army Jean](/facilities/fac-epoch-gc-u-s-army-jean)

US

2.57 × 10²¹ — view full provenance

2.35 × 10²¹ — view full provenance

1.09×

9% normalized

280 GPUs

0 MW

[Yandex Lyapunov](/facilities/fac-epoch-gc-yandex-lyapunov)

RU

1.00 × 10²² — view full provenance

9.19 × 10²¹ — view full provenance

1.09×

9% normalized

1,096 GPUs

1 MW

[MTS Grom](/facilities/fac-epoch-gc-mts-grom)

RU

1.47 × 10²¹ — view full provenance

1.34 × 10²¹ — view full provenance

1.09×

9% normalized

160 GPUs

0 MW

[NVIDIA Cambridge-1](/facilities/fac-epoch-gc-nvidia-cambridge-1)

GB

5.87 × 10²¹ — view full provenance

5.37 × 10²¹ — view full provenance

1.09×

9% normalized

640 GPUs

1 MW

[Recursion BioHive-1](/facilities/fac-epoch-gc-recursion-biohive-1)

US

2.93 × 10²¹ — view full provenance

2.68 × 10²¹ — view full provenance

1.09×

9% normalized

320 GPUs

0 MW

[Saudi Aramco Dammam-7](/facilities/fac-epoch-gc-saudi-aramco-dammam-7)

SA

2.91 × 10²² — view full provenance

2.66 × 10²² — view full provenance

1.09×

9% normalized

7,912 GPUs

5 MW

[SberCloud Christofari Neo](/facilities/fac-epoch-gc-sbercloud-christofari-neo)

RU

7.26 × 10²¹ — view full provenance

6.64 × 10²¹ — view full provenance

1.09×

9% normalized

792 GPUs

1 MW

[SiDi IARA](/facilities/fac-epoch-gc-sidi-iara)

BR

1.83 × 10²¹ — view full provenance

1.68 × 10²¹ — view full provenance

1.09×

9% normalized

200 GPUs

0 MW

[Tesla Auto-Labeling Cluster](/facilities/fac-epoch-gc-tesla-auto-labeling-cluster)

US

1.61 × 10²² — view full provenance

1.47 × 10²² — view full provenance

1.09×

9% normalized

1,752 GPUs

2 MW

[Tesla Training Cluster](/facilities/fac-epoch-gc-tesla-training-cluster)

US

3.70 × 10²² — view full provenance

3.38 × 10²² — view full provenance

1.09×

9% normalized

4,032 GPUs

3 MW

[Texas A&M Grace](/facilities/fac-epoch-gc-texas-a-m-grace)

US

1.83 × 10²¹ — view full provenance

1.68 × 10²¹ — view full provenance

1.09×

9% normalized

200 GPUs

0 MW

[Vingroup VinAI Research Superpod](/facilities/fac-epoch-gc-vingroup-vinai-research-superpod)

VN

1.47 × 10²¹ — view full provenance

1.34 × 10²¹ — view full provenance

1.09×

9% normalized

160 GPUs

0 MW

[Yandex Chervonenkis](/facilities/fac-epoch-gc-yandex-chervonenkis)

RU

1.46 × 10²² — view full provenance

1.34 × 10²² — view full provenance

1.09×

9% normalized

1,592 GPUs

1 MW

[Yandex Galushkin](/facilities/fac-epoch-gc-yandex-galushkin)

RU

9.97 × 10²¹ — view full provenance

9.12 × 10²¹ — view full provenance

1.09×

9% normalized

1,088 GPUs

1 MW

[Microsoft Azure Pioneer-EUS](/facilities/fac-epoch-gc-microsoft-azure-pioneer-eus)

US

1.20 × 10²² — view full provenance

1.10 × 10²² — view full provenance

1.09×

9% normalized

1,312 GPUs

1 MW

[Microsoft Azure Pioneer-SCUS](/facilities/fac-epoch-gc-microsoft-azure-pioneer-scus)

US

1.20 × 10²² — view full provenance

1.10 × 10²² — view full provenance

1.09×

9% normalized

1,312 GPUs

1 MW

[Microsoft Azure Pioneer-WEU](/facilities/fac-epoch-gc-microsoft-azure-pioneer-weu)

NL

1.20 × 10²² — view full provenance

1.10 × 10²² — view full provenance

1.09×

9% normalized

1,312 GPUs

1 MW

[Microsoft Azure Pioneer-WUS2](/facilities/fac-epoch-gc-microsoft-azure-pioneer-wus2)

US

1.20 × 10²² — view full provenance

1.10 × 10²² — view full provenance

1.09×

9% normalized

1,312 GPUs

1 MW

[Microsoft Azure Voyager-EUS2](/facilities/fac-epoch-gc-microsoft-azure-voyager-eus2)

US

1.94 × 10²² — view full provenance

1.77 × 10²² — view full provenance

1.09×

9% normalized

2,112 GPUs

2 MW

[University of Minnesota Agate](/facilities/fac-epoch-gc-university-of-minnesota-agate)

US

2.42 × 10²¹ — view full provenance

2.21 × 10²¹ — view full provenance

1.09×

9% normalized

264 GPUs

0 MW

[Karlsruher Institut für Technologie HoreKa](/facilities/fac-epoch-gc-karlsruher-institut-f-r-technologie-horeka)

DE

6.78 × 10²¹ — view full provenance

6.17 × 10²¹ — view full provenance

1.10×

9% normalized

740 GPUs

1 MW

[NSCC ASPIRE 2A Phase 2](/facilities/fac-epoch-gc-nscc-aspire-2a-phase-2)

SG

3.23 × 10²¹ — view full provenance

2.93 × 10²¹ — view full provenance

1.10×

10% normalized

352 GPUs

0 MW

[AGH Cyfronet Athena](/facilities/fac-epoch-gc-agh-cyfronet-athena)

PL

3.52 × 10²¹ — view full provenance

3.19 × 10²¹ — view full provenance

1.10×

10% normalized

384 GPUs

0 MW

[Ezra-1 Stability AI AWS Cluster](/facilities/fac-epoch-gc-ezra-1-stability-ai-aws-cluster)

—

3.67 × 10²² — view full provenance

3.33 × 10²² — view full provenance

1.10×

10% normalized

4,000 GPUs

3 MW

[IBM Vela](/facilities/fac-epoch-gc-ibm-vela)

US

2.64 × 10²² — view full provenance

2.39 × 10²² — view full provenance

1.10×

10% normalized

2,880 GPUs

2 MW

[Indiana University Bloomington Jetstream2](/facilities/fac-epoch-gc-indiana-university-bloomington-jetstream2)

US

3.30 × 10²¹ — view full provenance

2.99 × 10²¹ — view full provenance

1.10×

10% normalized

360 GPUs

0 MW

[Microsoft GPT-4 cluster](/facilities/fac-epoch-gc-microsoft-gpt-4-cluster)

US

2.29 × 10²³ — view full provenance

2.08 × 10²³ — view full provenance

1.10×

10% normalized

25,000 GPUs

21 MW

[TACC Lonestar6](/facilities/fac-epoch-gc-tacc-lonestar6)

US

2.31 × 10²¹ — view full provenance

2.10 × 10²¹ — view full provenance

1.10×

10% normalized

252 GPUs

0 MW

[Aleph Alpha alpha ONE](/facilities/fac-epoch-gc-aleph-alpha-alpha-one)

DE

4.69 × 10²¹ — view full provenance

4.26 × 10²¹ — view full provenance

1.10×

10% normalized

512 GPUs

0 MW

[Tesla A100 Cluster Phase 2](/facilities/fac-epoch-gc-tesla-a100-cluster-phase-2)

US

6.75 × 10²² — view full provenance

6.12 × 10²² — view full provenance

1.10×

10% normalized

7,360 GPUs

6 MW

[Microsoft Azure Meta AI Rental](/facilities/fac-epoch-gc-microsoft-azure-meta-ai-rental)

US

4.95 × 10²² — view full provenance

4.49 × 10²² — view full provenance

1.10×

10% normalized

5,400 GPUs

5 MW

[GENCI Adastra](/facilities/fac-epoch-gc-genci-adastra)

FR

1.52 × 10²² — view full provenance

1.38 × 10²² — view full provenance

1.10×

10% normalized

1,352 GPUs

2 MW

[G42 Microsoft 100 MW UAE Cluster](/facilities/fac-epoch-gc-g42-microsoft-100mw-uae-cluster)

