AI data centers in the U.S.

How this count works

No government agency publishes a count of AI data centers, and there is no single definition. The International Energy Agency says a hyperscale, AI-focused data center can be 100 MW or more[1], and Executive Order 14318 defines a “Data Center Project” as “a facility that requires greater than 100 megawatts (MW) of new load dedicated to AI inference, training, simulation, or synthetic data generation”[2]. What is an AI data center? explains how they differ from conventional ones.

The Census counts a project as an AI data center only when its own record shows it: a cited source whose headline describes that project as AI, a dated fact or quoted passage from a cited source that does, or a developer or tenant whose business is building AI models (OpenAI, Anthropic or xAI), named in a cited headline. A headline about a region’s AI boom in general does not count. Each project’s evidence is in the table above, and the classification is published in the open data.

That makes this a floor. Many large cloud campuses run AI workloads without their sources saying so, and they are not counted until a source does. When a record gains that evidence, the count updates on the next build. Paused, cancelled and denied projects stay on the list so the record is complete, but they are not in the active count.

Reviewed but not counted

• QTS Fayetteville data center campus: Its AI evidence is the Microsoft Fairwater building on the campus, which the Census tracks as its own record (Microsoft Fairwater Atlanta). Counting both would count that building twice.
• Google Kansas City Data Center (Hunt Midwest Business Center): The only AI reference is a headline about Kansas City's data center boom in general, not this project.
• Project Mica: The only AI reference is a headline about Kansas City's data center boom in general, not this project.
• Eco-Yotta Oneonta Data Center: The only AI reference is a headline about the town's moratorium on AI data centers generally, not a description of this project.
• Viridian Data Center Campus (former GM/JATCO site): The headline ties the site's redevelopment interest to AI in general terms without describing this project's use.
• Global AI Data Center: AI appears only in the developer's company name; no cited source in the record states what the campus will run.
• Meta El Paso Data Center: The record's summary calls it an AI data center, but no cited headline, dated event or quoted passage in the record says so yet.
• Port Washington (Lighthouse) Data Center Campus: The record names OpenAI as a tenant, but no cited headline, dated event or quoted passage in the record says so yet.
• Project Tembo: The only AI description is the 2025 announcement by Crusoe and Tallgrass, who have since left the project; no cited source describes Google's current plan as AI.
• Cloverleaf, Village of Wrightstown: Its only AI reference is a headline about Cloverleaf's separate, withdrawn Greenleaf proposal.
• Microsoft Mega Park: Its only AI reference is a headline about AI tensions in general; no cited source describes this campus as built for AI.
• Fleet Data Centers TRIC campuses (South Valley and Peru Ridge): Its only AI reference is a headline about who should pay for the industry's AI buildout in general; no cited source describes the South Valley or Peru Ridge campuses themselves as built for AI.
• Stillwater Technology Park: Cancelled; no cited source describes this project itself as built for AI. The only AI reference is the title of a roundup about OpenAI's Georgia plans.
• Cologix Johnstown Campus: Sources call the buildings AI-ready, but no cited headline or company statement says the campus is built for a named AI workload.
• Hyperscale Data Dowagiac Campus: The site mines cryptocurrency today; the company says it will shift to AI computing and announced a 20 MW AI compute agreement, but no AI capacity is operating yet.

Put this count on your site

The live count is available as a free widget for articles, council pages and newsletters. It updates itself when the record changes and credits the Census. Get the embed code →

Related: How much water does AI use? · How many data centers are in the U.S.? · AI data center news · Largest projects · Companies

Sources

1. Understanding the energy-AI nexus – Energy and AI
   International Energy Agency · 2025-04-10
   Source excerpt

   “A hyperscale, AI-focused data centre can have a capacity of 100 MW or more, consuming as much electricity annually as 100 000 households.”

2. Accelerating Federal Permitting of Data Center Infrastructure
   The White House · 2025-07-23
   Source excerpt

   “a facility that requires greater than 100 megawatts (MW) of new load dedicated to AI inference, training, simulation, or synthetic data generation. ... a Data Center Project or Covered Component Project for which the Project Sponsor has committed at least $500 million in capital expenditures as determined by the Secretary of Commerce”