Top 50
The largest AI data centers in the world
Ranked by installed AI compute, measured in H100-equivalents.
Colossus 2 leads with 1.1M H100e, ahead of Microsoft Fairwater Atlanta (768.8K H100e) and Meta Prometheus (763K H100e).
The top five account for 31% of the 12.2M H100e tracked across this ranking.
Compute capacity is the metric that matters for training: it is what decides how large the next frontier run can be. None of these figures is disclosed by the operators — they are reconstructed from chip orders, construction permits, satellite imagery and power contracts, and carry roughly 1.4x to 1.6x of uncertainty.
| # | Name | Owner | Country | Compute |
|---|---|---|---|---|
| 01 | Colossus 2 | SpaceXAI | United States | 1.1M H100e |
| 02 | Microsoft Fairwater Atlanta | Microsoft | United States | 768.8K H100e |
| 03 | Meta Prometheus | Meta | United States | 763K H100e |
| 04 | Anthropic-Amazon New Carlisle | Amazon | United States | 685.9K H100e |
| 05 | OpenAI Stargate Abilene | Oracle | United States | 509.3K H100e |
| 06 | Microsoft Fairwater Wisconsin | Microsoft | United States | 445.7K H100e |
| 07 | Google Columbus | United States | 408.8K H100e | |
| 08 | Google New Albany | United States | 351.7K H100e | |
| 09 | Google Council Bluffs (East) | United States | 335K H100e | |
| 10 | Google Lincoln | United States | 287.5K H100e | |
| 11 | Google Bristow | United States | 283.7K H100e | |
| 12 | DayOne Nusajaya | — | Malaysia | 281.7K H100e |
| 13 | Colossus 1 | SpaceXAI | United States | 275.8K H100e |
| 14 | Google Omaha | United States | 267.8K H100e | |
| 15 | CoreWeave Denton TX | CoreWeave | United States | 252.7K H100e |
| 16 | QTS Richmond 1 | — | United States | 242.5K H100e |
| 17 | Meta Rosemount | Meta | United States | 215.8K H100e |
| 18 | Amazon Madison Mega Site | Amazon | United States | 214.3K H100e |
| 19 | Microsoft Goodyear | Microsoft | United States | 205.2K H100e |
| 20 | Meta Jeffersonville | Meta | United States | 203.1K H100e |
| 21 | Google The Dalles | United States | 201.6K H100e | |
| 22 | Google Midlothian | United States | 191K H100e | |
| 23 | QTS Richmond 2 | — | United States | 182.9K H100e |
| 24 | QTS Richmond 3 | — | United States | 174.8K H100e |
| 25 | Meta Montgomery | Meta | United States | 174.5K H100e |
| 26 | Meta Kuna | Meta | United States | 173.3K H100e |
| 27 | Meta Temple | Meta | United States | 173.2K H100e |
| 28 | Meta Cheyenne | Meta | United States | 172.8K H100e |
| 29 | Amazon Ridgeland | Amazon | United States | 171.3K H100e |
| 30 | Meta Aiken | Meta | United States | 161.7K H100e |
| 31 | Coreweave Helios | CoreWeave | United States | 159.6K H100e |
| 32 | Google Arcola | United States | 157.2K H100e | |
| 33 | Microsoft Project Osmium | Microsoft | United States | 155.6K H100e |
| 34 | Google Storey County | United States | 153.6K H100e | |
| 35 | Meta Eagle Mountain | Meta | United States | 151.4K H100e |
| 36 | Google Lancaster | United States | 130.4K H100e | |
| 37 | Fluidstack Lake Mariner | Fluidstack | United States | 130.4K H100e |
| 38 | Google Mesa | United States | 125.8K H100e | |
| 39 | Google Waltham Cross | United Kingdom | 123.8K H100e | |
| 40 | Huawei Horinger | Huawei | China | 123.8K H100e |
| 41 | VNET Bayin Ulanqab | VNET | China | 122.8K H100e |
| 42 | Google Red Oak | United States | 108.1K H100e | |
| 43 | Meta Gallatin | Meta | United States | 99.5K H100e |
| 44 | Meta Los Lunas | Meta | United States | 99.5K H100e |
| 45 | CoreWeave Ellendale ND | CoreWeave | United States | 93.5K H100e |
| 46 | CoreWeave Chester VA | CoreWeave | United States | 88.4K H100e |
| 47 | Oracle Batam | Oracle | Indonesia | 84.4K H100e |
| 48 | Google Papillion | United States | 80.3K H100e | |
| 49 | CoreWeave Marble NC | CoreWeave | United States | 78.8K H100e |
| 50 | Microsoft SAT40 | Microsoft | United States | 64.7K H100e |
SourceEpoch AI, 'AI Models'. Published online at epoch.ai. Retrieved 2026-07-29 from https://epoch.ai/data/ai-models. Licensed under CC BY 4.0.
Browse all ai data centers →Other rankings of ai data centers