AE

1.60 × 10²⁴ — view full provenance

1.77 × 10²⁴ — view full provenance

1.11×

10% normalized

55,000 GPUs

100 MW

[Calcul Québec Béluga](/facilities/fac-epoch-gc-calcul-qu-bec-b-luga)

CA

2.53 × 10²¹ — view full provenance

2.28 × 10²¹ — view full provenance

1.11×

10% normalized

688 GPUs

0 MW

[Nebius ISEG](/facilities/fac-epoch-gc-nebius-iseg)

FI

4.42 × 10²² — view full provenance

3.97 × 10²² — view full provenance

1.11×

11% normalized

1,520 GPUs

2 MW

[JUWELS-Booster](/facilities/fac-epoch-gc-juwels-booster)

DE

3.43 × 10²² — view full provenance

3.08 × 10²² — view full provenance

1.11×

11% normalized

3,744 GPUs

3 MW

[ExxonMobil Discovery 5](/facilities/fac-epoch-gc-exxonmobil-discovery-5)

US

1.83 × 10²² — view full provenance

1.64 × 10²² — view full provenance

1.12×

11% normalized

2,000 GPUs

2 MW

[CoreWeave Muskogee](/facilities/fac-epoch-gc-coreweave-muskogee)

US

2.94 × 10²⁴ — view full provenance

2.61 × 10²⁴ — view full provenance

1.13×

12% normalized

40,000 GPUs

100 MW

[NVIDIA Selene Phase 2](/facilities/fac-epoch-gc-nvidia-selene-phase-2)

US

4.11 × 10²² — view full provenance

4.63 × 10²² — view full provenance

1.13×

12% normalized

4,480 GPUs

5 MW

[TU Dresden AlphaCentauri](/facilities/fac-epoch-gc-tu-dresden-alphacentauri)

DE

2.49 × 10²¹ — view full provenance

2.81 × 10²¹ — view full provenance

1.13×

12% normalized

272 GPUs

0 MW

[DeepL Mercury](/facilities/fac-epoch-gc-deepl-mercury)

SE

1.58 × 10²² — view full provenance

1.40 × 10²² — view full provenance

1.13×

12% normalized

544 GPUs

1 MW

[Microsoft Azure Eagle](/facilities/fac-epoch-gc-microsoft-azure-eagle)

US

4.19 × 10²³ — view full provenance

3.70 × 10²³ — view full provenance

1.13×

12% normalized

14,400 GPUs

21 MW

[AWS EC2 P5 UltraClusters](/facilities/fac-epoch-gc-aws-ec2-p5-ultraclusters)

US

5.81 × 10²³ — view full provenance

5.14 × 10²³ — view full provenance

1.13×

12% normalized

20,000 GPUs

29 MW

[Google TPUv5e](/facilities/fac-epoch-gc-google-tpuv5e)

—

2.95 × 10²³ — view full provenance

2.61 × 10²³ — view full provenance

1.13×

12% normalized

50,944 GPUs

24 MW

[Imbue 10k Cluster](/facilities/fac-epoch-gc-imbue-10k-cluster)

US

2.91 × 10²³ — view full provenance

2.57 × 10²³ — view full provenance

1.13×

12% normalized

10,000 GPUs

15 MW

[Lambda Labs H100/H200](/facilities/fac-epoch-gc-lambda-labs-h100-h200)

US

9.30 × 10²³ — view full provenance

8.22 × 10²³ — view full provenance

1.13×

12% normalized

32,000 GPUs

46 MW

[Microsoft Azure ND H100 v5 VM](/facilities/fac-epoch-gc-microsoft-azure-nd-h100-v5-vm)

US

1.16 × 10²³ — view full provenance

1.03 × 10²³ — view full provenance

1.13×

12% normalized

4,000 GPUs

6 MW

[Mistral Large Model Implied Cluster](/facilities/fac-epoch-gc-mistral-large-model-implied-cluster)

—

1.16 × 10²³ — view full provenance

1.03 × 10²³ — view full provenance

1.13×

12% normalized

4,000 GPUs

6 MW

[NHN Cloud's National AI Data Center](/facilities/fac-epoch-gc-nhn-cloud-s-national-ai-data-center)

KR

2.91 × 10²² — view full provenance

2.57 × 10²² — view full provenance

1.13×

12% normalized

1,000 GPUs

1 MW

[NVIDIA Eos Phase 2](/facilities/fac-epoch-gc-nvidia-eos-phase-2)

US

1.34 × 10²³ — view full provenance

1.18 × 10²³ — view full provenance

1.13×

12% normalized

4,608 GPUs

7 MW

[Oracle OCI Supercluster A100s](/facilities/fac-epoch-gc-oracle-oci-supercluster-a100s)

US

3.00 × 10²³ — view full provenance

2.66 × 10²³ — view full provenance

1.13×

12% normalized

32,768 GPUs

27 MW

[Scaleway Nabuchodonosor](/facilities/fac-epoch-gc-scaleway-nabuchodonosor)

FR

2.95 × 10²² — view full provenance

2.61 × 10²² — view full provenance

1.13×

12% normalized

1,016 GPUs

1 MW

[Softbank SuperPOD](/facilities/fac-epoch-gc-softbank-superpod)

JP

1.88 × 10²² — view full provenance

1.66 × 10²² — view full provenance

1.13×

12% normalized

2,048 GPUs

2 MW

[Tesla 10k H100 Cluster](/facilities/fac-epoch-gc-tesla-10k-h100-cluster)

US

2.91 × 10²³ — view full provenance

2.57 × 10²³ — view full provenance

1.13×

12% normalized

10,000 GPUs

15 MW

[Together AI H100 Cluster](/facilities/fac-epoch-gc-together-ai-h100-cluster)

US

1.29 × 10²³ — view full provenance

1.14 × 10²³ — view full provenance

1.13×

12% normalized

4,424 GPUs

6 MW

[Amazon Titan training cluster](/facilities/fac-epoch-gc-amazon-titan-training-cluster)

US

1.26 × 10²³ — view full provenance

1.11 × 10²³ — view full provenance

1.13×

12% normalized

13,760 GPUs

11 MW

[Google Hypercomputer TPU v5p pod](/facilities/fac-epoch-gc-google-hypercomputer-tpu-v5p-pod)

—

1.21 × 10²³ — view full provenance

1.07 × 10²³ — view full provenance

1.13×

12% normalized

8,960 GPUs

10 MW

[MosaicML MPT training cluster](/facilities/fac-epoch-gc-mosaicml-mpt-training-cluster)

—

4.03 × 10²¹ — view full provenance

3.57 × 10²¹ — view full provenance

1.13×

12% normalized

440 GPUs

0 MW

[MPT-30B training cluster](/facilities/fac-epoch-gc-mpt-30b-training-cluster)

—

7.44 × 10²¹ — view full provenance

6.58 × 10²¹ — view full provenance

1.13×

12% normalized

256 GPUs

0 MW

[NAVER Corp Sejong](/facilities/fac-epoch-gc-naver-corp-sejong)

KR

2.05 × 10²² — view full provenance

1.81 × 10²² — view full provenance

1.13×

12% normalized

2,240 GPUs

2 MW

[NEC Corp Japan Supercomputer](/facilities/fac-epoch-gc-nec-corp-japan-supercomputer)

JP

8.51 × 10²¹ — view full provenance

7.52 × 10²¹ — view full provenance

1.13×

12% normalized

928 GPUs

1 MW

[NVIDIA Helios](/facilities/fac-epoch-gc-nvidia-helios)

—

2.98 × 10²² — view full provenance

2.63 × 10²² — view full provenance

1.13×

12% normalized

1,024 GPUs

1 MW

[SK Telecom Titan Phase 2](/facilities/fac-epoch-gc-sk-telecom-titan-phase-2)

KR

9.53 × 10²¹ — view full provenance

8.43 × 10²¹ — view full provenance

1.13×

12% normalized

1,040 GPUs

1 MW

[AIRAWAT-PSAI Phase 2](/facilities/fac-epoch-gc-airawat-psai-phase-2)

IN

6.01 × 10²¹ — view full provenance

5.32 × 10²¹ — view full provenance

1.13×

12% normalized

656 GPUs

1 MW

[hessian.AI fortytwo](/facilities/fac-epoch-gc-hessian-ai-fortytwo)

