Hyperscale AI Data Centers

Frequently asked questions

What counts as a hyperscale AI data center?

This index defines hyperscale AI data centers as facilities with announced capacity of 500 MW or more. At that scale, a single campus rivals the electricity consumption of a small city and is purpose-built to run AI training or inference workloads at continuous high utilization. The 90 facilities in this view represent only 25% of tracked facilities but account for 77% of total tracked capacity.

How many gigawatt-scale AI data centers exist?

The AI Data Center Index tracks 46 facilities with announced capacity of 1 GW (1,000 MW) or more. These represent a historic step change in infrastructure scale — each is larger than most national data center markets were just five years ago. 5 facilities have announced capacities of 5 GW or above, including the largest sovereign AI campuses and hyperscaler master-planned developments.

Who builds the most hyperscale AI data centers?

The leading hyperscale builders by facility count are Microsoft (18), Amazon Web Services (16), Meta (9). At hyperscale, a handful of hyperscalers account for most announced capacity — the concentration is significantly higher than in the broader AI data center market.

Where are the largest AI data centers being built?

The countries with the most hyperscale AI data centers are United States (52), Brazil (3), Canada (3). The United States dominates through hyperscaler concentration in Virginia, Texas, and the Pacific Northwest. Emerging markets — Saudi Arabia, UAE, Japan, and India — are rapidly scaling with sovereign and hyperscaler-backed campuses.