Mega-Scale AI Campus Distribution
Data Source: Compiled from public filings, satellite imagery, and energy permits
| Rank | Owner | Location | Hardware | Chip Count | Current Power (MW) | Max Planned (MW) | Comments |
|---|
| 1 | Google | Columbus, OH | TPU (Multi-gen) | 200,000 | 500 | 1000 | Uses multi-datacenter training. |
| 2 | Google | Omaha, NE | TPU (Latest) | 200,000 | 500 | 1000 | Dispersed campus connected by fiber. |
| 3 | Meta | Columbus, OH | Mixed / GPU | - | 500 | - | Hybrid site using standard buildings. |
| 4 | AWS | New Carlisle, IN | Trainium 2 | 500,000 | 420 | 2000 | Project Rainier scale. |
| 5 | Microsoft | Atlanta, GA | Nvidia GB200 | 150,000 | 350 | 700 | "Fairwater" type campus. |
| 6 | xAI | Memphis, TN | Nvidia GB200 | 110,000 | 400 | 1000 | Colossus 2; rapid construction. |
| 7 | Microsoft | Mt. Pleasant, WI | Nvidia GB200 | 150,000 | 350 | 2000 | Massive expansion plans. |
| 8 | AWS | Canton, MS | Trainium 2 | 200,000 | 300 | 1000 | Located next to fulfillment center. |
| 9 | OpenAI | Abilene, TX | Nvidia GB200 | 100,000 | 200 | - | Project Stargate ready by 2026. |
| 10 | xAI | Memphis, TN | Hopper / Blackwell | 230,000 | 300 | 300 | Colossus 1 factory conversion. |