DE

5.79 × 10²¹ — view full provenance

5.12 × 10²¹ — view full provenance

1.13×

12% normalized

632 GPUs

1 MW

[KT SuperPOD](/facilities/fac-epoch-gc-kt-superpod)

KR

7.26 × 10²¹ — view full provenance

6.42 × 10²¹ — view full provenance

1.13×

12% normalized

792 GPUs

1 MW

[Meta Research SuperCluster (RSC-1) Phase 2](/facilities/fac-epoch-gc-meta-research-supercluster-rsc-1-phase-2)

US

1.47 × 10²³ — view full provenance

1.30 × 10²³ — view full provenance

1.13×

12% normalized

16,000 GPUs

13 MW

[Meta Research SuperCluster 2 (RSC-2)](/facilities/fac-epoch-gc-meta-research-supercluster-2-rsc-2)

US

7.33 × 10²² — view full provenance

6.48 × 10²² — view full provenance

1.13×

12% normalized

8,000 GPUs

7 MW

[NSC Berzelius Phase 2](/facilities/fac-epoch-gc-nsc-berzelius-phase-2)

SE

6.89 × 10²¹ — view full provenance

6.09 × 10²¹ — view full provenance

1.13×

12% normalized

752 GPUs

1 MW

[KAUST Shaheen-III](/facilities/fac-epoch-gc-kaust-shaheen-iii)

SA

8.19 × 10²² — view full provenance

9.38 × 10²² — view full provenance

1.15×

14% normalized

2,816 GPUs

5 MW

[RPI AiMOS](/facilities/fac-epoch-gc-rpi-aimos)

US

5.55 × 10²¹ — view full provenance

4.84 × 10²¹ — view full provenance

1.15×

14% normalized

1,512 GPUs

1 MW

[Meta GenAI 2024a](/facilities/fac-epoch-gc-meta-genai-2024a)

US

7.14 × 10²³ — view full provenance

6.20 × 10²³ — view full provenance

1.15×

14% normalized

24,576 GPUs

35 MW

[Meta GenAI 2024b](/facilities/fac-epoch-gc-meta-genai-2024b)

US

7.14 × 10²³ — view full provenance

6.20 × 10²³ — view full provenance

1.15×

14% normalized

24,576 GPUs

35 MW

[Saudi Aramco Tuwaiq-1](/facilities/fac-epoch-gc-saudi-aramco-tuwaiq-1)

SA

1.21 × 10²² — view full provenance

1.05 × 10²² — view full provenance

1.15×

14% normalized

416 GPUs

1 MW

[Wroclaw Centre for Networking and Supercomputing Lem](/facilities/fac-epoch-gc-wroclaw-centre-for-networking-and-supercomputing-lem)

PL

8.84 × 10²¹ — view full provenance

7.67 × 10²¹ — view full provenance

1.15×

14% normalized

304 GPUs

0 MW

[Core42 SuperPOD](/facilities/fac-epoch-gc-core42-superpod)

AE

3.70 × 10²² — view full provenance

3.21 × 10²² — view full provenance

1.15×

14% normalized

1,272 GPUs

2 MW

[CSIRO Virga](/facilities/fac-epoch-gc-csiro-virga)

AU

1.30 × 10²² — view full provenance

1.13 × 10²² — view full provenance

1.15×

14% normalized

448 GPUs

1 MW

[Denvr Dataworks H100](/facilities/fac-epoch-gc-denvr-dataworks-h100)

US

2.98 × 10²² — view full provenance

2.59 × 10²² — view full provenance

1.15×

14% normalized

1,024 GPUs

1 MW

[GreenNode Bangkok Cluster](/facilities/fac-epoch-gc-greennode-bangkok-cluster)

TH

2.98 × 10²² — view full provenance

2.59 × 10²² — view full provenance

1.15×

14% normalized

1,024 GPUs

1 MW

[IBM Blue Vela](/facilities/fac-epoch-gc-ibm-blue-vela)

US

2.98 × 10²² — view full provenance

2.59 × 10²² — view full provenance

1.15×

14% normalized

1,024 GPUs

1 MW

[LeptonAI H100 Cluster](/facilities/fac-epoch-gc-leptonai-h100-cluster)

US

2.98 × 10²² — view full provenance

2.59 × 10²² — view full provenance

1.15×

14% normalized

1,024 GPUs

1 MW

[Meta 100k](/facilities/fac-epoch-gc-meta-100k)

US

2.91 × 10²⁴ — view full provenance

2.52 × 10²⁴ — view full provenance

1.15×

14% normalized

100,000 GPUs

143 MW

[Neevcloud cluster 1](/facilities/fac-epoch-gc-neevcloud-cluster-1)

IN

2.98 × 10²² — view full provenance

2.59 × 10²² — view full provenance

1.15×

14% normalized

1,024 GPUs

1 MW

[NexGen Cloud Hyperstack AQ Compute Supercomputer](/facilities/fac-epoch-gc-nexgen-cloud-hyperstack-aq-compute-supercomputer)

NO

4.76 × 10²³ — view full provenance

4.14 × 10²³ — view full provenance

1.15×

14% normalized

16,384 GPUs

23 MW

[Northern Data Group Njored Taiga Cloud Island 5](/facilities/fac-epoch-gc-northern-data-group-njored-taiga-cloud-island-5)

GB

5.95 × 10²² — view full provenance

5.17 × 10²² — view full provenance

1.15×

14% normalized

2,048 GPUs

3 MW

[Northern Data Group Taiga Cloud Island 3](/facilities/fac-epoch-gc-northern-data-group-taiga-cloud-island-3)

—

5.95 × 10²² — view full provenance

5.17 × 10²² — view full provenance

1.15×

14% normalized

2,048 GPUs

3 MW

[Northern Data Group Taiga Cloud Island 4](/facilities/fac-epoch-gc-northern-data-group-taiga-cloud-island-4)

—

5.95 × 10²² — view full provenance

5.17 × 10²² — view full provenance

1.15×

14% normalized

2,048 GPUs

3 MW

[Northern Data Group Taiga Cloud NO1 Island 1](/facilities/fac-epoch-gc-northern-data-group-taiga-cloud-no1-island-1)

NO

5.95 × 10²² — view full provenance

5.17 × 10²² — view full provenance

1.15×

14% normalized

2,048 GPUs

3 MW

[Northern Data Group Taiga Cloud NO1 Island 2](/facilities/fac-epoch-gc-northern-data-group-taiga-cloud-no1-island-2)

NO

5.95 × 10²² — view full provenance

5.17 × 10²² — view full provenance

1.15×

14% normalized

2,048 GPUs

3 MW

[NVIDIA Israel-1 Phase 2](/facilities/fac-epoch-gc-nvidia-israel-1-phase-2)

IL

5.95 × 10²² — view full provenance

5.17 × 10²² — view full provenance

1.15×

14% normalized

2,048 GPUs

3 MW

[OpenAI/Microsoft Goodyear Arizona](/facilities/fac-epoch-gc-openai-microsoft-goodyear-arizona)

US

2.91 × 10²⁴ — view full provenance

2.52 × 10²⁴ — view full provenance

1.15×

14% normalized

100,000 GPUs

143 MW

[Oracle OCI MI300x](/facilities/fac-epoch-gc-oracle-oci-mi300x)

US

6.29 × 10²³ — view full provenance

5.47 × 10²³ — view full provenance

1.15×

14% normalized

16,384 GPUs

25 MW

[Oracle OCI Supercluster H100s](/facilities/fac-epoch-gc-oracle-oci-supercluster-h100s)

US

4.76 × 10²³ — view full provenance

4.14 × 10²³ — view full provenance

1.15×

14% normalized

16,384 GPUs

23 MW

[Oracle OCI Supercluster H200s](/facilities/fac-epoch-gc-oracle-oci-supercluster-h200s)

US

1.90 × 10²⁴ — view full provenance

1.65 × 10²⁴ — view full provenance

1.15×

14% normalized

65,536 GPUs

94 MW

[Ori Global Cloud H100 Cluster](/facilities/fac-epoch-gc-ori-global-cloud-h100-cluster)

—

2.98 × 10²² — view full provenance

2.59 × 10²² — view full provenance

1.15×

14% normalized

1,024 GPUs

1 MW

[PCSS Poznan Proxima](/facilities/fac-epoch-gc-pcss-poznan-proxima)

