What Is an AI Data Center?

The simple definition

An AI data center is a facility built to train and run AI models. Instead of rows of ordinary servers, it's packed with racks of GPUs and other AI accelerators, chips designed to do enormous numbers of calculations in parallel. Some AI data centers mainly train new models, which takes months of continuous computing. Others mainly run inference, answering the prompts people send every second.

How it's different from a regular data center

• Power density. Many traditional server racks draw under 10 kilowatts. AI racks commonly draw 50 to 70 kilowatts, and the newest AI systems can exceed 200 kilowatts per rack (Data Center Knowledge, Schneider Electric).
• Cooling. Air cooling stops being practical at those densities, so AI data centers increasingly pipe liquid directly to the chips through cold plates.
• Networking. Thousands of GPUs have to work as one machine, linked by very fast connections, so layout and cabling matter more than in a typical facility.
• Scale. AI campuses are planned in hundreds of megawatts, and the largest proposals in gigawatts, as much power as a small city.

How much energy do they use?

Data centers of all kinds used about 415 terawatt-hours of electricity worldwide in 2024, around 1.5% of global electricity, and the International Energy Agency expects that to roughly double by 2030, with AI the main driver (IEA). In the US, data centers used about 4.4% of the country's electricity in 2023 and could reach 6.7% to 12% by 2028 (US Department of Energy). The per-prompt picture, including water use, is in Is AI bad for the environment?

Where they get built

AI data centers go where they can get large amounts of reliable power, land, fiber connections, and approvals. That's why they cluster in certain regions, with Northern Virginia the largest market, and why new projects are spreading to states with available power and land. Because they need so much electricity, the power supply now often decides the location more than anything else.

Is there an AI data center near me?

The best sources are local: news coverage, your county's planning and zoning agendas, and your utility's filings with the state regulator, where large new power requests show up. Proposed projects usually go through public hearings, which is the time to ask about power, water, noise, and tax deals.

The local trade-offs

• For the area: large construction investment, property tax revenue, and grid upgrades.
• Against: relatively few permanent jobs for the size, strain on the power grid, water use in dry regions, noise, and land use.

Whether a project is a good deal depends on the details, like where the power comes from, how it's cooled, and what the community gets in return. I've sketched two speculative alternatives in the Lab: cooling a data center with rainwater and burying one under power lines, cooled by the soil.