Data centers—vast facilities that power the artificial intelligence revolution—are rising faster than many realize, often behind closed doors. To bring transparency to this rapid and largely opaque buildout, Epoch AI, a nonprofit research institute, has launched the Frontier Data Centers map, an open database that tracks the location, power use and construction progress of large AI facilities using satellite imagery and permitting data.
According to Epoch, the tool focuses on gigawatt-scale centers—those capable of sustaining the colossal computing demands of today’s generative AI systems. Many tracked facilities are designed to reach full capacity within two years, with construction-to-operation timelines ranging from just over a year to about 3.6 years. The group follows three milestones for each project: initial site work, the first sign of live electrical load and the point at which the center reaches one gigawatt of total power.
Such data adds context to a shifting landscape where progress isn’t always straightforward. In Wilmington, Ohio, for instance, the local planning commission recently tabled Amazon Web Services’ proposed data center due to “unanswered questions,” the area's new station, WKRC-TV, reported. Elsewhere, so-called “phantom data centers” have sought energy allocations without firm construction plans, further clouding the picture of where development is truly occurring.
The map also illustrates how quickly facilities can evolve. Meta’s Prometheus center, Epoch AI noted, began as a conventional data center in 2017 but pivoted in early 2025 to support frontier AI workloads—driving a power surge from 319 megawatts to more than 1,360 megawatts by late 2026.
Epoch AI’s data is freely accessible on its website as downloadable CSV files or interactive visualizations, offering one of the clearest looks yet at an industry racing to keep up with AI’s accelerating demands—though the group cautions that its focus on the largest projects means the dataset is not comprehensive.
© Arc, All Rights Reserved. Request academic re-use from www.copyright.com. All other uses, submit a request to TMSalesOperations@arc-network.com. For more information visit Asset & Logo Licensing.