PL

1.01 × 10²² — view full provenance

8.79 × 10²¹ — view full provenance

1.15×

14% normalized

348 GPUs

0 MW

[Sesterce H100s Phase 2](/facilities/fac-epoch-gc-sesterce-h100s-phase-2)

FR

1.19 × 10²³ — view full provenance

1.03 × 10²³ — view full provenance

1.15×

14% normalized

4,096 GPUs

6 MW

[Sesterce Nordics](/facilities/fac-epoch-gc-sesterce-nordics)

—

2.38 × 10²³ — view full provenance

2.07 × 10²³ — view full provenance

1.15×

14% normalized

8,192 GPUs

12 MW

[SIAM AI HGX](/facilities/fac-epoch-gc-siam-ai-hgx)

TH

2.98 × 10²² — view full provenance

2.59 × 10²² — view full provenance

1.15×

14% normalized

1,024 GPUs

1 MW

[SoftBank CHIE-2](/facilities/fac-epoch-gc-softbank-chie-2)

JP

5.93 × 10²² — view full provenance

5.15 × 10²² — view full provenance

1.15×

14% normalized

2,040 GPUs

3 MW

[SoftBank CHIE-3](/facilities/fac-epoch-gc-softbank-chie-3)

JP

5.93 × 10²² — view full provenance

5.15 × 10²² — view full provenance

1.15×

14% normalized

2,040 GPUs

3 MW

[Tesla Earnings Call Claim](/facilities/fac-epoch-gc-tesla-earnings-call-claim)

US

1.02 × 10²⁴ — view full provenance

8.84 × 10²³ — view full provenance

1.15×

14% normalized

35,000 GPUs

50 MW

[Ahrefs Yep1](/facilities/fac-epoch-gc-ahrefs-yep1)

US

1.47 × 10²² — view full provenance

1.27 × 10²² — view full provenance

1.15×

14% normalized

504 GPUs

1 MW

[Andreessen Horowitz Oxygen](/facilities/fac-epoch-gc-andreessen-horowitz-oxygen)

US

5.81 × 10²³ — view full provenance

5.05 × 10²³ — view full provenance

1.15×

14% normalized

20,000 GPUs

29 MW

[BNY Mellon Supercomputer](/facilities/fac-epoch-gc-bny-mellon-supercomputer)

—

1.45 × 10²² — view full provenance

1.26 × 10²² — view full provenance

1.15×

14% normalized

500 GPUs

1 MW

[CoreWeave H200s](/facilities/fac-epoch-gc-coreweave-h200s)

US

1.22 × 10²⁴ — view full provenance

1.06 × 10²⁴ — view full provenance

1.15×

14% normalized

42,000 GPUs

60 MW

[Fastweb NeXXt AI Factory](/facilities/fac-epoch-gc-fastweb-nexxt-ai-factory)

IT

7.21 × 10²¹ — view full provenance

6.26 × 10²¹ — view full provenance

1.15×

14% normalized

248 GPUs

0 MW

[Hut 8 H100 Cluster](/facilities/fac-epoch-gc-hut-8-h100-cluster)

US

2.91 × 10²² — view full provenance

2.52 × 10²² — view full provenance

1.15×

14% normalized

1,000 GPUs

1 MW

[Iris Energy Prince George cluster](/facilities/fac-epoch-gc-iris-energy-prince-george-cluster)

CA

2.37 × 10²² — view full provenance

2.06 × 10²² — view full provenance

1.15×

14% normalized

816 GPUs

1 MW

[Opera Iceland KEF-1 SuperPOD](/facilities/fac-epoch-gc-opera-iceland-kef-1-superpod)

IS

7.21 × 10²¹ — view full provenance

6.26 × 10²¹ — view full provenance

1.15×

14% normalized

248 GPUs

0 MW

[Recursion BioHive-2](/facilities/fac-epoch-gc-recursion-biohive-2)

US

1.47 × 10²² — view full provenance

1.27 × 10²² — view full provenance

1.15×

14% normalized

504 GPUs

1 MW

[Vultr Chicago Cluster](/facilities/fac-epoch-gc-vultr-chicago-cluster)

US

1.15 × 10²³ — view full provenance

1.00 × 10²³ — view full provenance

1.15×

14% normalized

3,000 GPUs

5 MW

[TSUBAME4.0](/facilities/fac-epoch-gc-tsubame4-0)

JP

2.79 × 10²² — view full provenance

2.41 × 10²² — view full provenance

1.16×

15% normalized

960 GPUs

1 MW

[Gcore data center Phase 2](/facilities/fac-epoch-gc-gcore-data-center-phase-2)

KR

2.91 × 10²² — view full provenance

3.37 × 10²² — view full provenance

1.16×

15% normalized

1,000 GPUs

2 MW

[University of Cambridge Wilkes-3](/facilities/fac-epoch-gc-university-of-cambridge-wilkes-3)

GB

2.93 × 10²¹ — view full provenance

2.52 × 10²¹ — view full provenance

1.16×

15% normalized

320 GPUs

0 MW

[Inflection AI Cluster](/facilities/fac-epoch-gc-inflection-ai-cluster)

US

6.39 × 10²³ — view full provenance

5.48 × 10²³ — view full provenance

1.17×

15% normalized

22,000 GPUs

31 MW

[Novo Nordisk Gefion](/facilities/fac-epoch-gc-novo-nordisk-gefion)

DK

4.44 × 10²² — view full provenance

5.19 × 10²² — view full provenance

1.17×

15% normalized

1,528 GPUs

3 MW

[Project Ceiba Phase 2](/facilities/fac-epoch-gc-project-ceiba-phase-2)

US

1.52 × 10²⁴ — view full provenance

1.30 × 10²⁴ — view full provenance

1.17×

16% normalized

20,736 GPUs

50 MW

[ExxonMobil Discovery 6](/facilities/fac-epoch-gc-exxonmobil-discovery-6)

US

1.17 × 10²³ — view full provenance

10.00 × 10²² — view full provenance

1.17×

16% normalized

4,032 GPUs

6 MW

[Google A3 VMs](/facilities/fac-epoch-gc-google-a3-vms)

US

7.56 × 10²³ — view full provenance

6.45 × 10²³ — view full provenance

1.17×

16% normalized

26,000 GPUs

36 MW

[Nebius Finland Phase 2](/facilities/fac-epoch-gc-nebius-finland-phase-2)

FI

1.74 × 10²⁴ — view full provenance

1.49 × 10²⁴ — view full provenance

1.17×

16% normalized

60,000 GPUs

84 MW

[Saudi Data & AI Authority Sovereign AI factory](/facilities/fac-epoch-gc-saudi-data-ai-authority-sovereign-ai-factory)

SA

3.67 × 10²³ — view full provenance

3.13 × 10²³ — view full provenance

1.17×

16% normalized

5,000 GPUs

12 MW

[Sesterce Valence](/facilities/fac-epoch-gc-sesterce-valence)

FR

2.94 × 10²⁴ — view full provenance

2.51 × 10²⁴ — view full provenance

1.17×

16% normalized

40,000 GPUs

96 MW

[together.ai 36k GB200s](/facilities/fac-epoch-gc-together-ai-36k-gb200s)

US

2.64 × 10²⁴ — view full provenance

2.26 × 10²⁴ — view full provenance

1.17×

16% normalized

36,000 GPUs

86 MW

[AIST ABCI 3.0](/facilities/fac-epoch-gc-aist-abci-3-0)

JP

1.78 × 10²³ — view full provenance

1.52 × 10²³ — view full provenance

1.17×

16% normalized

6,128 GPUs

9 MW

[AIST ABCI-Q](/facilities/fac-epoch-gc-aist-abci-q)

JP

5.95 × 10²² — view full provenance

5.08 × 10²² — view full provenance

1.17×

16% normalized

2,048 GPUs

3 MW

[CEA EXA1-HE Phase 3](/facilities/fac-epoch-gc-cea-exa1-he-phase-3)

FR

7.81 × 10²² — view full provenance

6.66 × 10²² — view full provenance

1.17×

16% normalized

2,688 GPUs

4 MW

[Foxconn Hon Hai Kaohsiung Supercomputer](/facilities/fac-epoch-gc-foxconn-hon-hai-kaohsiung-supercomputer)

