🧠 Largest AI Supercomputers
The biggest operational AI supercomputers, measured in NVIDIA H100-equivalent chips, with their owners.
1
🇺🇸xAI Colossus Memphis Phase 3xAI
275,796
2
🇺🇸Meta 100kMeta AI
100,000
3
🇺🇸OpenAI/Microsoft Goodyear ArizonaMicrosoft, OpenAI
100,000
4
🇺🇸Oracle OCI Supercluster H200sOracle
65,536
5
🇺🇸Tesla Cortex Phase 1Tesla
50,000
6
🇺🇸Lawrence Livermore NL El Capitan Phase 2US Department of Energy
44,143
7
🇺🇸CoreWeave H200sCoreWeave
42,000
8
🇺🇸Meta GenAI 2024aMeta AI
24,576
9
🇺🇸Meta GenAI 2024bMeta AI
24,576
10
🇩🇪"Jupiter, Jülich"EuroHPC JU, Julich Supercomputing Center
23,536
11
🇺🇸Oracle OCI MI300xOracle
21,649
12
🇳🇴NexGen Cloud Hyperstack AQ Compute SupercomputerNexGen Cloud
16,384
13
🇺🇸Oracle OCI Supercluster H100sOracle
16,384
14
🇺🇸Microsoft Azure EagleMicrosoft
14,400
15
🇨🇭Alps Supercomputer Phase 2ETH Domain
10,752
16
🇺🇸NVIDIA CoreWeave Eos-DFW Phase 1Nvidia, CoreWeave
10,752
17
🇺🇸Oracle OCI Supercluster A100sOracle
10,332
18
🇨🇳Anonymized Chinese System
10,000
19
🇨🇳Anonymized Chinese System #2
10,000
20
🇺🇸Imbue 10k ClusterImbue
10,000
21
🇺🇸Tesla 10k H100 ClusterTesla
10,000
22
🇦🇪Sesterce NordicsSesterce
8,192
23
🇨🇳Anonymized Chinese System #3
8,000
24
🇺🇸Magic G4 Google Cloud RentalGoogle
8,000
25
🇫🇮Nebius 8k Finland Phase 1Nebius AI
8,000
Source: Epoch AI, GPU Clusters · CC BY 4.0 (Epoch AI) · fetched Oct 14, 2026 — Operational clusters only. Compute is expressed in H100-equivalents; some clusters are estimates (Epoch marks them 'Likely').