How much electricity do America's data centers actually use? Explore every state's consumption, see where the AI building boom is headed, and find out what it means for electricity rates.
Click any state for full details. Toggle views to explore consumption, density, or city-level heatmap. Zoom and pan to navigate.
Toggle between three views using the controls above the map.
Total electricity consumed by data centers in each state per year, per LBNL estimates. Virginia leads at 24 TWh, followed by Texas (17) and Illinois (12). (Note: some analyses, such as EPRINC, estimate Virginia's share as high as 32 TWh depending on methodology.)
Number of data center facilities in each state. Virginia leads with 665, followed by Texas (413) and California (321).
City-level power usage as a heat gradient. Reveals that Ashburn, VA alone uses more electricity than most entire states.
1 TWh is one trillion watt-hours of electricity — here's what that looks like.
U.S. data centers use 176 TWh per year — enough to power 16 million homes.
Ranked by electricity consumption. Virginia's "Data Center Alley" alone uses more power than many countries.
Select two states to see a head-to-head comparison of their data center footprints.
Complete breakdown of every state. Click column headers to sort. Click a row to see details.
| # ▲ | State | Data Centers | TWh/Year | Major Hubs & Operators |
|---|
The AI boom is driving unprecedented data center construction. Here are the biggest projects announced or underway.
U.S. data center electricity demand has tripled in the past decade — and is projected to double again by 2028.
Data center growth is reshaping electricity markets across the country.
In Virginia, data centers now consume more than 1 in 4 kilowatt-hours of the state's electricity — an estimated 32 TWh out of 128 TWh total in 2023, per an EPRINC analysis. (At Dominion Energy specifically, data centers accounted for 21% of sales as of late 2022.) That kind of demand can strain the grid — and when utilities build new infrastructure to keep up, everyone's rates can go up.
Counterpoint: Dominion Energy has stated that data centers "do not currently have an outside influence on customer energy bills." Virginia's Joint Legislative Audit and Review Commission (JLARC) agreed in a December 2024 study that data centers are currently paying their fair share — but warned residential customers could begin bearing some costs if changes aren't made. Dominion has since created a dedicated rate class (GS-5) for data center customers over 25 MW.
In power-rich areas like West Texas and Indiana, the opposite happens. More large customers sharing fixed grid costs can lower rates for residents. Amazon's Indiana project alone is expected to save local households $1 billion over 15 years, with NIPSCO confirming Amazon will cover all infrastructure build-out costs without passing expenses to existing customers.
Water claims around data centers have become highly politicized. Some are legitimate concerns; others are misleading statistics that have been repeated uncritically for years. Here's what the evidence actually says.
This claim is a misreading of a UC Riverside study by Shaolei Ren, which found that 20–50 ChatGPT queries use about 500 ml — not one query. That works out to roughly 10–25 ml per query (a few teaspoons). OpenAI CEO Sam Altman put the figure even lower at ~0.32 ml per query (~1/15th of a teaspoon), though he didn't cite a source. The real number depends on the data center's cooling method and location, but "a bottle per query" was never what the research said.
Many widely-cited figures combine "direct" on-site cooling water with "indirect" water consumed upstream by power plants generating the data center's electricity. According to HyperFRAME Research, roughly 80% of a data center's total water footprint comes from the power grid, not the facility itself. That's real water, but attributing it to "the data center" is like blaming your house for the water used at the power plant that keeps your lights on. Direct U.S. data center water use is approximately 17 billion gallons per year — about 0.3% of public water supply (AEI).
Nationally, data center water use is a rounding error — but national averages mask local strain. In The Dalles, Oregon, Google's water use grew 316% while the town's population grew 12%. That's a real impact. Drought, aging infrastructure, and agriculture all contribute — but two-thirds of new hyperscale campuses built since 2022 are in high water-stress counties, and 1 in 5 servers sit in water-stressed watersheds. The concern isn't about the national picture; it's about where the water is being drawn.
Google replenished 4.5 billion gallons of water in 2024. Microsoft has invested $34 million+ in 76 replenishment projects globally, cut potable water use 97% at its Quincy, WA facility, and is now building zero-water-evaporation data centers using closed-loop liquid cooling. Industry-wide, water usage effectiveness has improved 39% since 2021. Modern facilities using closed-loop cooling operate at less than 0.2 L/kWh, compared to 20 L/kWh for traditional power generation. The trajectory is clearly toward water-neutral operations.
Electricity consumption and facility data sourced from the Lawrence Berkeley National Laboratory's 2024 United States Data Center Energy Usage Report (LBNL-2001637, U.S. Department of Energy, December 2024) and the U.S. Energy Information Administration's commercial electricity data (2023–24) and January 2026 demand forecast. National total of 176 TWh (4.4% of U.S. electricity) per LBNL and the Congressional Research Service (R48646, August 2025). Virginia data center share per EPRINC (Nov 2024) and JLARC (Dec 2024). Water data per the EIA, Google Environmental Report (2024), and Microsoft Sustainability Data. Household water comparison per International Energy Agency via CRS. Upcoming projects sourced from verified company press releases and governor announcements. Last updated: March 2026.