TW

3.38 × 10²³ — view full provenance

2.89 × 10²³ — view full provenance

1.17×

16% normalized

4,608 GPUs

11 MW

[Horizon Compute Baobab Phase 2](/facilities/fac-epoch-gc-horizon-compute-baobab-phase-2)

US

5.95 × 10²² — view full provenance

5.08 × 10²² — view full provenance

1.17×

16% normalized

2,048 GPUs

3 MW

[iGenius Colosseum](/facilities/fac-epoch-gc-igenius-colosseum)

IT

4.23 × 10²³ — view full provenance

3.61 × 10²³ — view full provenance

1.17×

16% normalized

5,760 GPUs

14 MW

[JCAHPC Miyabi](/facilities/fac-epoch-gc-jcahpc-miyabi)

JP

3.26 × 10²² — view full provenance

2.78 × 10²² — view full provenance

1.17×

16% normalized

1,120 GPUs

2 MW

[KDDI Sharp Sakai](/facilities/fac-epoch-gc-kddi-sharp-sakai)

JP

1.47 × 10²³ — view full provenance

1.25 × 10²³ — view full provenance

1.17×

16% normalized

2,000 GPUs

5 MW

[MITRE Federal AI Sandbox](/facilities/fac-epoch-gc-mitre-federal-ai-sandbox)

US

7.44 × 10²¹ — view full provenance

6.35 × 10²¹ — view full provenance

1.17×

16% normalized

256 GPUs

0 MW

[PanaAI AUS AISF](/facilities/fac-epoch-gc-panaai-aus-aisf)

AU

1.19 × 10²³ — view full provenance

1.01 × 10²³ — view full provenance

1.17×

16% normalized

4,088 GPUs

6 MW

[S. Korea 6th national supercomputer](/facilities/fac-epoch-gc-s-korea-6th-national-supercomputer)

KR

2.56 × 10²³ — view full provenance

2.18 × 10²³ — view full provenance

1.17×

16% normalized

8,800 GPUs

12 MW

[Sesterce Pegasus](/facilities/fac-epoch-gc-sesterce-pegasus)

—

7.71 × 10²³ — view full provenance

6.58 × 10²³ — view full provenance

1.17×

16% normalized

10,496 GPUs

25 MW

[Sesterce Synapse Phase 2](/facilities/fac-epoch-gc-sesterce-synapse-phase-2)

FR

1.19 × 10²³ — view full provenance

1.02 × 10²³ — view full provenance

1.17×

16% normalized

4,096 GPUs

6 MW

[TensorWave MI300X Cluster 1 Phase 2](/facilities/fac-epoch-gc-tensorwave-mi300x-cluster-1-phase-2)

US

3.84 × 10²³ — view full provenance

3.28 × 10²³ — view full provenance

1.17×

16% normalized

10,000 GPUs

15 MW

[TensorWave MI300X Cluster 2](/facilities/fac-epoch-gc-tensorwave-mi300x-cluster-2)

US

3.84 × 10²³ — view full provenance

3.28 × 10²³ — view full provenance

1.17×

16% normalized

10,000 GPUs

15 MW

[Voltage Park Location 5](/facilities/fac-epoch-gc-voltage-park-location-5)

US

1.19 × 10²³ — view full provenance

1.01 × 10²³ — view full provenance

1.17×

16% normalized

4,088 GPUs

6 MW

[Voltage Park Location 6](/facilities/fac-epoch-gc-voltage-park-location-6)

US

1.19 × 10²³ — view full provenance

1.01 × 10²³ — view full provenance

1.17×

16% normalized

4,088 GPUs

6 MW

[Voltage Park Texas Phase 2](/facilities/fac-epoch-gc-voltage-park-texas-phase-2)

US

1.19 × 10²³ — view full provenance

1.01 × 10²³ — view full provenance

1.17×

16% normalized

4,088 GPUs

6 MW

[Voltage Park Utah](/facilities/fac-epoch-gc-voltage-park-utah)

US

1.19 × 10²³ — view full provenance

1.01 × 10²³ — view full provenance

1.17×

16% normalized

4,088 GPUs

6 MW

[Voltage Park Virginia](/facilities/fac-epoch-gc-voltage-park-virginia)

US

1.19 × 10²³ — view full provenance

1.01 × 10²³ — view full provenance

1.17×

16% normalized

4,088 GPUs

6 MW

[Voltage Park Washington](/facilities/fac-epoch-gc-voltage-park-washington)

US

1.19 × 10²³ — view full provenance

1.01 × 10²³ — view full provenance

1.17×

16% normalized

4,088 GPUs

6 MW

[Yotta Shakti Cloud D1](/facilities/fac-epoch-gc-yotta-shakti-cloud-d1)

IN

4.76 × 10²³ — view full provenance

4.06 × 10²³ — view full provenance

1.17×

16% normalized

16,384 GPUs

23 MW

[Yotta Shakti Cloud NM1 Phase 2](/facilities/fac-epoch-gc-yotta-shakti-cloud-nm1-phase-2)

IN

4.76 × 10²³ — view full provenance

4.06 × 10²³ — view full provenance

1.17×

16% normalized

16,384 GPUs

23 MW

[YTL AI Johor](/facilities/fac-epoch-gc-ytl-ai-johor)

MY

1.13 × 10²⁴ — view full provenance

9.67 × 10²³ — view full provenance

1.17×

16% normalized

15,428 GPUs

37 MW

[CoreWeave LiquidLab](/facilities/fac-epoch-gc-coreweave-liquidlab)

US

1.16 × 10²³ — view full provenance

9.92 × 10²² — view full provenance

1.17×

16% normalized

4,000 GPUs

6 MW

[Corvex B200s](/facilities/fac-epoch-gc-corvex-b200s)

US

1.69 × 10²² — view full provenance

1.44 × 10²² — view full provenance

1.17×

16% normalized

256 GPUs

1 MW

[FPT AI Factory Japan](/facilities/fac-epoch-gc-fpt-ai-factory-japan)

JP

8.72 × 10²² — view full provenance

7.44 × 10²² — view full provenance

1.17×

16% normalized

3,000 GPUs

4 MW

[FPT AI Factory Vietnam](/facilities/fac-epoch-gc-fpt-ai-factory-vietnam)

VN

8.72 × 10²² — view full provenance

7.44 × 10²² — view full provenance

1.17×

16% normalized

3,000 GPUs

4 MW

[Huawei Pangu Ultra MoE 910Bs](/facilities/fac-epoch-gc-huawei-pangu-ultra-moe-910bs)

—

6.61 × 10²² — view full provenance

5.64 × 10²² — view full provenance

1.17×

16% normalized

6,000 GPUs

5 MW

[Magic G4 Google Cloud Rental](/facilities/fac-epoch-gc-magic-g4-google-cloud-rental)

US

2.33 × 10²³ — view full provenance

1.98 × 10²³ — view full provenance

1.17×

16% normalized

8,000 GPUs

11 MW

[Nebius ISEG2](/facilities/fac-epoch-gc-nebius-iseg2)

IS

1.45 × 10²³ — view full provenance

1.24 × 10²³ — view full provenance

1.17×

16% normalized

4,992 GPUs

7 MW

[NVIDIA CoreWeave Eos-DFW Rumored Phase 2](/facilities/fac-epoch-gc-nvidia-coreweave-eos-dfw-rumored-phase-2)

US

9.30 × 10²³ — view full provenance

7.93 × 10²³ — view full provenance

1.17×

16% normalized

32,000 GPUs

45 MW

[OneAsia OBON Clusters](/facilities/fac-epoch-gc-oneasia-obon-clusters)

TH

1.16 × 10²³ — view full provenance

9.92 × 10²² — view full provenance

1.17×

16% normalized

4,000 GPUs

6 MW

[Oracle OCI Supercluster B200s](/facilities/fac-epoch-gc-oracle-oci-supercluster-b200s)

US

8.66 × 10²⁴ — view full provenance

7.39 × 10²⁴ — view full provenance

1.17×

16% normalized

131,072 GPUs

262 MW

[Poolside 10k Cluster](/facilities/fac-epoch-gc-poolside-10k-cluster)

—

2.91 × 10²³ — view full provenance

2.48 × 10²³ — view full provenance

1.17×

16% normalized

10,000 GPUs

14 MW

[Quebec 2k H100 Cluster](/facilities/fac-epoch-gc-quebec-2k-h100-cluster)

