Hyperscale and AI-focused datacenters. GPU cluster estimates and power capacity.
| Facility | Operator | Location | Type | GPU / Accelerator | Est. GPUs | Est. Power | Status | CapEx |
|---|---|---|---|---|---|---|---|---|
| Microsoft Wisconsin — GB200 Cluster | Microsoft | Mount Pleasant, United States | AI Cluster | GB200 NVL72 | ~60,000 | 700 MW | Under Construction | $8.0B |
| Azure West US 3 — AI Cluster | Microsoft | Phoenix, United States | AI Cluster | H100 | ~50,000 | 500 MW | Operational | $5.0B |
| AWS Oregon — AI Training Hub | Amazon (AWS) | Boardman, United States | AI Cluster | H200 | ~40,000 | 420 MW | Operational | $4.5B |
| Google The Dalles — TPU v5 | The Dalles, United States | Hyperscale | TPU v5e | — | 350 MW | Operational | $3.2B | |
| Meta Eagle Mountain — AI Cluster | Meta | Eagle Mountain, United States | AI Cluster | H100 | ~35,000 | 350 MW | Operational | $3.8B |
| AWS Singapore — APAC AI | Amazon (AWS) | Singapore, Singapore | AI Cluster | H100 | ~20,000 | 180 MW | Operational | $2.2B |
| xAI Memphis — Colossus Cluster | xAI | Memphis, United States | AI Cluster | H100 | ~100,000 | 150 MW | Operational | $2.0B |
| Google Belgium — EU AI Hub | Saint-Ghislain, Belgium | Hyperscale | TPU v5p | — | 300 MW | Expanding | $2.4B |
Microsoft
1,200 MW est.
~110,000 GPUs est.
2 facilities
Amazon (AWS)
600 MW est.
~60,000 GPUs est.
2 facilities
650 MW est.
2 facilities
Meta
350 MW est.
~35,000 GPUs est.
1 facility
xAI
150 MW est.
~100,000 GPUs est.
1 facility
* All GPU counts, power figures and CapEx estimates are illustrative demo data. Sourced from public announcements and analyst estimates (approximated). Not for investment use.