Big tech AI clusters: Meta, Microsoft, AWS, Tesla and others

21 clusters. 17 are running or partly running. Installed GPUs on record add up to 133,952, with a further 1,002,048 planned across 2 clusters. The largest is Meta Prometheus (1,000,000 GPUs planned).

The hyperscalers and large platform companies build the biggest training clusters, but publish counts mainly in engineering posts and earnings material, so some rows carry a model and a power figure but no installed total.

Last verified

Cluster Owner / operator GPU model GPUs installed GPUs planned Location Status
Meta GenAI InfiniBand cluster Meta NVIDIA H100 24,576 — — operational
Meta GenAI RoCE cluster Meta NVIDIA H100 24,576 — — operational
Project Ceiba AWS NVIDIA GB200 (Blackwell) 20,736 — — operational
Meta AI Research SuperCluster (RSC) Meta NVIDIA A100 16,000 — — operational
Microsoft Azure Eagle Microsoft NVIDIA H100 14,400 — — operational
ByteDance MegaScale Ampere cluster ByteDance NVIDIA Ampere 12,288 — — operational
NVIDIA Eos (10,752-GPU system) NVIDIA NVIDIA H100 10,752 — — operational
NVIDIA Eos (4,608-GPU system) NVIDIA NVIDIA H100 4,608 — — operational
NAVER Sejong NAVER NVIDIA A100 2,240 — Sejong, South Korea operational
Yandex Chervonenkis Yandex NVIDIA A100 80GB 1,592 — Sasovo, Ryazan Oblast, Russia not stated
Yandex Lyapunov Yandex NVIDIA A100 40GB 1,096 — Sasovo, Ryazan Oblast, Russia not stated
Yandex Galushkin Yandex NVIDIA A100 80GB 1,088 — Vladimir, Vladimir Oblast, Russia not stated
Meta Prometheus Meta NVIDIA GB200 — 1,000,000 New Albany, Ohio, United States partly operational
NVIDIA Israel-1 NVIDIA NVIDIA H100 80GB — 2,048 Israel partly operational
IBM Blue Vela IBM NVIDIA H100 — — — operational
Meta Hyperion Meta — — — Richland Parish, Louisiana, United States under construction
Microsoft Fairwater Atlanta Microsoft NVIDIA GB200 NVL72 — — Atlanta area, Georgia, United States operational
Microsoft Fairwater Wisconsin Microsoft NVIDIA GPUs — — Mount Pleasant, Wisconsin, United States partly operational
SoftBank Blackwell DGX SuperPOD SoftBank Corp. NVIDIA B200 — — — operational
Tesla Cortex 1 Tesla — — — Gigafactory Texas, Austin, Texas, United States operational
Tesla Cortex 2 Tesla — — — Gigafactory Texas, Austin, Texas, United States operational

Other lists