CA

5.81 × 10²² — view full provenance

4.96 × 10²² — view full provenance

1.17×

16% normalized

2,000 GPUs

3 MW

[Samsung SSC4](/facilities/fac-epoch-gc-samsung-ssc4)

KR

7.26 × 10²² — view full provenance

6.19 × 10²² — view full provenance

1.17×

16% normalized

2,496 GPUs

3 MW

[SoftBank Planned B200 Superpod](/facilities/fac-epoch-gc-softbank-planned-b200-superpod)

JP

2.64 × 10²³ — view full provenance

2.26 × 10²³ — view full provenance

1.17×

16% normalized

4,000 GPUs

8 MW

[Sustainable Metal Cloud Singapore Phase 2](/facilities/fac-epoch-gc-sustainable-metal-cloud-singapore-phase-2)

SG

1.45 × 10²³ — view full provenance

1.24 × 10²³ — view full provenance

1.17×

16% normalized

5,000 GPUs

7 MW

[Sweden 4k H100 Cluster](/facilities/fac-epoch-gc-sweden-4k-h100-cluster)

SE

1.16 × 10²³ — view full provenance

9.92 × 10²² — view full provenance

1.17×

16% normalized

4,000 GPUs

6 MW

[University of Bristol Isambard-AI](/facilities/fac-epoch-gc-university-of-bristol-isambard-ai)

GB

1.58 × 10²³ — view full provenance

1.35 × 10²³ — view full provenance

1.17×

16% normalized

5,448 GPUs

8 MW

[TACC Frontera](/facilities/fac-epoch-gc-tacc-frontera)

US

2.97 × 10²¹ — view full provenance

2.52 × 10²¹ — view full provenance

1.18×

16% normalized

808 GPUs

0 MW

[Princeton Della Phase 2](/facilities/fac-epoch-gc-princeton-della-phase-2)

US

5.61 × 10²¹ — view full provenance

6.64 × 10²¹ — view full provenance

1.18×

17% normalized

612 GPUs

1 MW

[NVIDIA Tethys](/facilities/fac-epoch-gc-nvidia-tethys)

US

1.47 × 10²¹ — view full provenance

1.23 × 10²¹ — view full provenance

1.19×

18% normalized

160 GPUs

0 MW

[Pawsey Supercomputing Centre Setonix](/facilities/fac-epoch-gc-pawsey-supercomputing-centre-setonix)

AU

8.64 × 10²¹ — view full provenance

7.12 × 10²¹ — view full provenance

1.21×

19% normalized

768 GPUs

1 MW

[G42 Microsoft 30 MW UAE Cluster A](/facilities/fac-epoch-gc-g42-microsoft-30mw-uae-cluster-a)

AE

4.36 × 10²³ — view full provenance

5.31 × 10²³ — view full provenance

1.22×

20% normalized

15,000 GPUs

30 MW

[G42 Microsoft 30 MW UAE Cluster B](/facilities/fac-epoch-gc-g42-microsoft-30mw-uae-cluster-b)

AE

4.36 × 10²³ — view full provenance

5.31 × 10²³ — view full provenance

1.22×

20% normalized

15,000 GPUs

30 MW

[Telangana Yotta Hyderbad AI City Cluster Phase 2](/facilities/fac-epoch-gc-telangana-yotta-hyderbad-ai-city-cluster-phase-2)

IN

7.27 × 10²³ — view full provenance

8.85 × 10²³ — view full provenance

1.22×

20% normalized

25,000 GPUs

50 MW

[NSTDA Supercomputer Center (ThaiSC) LANTA](/facilities/fac-epoch-gc-nstda-supercomputer-center-thaisc-lanta)

TH

6.45 × 10²¹ — view full provenance

5.29 × 10²¹ — view full provenance

1.22×

20% normalized

704 GPUs

1 MW

[Mare Nostrum 5](/facilities/fac-epoch-gc-mare-nostrum-5)

ES

1.30 × 10²³ — view full provenance

1.06 × 10²³ — view full provenance

1.23×

20% normalized

4,480 GPUs

6 MW

[Microsoft Goodyear](/facilities/fac-epoch-microsoft-goodyear)

US

2.91 × 10²⁴ — view full provenance

3.57 × 10²⁴ — view full provenance

1.23×

21% normalized

100,000 GPUs

202 MW

[KT Internal MI250 Cluster](/facilities/fac-epoch-gc-kt-internal-mi250-cluster)

KR

1.35 × 10²² — view full provenance

1.09 × 10²² — view full provenance

1.23×

21% normalized

1,200 GPUs

1 MW

[Oak Ridge NL Frontier](/facilities/fac-epoch-gc-oak-ridge-nl-frontier)

US

4.23 × 10²³ — view full provenance

3.42 × 10²³ — view full provenance

1.24×

21% normalized

37,632 GPUs

40 MW

[University of Illinois NCSA Delta](/facilities/fac-epoch-gc-university-of-illinois-ncsa-delta)

US

8.07 × 10²¹ — view full provenance

6.49 × 10²¹ — view full provenance

1.24×

22% normalized

880 GPUs

1 MW

[NFDG Andromeda Phase 2](/facilities/fac-epoch-gc-nfdg-andromeda-phase-2)

US

1.28 × 10²³ — view full provenance

1.03 × 10²³ — view full provenance

1.24×

22% normalized

4,400 GPUs

6 MW

[Foxconn Big Innovation Cloud AI factory](/facilities/fac-epoch-gc-foxconn-big-innovation-cloud-ai-factory)

TW

7.34 × 10²³ — view full provenance

5.82 × 10²³ — view full provenance

1.26×

23% normalized

10,000 GPUs

26 MW

[Microsoft Explorer-WUS3](/facilities/fac-epoch-gc-microsoft-explorer-wus3)

US

2.16 × 10²² — view full provenance

1.71 × 10²² — view full provenance

1.27×

24% normalized

1,920 GPUs

2 MW

[Reliance Industries Supercomputer](/facilities/fac-epoch-gc-reliance-industries-supercomputer)

IN

3.30 × 10²⁵ — view full provenance

2.61 × 10²⁵ — view full provenance

1.27×

24% normalized

450,000 GPUs

1,000 MW

[Applied Digital Ellendale Possible Phase 3](/facilities/fac-epoch-gc-applied-digital-ellendale-possible-phase-3)

US

1.32 × 10²⁵ — view full provenance

1.04 × 10²⁵ — view full provenance

1.27×

24% normalized

180,000 GPUs

400 MW

[NVIDIA DGX SuperPOD 2019](/facilities/fac-epoch-gc-nvidia-dgx-superpod-2019)

—

5.64 × 10²¹ — view full provenance

4.45 × 10²¹ — view full provenance

1.27×

24% normalized

1,536 GPUs

1 MW

[Eni HPC5](/facilities/fac-epoch-gc-eni-hpc5)

IT

2.67 × 10²² — view full provenance

2.10 × 10²² — view full provenance

1.27×

24% normalized

7,280 GPUs

4 MW

[Petrobras Atlas](/facilities/fac-epoch-gc-petrobras-atlas)

BR

4.00 × 10²¹ — view full provenance

5.11 × 10²¹ — view full provenance

1.28×

24% normalized

1,088 GPUs

1 MW

[Eni HPC6](/facilities/fac-epoch-gc-eni-hpc6)

IT

1.56 × 10²³ — view full provenance

1.21 × 10²³ — view full provenance

1.29×

25% normalized

13,888 GPUs

14 MW

[XTX Markets Cluster](/facilities/fac-epoch-gc-xtx-markets-cluster)

—

1.83 × 10²³ — view full provenance

1.42 × 10²³ — view full provenance

1.29×

26% normalized

20,000 GPUs

15 MW

[EuroHPC LUMI](/facilities/fac-epoch-gc-eurohpc-lumi)

FI

1.34 × 10²³ — view full provenance

1.04 × 10²³ — view full provenance

1.29×

26% normalized

11,912 GPUs

12 MW

[OpenAI/Microsoft Mt Pleasant, Wisconsin Phase 2](/facilities/fac-epoch-gc-openai-microsoft-mt-pleasant-wisconsin-phase-2)

US

5.14 × 10²⁵ — view full provenance

3.91 × 10²⁵ — view full provenance

1.31×

27% normalized

700,000 GPUs

1,500 MW

[Lawrence Livermore NL Tuolumne](/facilities/fac-epoch-gc-lawrence-livermore-nl-tuolumne)

