Compute per power capacity
Sum H100e / sum MW over the same Existing clusters with both positive values. Calculated capacity takes precedence over disclosed capacity. This is a capacity ratio, not measured efficiency or PUE.
Compute per power capacity
Total 404 · Page 16/17 · Observed 9/25/2026, 3:30:45 AM
| Name | Group | Reported value | Date | Notes | Source |
|---|---|---|---|---|---|
| TACC Frontera | United States of America | 67.09 H100e/MW | 2019-08-31 | University of Texas at Austin | Source ↗ |
| Google MLPerf 0.7 Submission | United States of America | 63.24 H100e/MW | 2020-07-29 | Source ↗ | |
| Jean Zay Supercomputer Phase 3 | France | 63.17 H100e/MW | 2022-06-01 | GENCI | Source ↗ |
| Eni HPC4 Phase 2 | Italy | 39.68 H100e/MW | 2021-10-30 | Eni | Source ↗ |
| Google TensorFlow Research Cloud | United States of America | 35.36 H100e/MW | 2017-05-17 | Source ↗ | |
| Anonymized Chinese System | China | 30 H100e/MW | 2017-11-15 | — | — |
| Anonymized Chinese System | China | 30 H100e/MW | 2017-11-15 | — | — |
| Argonne NL Aurora | United States of America | 21.66 H100e/MW | 2024-05-13 | US Department of Energy | Source ↗ |
| Anonymized Chinese System | China | 20 H100e/MW | 2012-03-15 | — | — |
| JAMSTEC ZettaScaler-2.2 Gyoukou | Japan | 19.18 H100e/MW | 2017-11-15 | Japan Agency for Marine-Earth Science and Technology | Source ↗ |
| JAMSTEC ZettaScaler-2.0 Gyoukou | Japan | 19.18 H100e/MW | 2017-06-15 | Japan Agency for Marine-Earth Science and Technology | Source ↗ |
| Anonymized Chinese System | China | 16.67 H100e/MW | 2020-11-15 | — | — |
| Eni HPC4 Phase 1 | Italy | 16.48 H100e/MW | 2018-01-18 | Eni | Source ↗ |
| CSCS Piz Daint Phase 2 | Switzerland | 16.4 H100e/MW | 2017-06-01 | ETH Domain | Source ↗ |
| CSCS Piz Daint Phase 1 | Switzerland | 16.15 H100e/MW | 2016-11-01 | ETH Domain | Source ↗ |
| Meta 2017 P100 Cluster | United States of America | 16.1 H100e/MW | 2017-06-01 | Meta AI | Source ↗ |
| NVIDIA SATURN V Phase 1 | United States of America | 15.85 H100e/MW | 2016-11-01 | Nvidia | Source ↗ |
| GSIC TSUBAME 3.0 | Japan | 15.22 H100e/MW | 2017-11-01 | Tokyo Institute of Technology | Source ↗ |
| Anonymized Chinese System | China | 15 H100e/MW | 2020-09-15 | — | — |
| Anonymized Chinese System | China | 6 H100e/MW | 2016-06-15 | — | — |
| Energy Company DD | United States of America | 5.74 H100e/MW | 2016-06-15 | NVIDIA Tesla K80 | Source ↗ |
| US Government Supercomputer 2 | United States of America | 5.2 H100e/MW | 2015-06-01 | United States Government | Source ↗ |
| US Government Supercomputer 1 | United States of America | 5.2 H100e/MW | 2014-11-01 | United States Government | Source ↗ |
| Anonymized Chinese System | China | 4.5 H100e/MW | 2017-11-15 | — | — |
| Moscow State University Lomonosov 2 | Russia | 4.37 H100e/MW | 2016-06-15 | Moscow State Univeristy | Source ↗ |
Data: Epoch AI · CC BY · View source · Automatically synced; not an exhaustive global sample