US

1.34 × 10²³ — view full provenance

1.00 × 10²³ — view full provenance

1.34×

29% normalized

4,608 GPUs

6 MW

[Samsung SSC-21 Scalable Module](/facilities/fac-epoch-gc-samsung-ssc-21-scalable-module)

KR

1.32 × 10²¹ — view full provenance

1.77 × 10²¹ — view full provenance

1.34×

29% normalized

144 GPUs

0 MW

[G42 Artemis](/facilities/fac-epoch-gc-g42-artemis)

AE

4.76 × 10²¹ — view full provenance

6.39 × 10²¹ — view full provenance

1.34×

29% normalized

1,296 GPUs

1 MW

[Intel Stability Gaudi 2](/facilities/fac-epoch-gc-intel-stability-gaudi-2)

US

1.16 × 10²³ — view full provenance

8.66 × 10²² — view full provenance

1.34×

29% normalized

4,000 GPUs

5 MW

[Sandia NL El Dorado](/facilities/fac-epoch-gc-sandia-nl-el-dorado)

US

4.42 × 10²² — view full provenance

3.28 × 10²² — view full provenance

1.35×

29% normalized

1,520 GPUs

2 MW

[Sesterce Southern France 250MW](/facilities/fac-epoch-gc-sesterce-southern-france-250mw)

FR

8.81 × 10²⁴ — view full provenance

6.52 × 10²⁴ — view full provenance

1.35×

30% normalized

120,000 GPUs

250 MW

[Petrobras Fênix Phase 2](/facilities/fac-epoch-gc-petrobras-f-nix-phase-2)

BR

2.64 × 10²¹ — view full provenance

3.64 × 10²¹ — view full provenance

1.38×

32% normalized

720 GPUs

1 MW

[Nscale Loughton](/facilities/fac-epoch-gc-nscale-loughton)

GB

1.69 × 10²⁴ — view full provenance

2.35 × 10²⁴ — view full provenance

1.39×

32% normalized

23,040 GPUs

90 MW

[Tesla Cortex Phase 3](/facilities/fac-epoch-gc-tesla-cortex-phase-3)

US

3.49 × 10²⁴ — view full provenance

2.48 × 10²⁴ — view full provenance

1.41×

34% normalized

120,000 GPUs

140 MW

[Fluidstack France Gigawatt Campus](/facilities/fac-epoch-gc-fluidstack-france-gigawatt-campus)

FR

3.67 × 10²⁵ — view full provenance

2.61 × 10²⁵ — view full provenance

1.41×

34% normalized

500,000 GPUs

1,000 MW

[Nebius New Jersey](/facilities/fac-epoch-gc-nebius-new-jersey)

US

1.10 × 10²⁵ — view full provenance

7.82 × 10²⁴ — view full provenance

1.41×

34% normalized

150,000 GPUs

300 MW

[OpenAI/Microsoft Mt Pleasant, Wisconsin Phase 1](/facilities/fac-epoch-gc-openai-microsoft-mt-pleasant-wisconsin-phase-1)

US

1.10 × 10²⁵ — view full provenance

7.82 × 10²⁴ — view full provenance

1.41×

34% normalized

150,000 GPUs

300 MW

[Sesterce Grand Est France A](/facilities/fac-epoch-gc-sesterce-grand-est-france-a)

FR

1.10 × 10²⁵ — view full provenance

7.82 × 10²⁴ — view full provenance

1.41×

34% normalized

150,000 GPUs

300 MW

[Sesterce Grand Est France B](/facilities/fac-epoch-gc-sesterce-grand-est-france-b)

FR

1.10 × 10²⁵ — view full provenance

7.82 × 10²⁴ — view full provenance

1.41×

34% normalized

150,000 GPUs

300 MW

[NVIDIA Taipei-1](/facilities/fac-epoch-gc-nvidia-taipei-1)

TW

2.23 × 10²² — view full provenance

1.57 × 10²² — view full provenance

1.42×

35% normalized

768 GPUs

1 MW

[SURF Snellius Phase 3](/facilities/fac-epoch-gc-surf-snellius-phase-3)

NL

1.86 × 10²² — view full provenance

1.30 × 10²² — view full provenance

1.43×

35% normalized

640 GPUs

1 MW

[AGH Cyfronet Helios](/facilities/fac-epoch-gc-agh-cyfronet-helios)

PL

1.35 × 10²² — view full provenance

9.37 × 10²¹ — view full provenance

1.44×

36% normalized

464 GPUs

1 MW

[Center for Advanced Intelligence Project, RIKEN, RAIDEN](/facilities/fac-epoch-gc-center-for-advanced-intelligence-project-riken-raiden)

JP

1.53 × 10²¹ — view full provenance

1.05 × 10²¹ — view full provenance

1.46×

37% normalized

416 GPUs

0 MW

[Petrobras Gaia](/facilities/fac-epoch-gc-petrobras-gaia)

BR

6.45 × 10²¹ — view full provenance

9.54 × 10²¹ — view full provenance

1.48×

39% normalized

704 GPUs

1 MW

[Alps Supercomputer Phase 2](/facilities/fac-epoch-gc-alps-supercomputer-phase-2)

CH

3.13 × 10²³ — view full provenance

2.11 × 10²³ — view full provenance

1.48×

39% normalized

10,752 GPUs

12 MW

[Oak Ridge NL Summit](/facilities/fac-epoch-gc-oak-ridge-nl-summit)

US

1.02 × 10²³ — view full provenance

6.78 × 10²² — view full provenance

1.50×

40% normalized

27,648 GPUs

13 MW

[Los Alamos NL Venado](/facilities/fac-epoch-gc-los-alamos-nl-venado)

US

7.44 × 10²² — view full provenance

4.91 × 10²² — view full provenance

1.51×

41% normalized

2,560 GPUs

3 MW

[CEA EXA1-HE Phase 2](/facilities/fac-epoch-gc-cea-exa1-he-phase-2)

FR

5.55 × 10²² — view full provenance

3.57 × 10²² — view full provenance

1.55×

43% normalized

1,908 GPUs

2 MW

[Gemini 1.0 Ultra training cluster A](/facilities/fac-epoch-gc-gemini-1-0-ultra-training-cluster-a)

US

2.32 × 10²³ — view full provenance

3.63 × 10²³ — view full provenance

1.57×

44% normalized

28,672 GPUs

20 MW

[Microsoft Ares/Maia](/facilities/fac-epoch-gc-microsoft-ares-maia)

US

5.95 × 10²² — view full provenance

3.76 × 10²² — view full provenance

1.58×

45% normalized

2,048 GPUs

2 MW

[Google Oklahoma TPU v4 Pods](/facilities/fac-epoch-gc-google-oklahoma-tpu-v4-pods)

US

2.65 × 10²³ — view full provenance

4.25 × 10²³ — view full provenance

1.61×

47% normalized

32,768 GPUs

24 MW

[Paper on PaLM](/facilities/fac-epoch-gc-paper-on-palm)

US

4.96 × 10²² — view full provenance

7.97 × 10²² — view full provenance

1.61×

47% normalized

6,144 GPUs

4 MW

[Google TPU v4 Pod](/facilities/fac-epoch-gc-google-tpu-v4-pod)

US

3.31 × 10²² — view full provenance

5.36 × 10²² — view full provenance

1.62×

47% normalized

4,096 GPUs

3 MW

[SK Group AWS Uslan Phase 2](/facilities/fac-epoch-gc-sk-group-aws-uslan-phase-2)

KR

4.41 × 10²⁴ — view full provenance

2.69 × 10²⁴ — view full provenance

1.64×

49% normalized

60,000 GPUs

103 MW

[Google TPUv3 POD Generic](/facilities/fac-epoch-gc-google-tpuv3-pod-generic)

—

2.98 × 10²² — view full provenance

1.81 × 10²² — view full provenance

1.64×

49% normalized

1,024 GPUs

1 MW

[Google MLPerf 0.7 Submission](/facilities/fac-epoch-gc-google-mlperf-0-7-submission)

US

1.19 × 10²³ — view full provenance

7.12 × 10²² — view full provenance

1.67×

50% normalized

4,096 GPUs

4 MW

[Argonne NL Aurora](/facilities/fac-epoch-gc-argonne-nl-aurora)

US

1.85 × 10²⁴ — view full provenance

1.06 × 10²⁴ — view full provenance

1.75×

54% normalized

63,744 GPUs

60 MW

[University of Oxford Jade2](/facilities/fac-epoch-gc-university-of-oxford-jade2)

GB

1.85 × 10²¹ — view full provenance

1.06 × 10²¹ — view full provenance

1.75×

55% normalized

504 GPUs

0 MW

[LUMI Supercomputer (CSC Finland)](/facilities/fac-lumi-kajaani)

FI

1.15 × 10²³ — view full provenance

6.08 × 10²² — view full provenance

1.90×

62% normalized

10,240 GPUs

7 MW

[Eclairion Bruyères-le-Châtel (Mistral Compute)](/facilities/fac-eclairion-bruyeres-mistral-compute)

FR

4.01 × 10²³ — view full provenance

7.79 × 10²³ — view full provenance

1.94×

64% normalized

13,800 GPUs

44 MW

[Oak Ridge NL Titan](/facilities/fac-epoch-gc-oak-ridge-nl-titan)

US

5.43 × 10²³ — view full provenance

2.80 × 10²³ — view full provenance

1.94×

64% normalized

18,688 GPUs

16 MW

[Tesla Dojo 1 Phase 1](/facilities/fac-epoch-gc-tesla-dojo-1-phase-1)

US

8.72 × 10²² — view full provenance

4.25 × 10²² — view full provenance

2.05×

69% normalized

3,000 GPUs

2 MW

[Lawrence Livermore NL El Capitan Phase 2](/facilities/fac-epoch-gc-lawrence-livermore-nl-el-capitan-phase-2)

US

1.29 × 10²⁴ — view full provenance

6.19 × 10²³ — view full provenance

2.09×

71% normalized

44,544 GPUs

35 MW

[CSCS Piz Daint Phase 2](/facilities/fac-epoch-gc-cscs-piz-daint-phase-2)

CH

1.66 × 10²³ — view full provenance

7.74 × 10²² — view full provenance

2.14×

73% normalized

5,704 GPUs

4 MW

[Jupiter, Jülich](/facilities/fac-epoch-gc-jupiter-j-lich)

DE

6.84 × 10²³ — view full provenance

3.18 × 10²³ — view full provenance

2.15×

73% normalized

23,536 GPUs

18 MW

[ORNL Frontier (Exascale HPC)](/facilities/fac-ornl-frontier)

US

4.26 × 10²³ — view full provenance

1.94 × 10²³ — view full provenance

2.19×

75% normalized

37,888 GPUs

23 MW

[GSIC TSUBAME 3.0](/facilities/fac-epoch-gc-gsic-tsubame-3-0)

JP

6.27 × 10²² — view full provenance

2.70 × 10²² — view full provenance

2.32×

80% normalized

2,156 GPUs

2 MW

[US Government Supercomputer 1](/facilities/fac-epoch-gc-us-government-supercomputer-1)

US

1.21 × 10²³ — view full provenance

5.11 × 10²² — view full provenance

2.37×

81% normalized

4,160 GPUs

3 MW

[US Government Supercomputer 2](/facilities/fac-epoch-gc-us-government-supercomputer-2)

US

1.21 × 10²³ — view full provenance

5.11 × 10²² — view full provenance

2.37×

81% normalized

4,160 GPUs

3 MW

[Meta 2017 P100 Cluster](/facilities/fac-epoch-gc-meta-2017-p100-cluster)

US

2.88 × 10²² — view full provenance

1.19 × 10²² — view full provenance

2.42×

83% normalized

992 GPUs

1 MW

[Google TensorFlow Research Cloud](/facilities/fac-epoch-gc-google-tensorflow-research-cloud)

US

1.19 × 10²³ — view full provenance

4.76 × 10²² — view full provenance

2.50×

86% normalized

4,096 GPUs

3 MW

[Jean Zay Supercomputer Phase 4](/facilities/fac-epoch-gc-jean-zay-supercomputer-phase-4)

FR

1.08 × 10²³ — view full provenance

4.28 × 10²² — view full provenance

2.52×

86% normalized

3,704 GPUs

2 MW

[GSIC TSUBAME 2.5](/facilities/fac-epoch-gc-gsic-tsubame-2-5)

JP

1.23 × 10²³ — view full provenance

4.76 × 10²² — view full provenance

2.58×

88% normalized

4,224 GPUs

3 MW

[Preferred Networks MN-1b](/facilities/fac-epoch-gc-preferred-networks-mn-1b)

JP

4.46 × 10²² — view full provenance

1.66 × 10²² — view full provenance

2.69×

91% normalized

1,536 GPUs

1 MW

[Moscow State University Lomonosov 2](/facilities/fac-epoch-gc-moscow-state-university-lomonosov-2)

RU

4.28 × 10²² — view full provenance

1.52 × 10²² — view full provenance

2.81×

95% normalized

1,472 GPUs

1 MW

[Eni HPC2](/facilities/fac-epoch-gc-eni-hpc2)

IT

8.72 × 10²² — view full provenance

2.50 × 10²² — view full provenance

3.49×

111% normalized

3,000 GPUs

1 MW

[Aramco Groq Inference Cluster](/facilities/fac-epoch-gc-aramco-groq-inference-cluster)

SA

5.73 × 10²³ — view full provenance

1.50 × 10²³ — view full provenance

3.82×

117% normalized

19,725 GPUs

8 MW

[MIT LLSC TX-GAIA](/facilities/fac-epoch-gc-mit-llsc-tx-gaia)

US

3.29 × 10²¹ — view full provenance

1.34 × 10²² — view full provenance

4.08×

121% normalized

896 GPUs

3 MW

[Universitaet Frankfurt Goethe-NHR](/facilities/fac-epoch-gc-universitaet-frankfurt-goethe-nhr)

DE

2.56 × 10²² — view full provenance

5.88 × 10²¹ — view full provenance

4.35×

125% normalized

880 GPUs

0 MW

[JAMSTEC ZettaScaler-2.2 Gyoukou](/facilities/fac-epoch-gc-jamstec-zettascaler-2-2-gyoukou)

JP

2.91 × 10²³ — view full provenance

4.60 × 10²² — view full provenance

6.32×

145% normalized

10,000 GPUs

3 MW

[Paper on AlphaZero](/facilities/fac-epoch-gc-paper-on-alphazero)

US

1.45 × 10²³ — view full provenance

1.56 × 10²² — view full provenance

9.33×

161% normalized

5,000 GPUs

1 MW

What this ranking means

-   Tight bounds (ratio ≤ 1.5×): the two independent inputs reconcile under the methodology defaults. That is a credibility signal with a known blind spot: a facility where both inputs are wrong in the same direction (both nameplate-only, both phase-1) reconciles tightly without being right.
-   Loose bounds (ratio > 3×) usually point to a gap in the record: a GPU model unknown but power disclosed (the H100 SXM5 default does not match deployed hardware), nameplate power capacity oversized for actual draw, or a partial GPU count (a phase-1 figure against a full-build power capacity); the loose rows are this register's investigation queue.
-   The cost path is excluded at the facility level: Scrutica's records have no per-facility total-investment figure (organization-level capex is annualized and pooled across facilities). The Capacity Estimator above runs all three paths against user-supplied inputs; this ranking is hardware-vs-power only.

Each facility's estimate runs the accelerator its own record discloses: 285 of 314 reconciled sites resolve a specific chip from the per-facility hardware inventory, and the remaining 29 fall back to the H100 SXM5 default because no recognisable model is recorded. Both paths apply the SAME chip for a given site (the power path anchors its TDP to it), so a divergence reconciles the disclosed pair (chip count, megawatts) against the shared constants, with no hardware mismatch between the two readings. Throughput cancels between the paths and TDP does not, so a row on the fallback spec has a divergence conditional on that assumption; those rows are the ones a disclosed chip model would move.

Model archaeology

### The seven strata under a FLOP figure

Everything above compresses a facility to a number. These are the seven physical strata that produce it (raw silicon, the fab, the accelerator, the cluster, the power drawn), dug down one model scale at a time, each layer cited.

Continue

[Thresholds](/capacity/thresholds) reads the same estimate against the lines that carry legal consequence: which of these sites hold enough compute to put a model across the EU AI Act’s 10²⁵ trigger, and how fast.