U.S. data centers directly consumed about 17.4 billion gallons of water for cooling in 2023 — and roughly 228 billion gallons once you add the water used to generate the electricity that powers them (Lawrence Berkeley National Laboratory). Both numbers are climbing fast as AI drives compute — and the heat that comes with it — to new highs. This is the definitive breakdown of data center water consumption: where it goes, how it’s measured, and what actually fixes it.
Direct vs. indirect: two very different numbers
“Data center water consumption” hides two separate figures, and mixing them up is where most of the confusion comes from:
- Direct (on-site cooling): the water evaporated in cooling towers and used for humidification. U.S. data centers consumed 17.4 billion gallons this way in 2023, and it’s projected to reach 38–73 billion gallons by 2028 (MOST Policy Initiative).
- Indirect (electricity generation): the water consumed off-site by power plants feeding the grid. This added roughly 211 billion gallons in 2023 — the larger share by far (CBS News).
Together that’s about 228 billion gallons in 2023, projected to grow to between 469 and 844 billion gallons by 2028. Data centers already draw around 4% of U.S. electricity, a figure that could hit 12% by 2028 — which is why the indirect number is so large.
How much water does a single data center use?
Scaled down to one facility, the numbers are still striking. A 150 MW hyperscale data center using evaporative cooling consumes roughly 340,000 gallons a day — about 120 million gallons a year (EPRI). At the water-intensive end, a water-cooled chiller running at a WUE of 2.8 L/kWh can consume 2.7 million gallons a day, or 970 million gallons a year — enough to fill nearly 390 Olympic pools.
Real facilities bear this out. Google’s data center in Council Bluffs, Iowa withdrew an average of 3.9 million gallons per day and consumed 2.8 million of it in 2024. With more than 4,000 data centers in the U.S. — about 37% of the world’s total — that consumption adds up quickly, especially in water-stressed regions.
The metric that matters: WUE
The industry measures this with Water Usage Effectiveness (WUE) — liters of water consumed per kilowatt-hour of IT energy. Lower is better.
The average U.S. data center ran at about 0.36 L/kWh in 2023, and that’s expected to rise to 0.45–0.48 L/kWh by 2028 as power-dense AI and hyperscale facilities take a bigger share. But site-level WUE varies enormously — anywhere from 0.1 to 9.0 L/kWh — depending on the cooling system and local climate.
Where the water goes — and why so much is lost
The water isn’t going into the servers; it’s rejecting heat. Most large facilities use evaporative cooling towers, which work by letting a portion of the water evaporate. That evaporation is what makes the cooling cheap and efficient — but it also means the water is gone for good. Up to 85% of the water a data center withdraws is consumed and never returns to the supply; Google reports about 78% consumed across its fleet.
The water that isn’t evaporated leaves as blowdown — deliberately dumped to keep dissolved minerals from scaling up the equipment. That discharge carries concentrated salts and silica that can affect drinking water, crops, and aquatic life downstream.
The energy–water tradeoff (why “just go dry” fails)
The two reflexive fixes both hit walls:
- Switch to dry or air cooling. It slashes water use, but it drives electricity consumption (and PUE) up and costs more to run. Operators use evaporative cooling because it’s the most efficient way to reject heat — not by accident.
- Just use reclaimed water. Treated wastewater helps where supply comfortably exceeds demand, but it has hard geographic limits. In the dense corridors where data centers cluster, that supply often isn’t there.
Chasing new water supply is a treadmill: the more compute you add, the more makeup water you need, and the harder each incremental gallon is to source.
The durable fix: recover the water you already have
There’s a better option than dumping and re-buying: recover the cooling water on-site. If you continuously pull the dissolved minerals back out of the loop, you no longer need to blow down water to control concentration — you keep circulating the water you already have.
That’s what NONA RECOVER does, using ICP — Ion Concentration Polarization, developed at MIT. Unlike reverse osmosis, ICP doesn’t force water through a membrane at high pressure; it pulls minerals aside with a low-pressure electric field, so there’s no membrane to foul. It bolts onto your existing cooling loop, works with your existing tower chemistry, and recovers 90% of the water that would otherwise be blown down. (For the full mechanism, see ICP vs. reverse osmosis; for the wider picture, see The AI Water Problem.)
ICP isn’t a lab concept: it was developed at MIT, its development was funded by the U.S. Army for field water production, and the first industrial RECOVER pilot is running today with Tracy Renewable Energy.
What this means for site selection
Water is quietly becoming a gating factor for where data centers can be built. Once you can recover most of your cooling water on-site, the math changes — circularity beats scarcity. You stop competing for new supply, you cut makeup-water and sewer costs, and you de-risk the sites where water, not power, is the binding constraint.
Data center water consumption is real, large, and growing. But the answer isn’t to abandon the cooling that works — it’s to stop throwing the water away.
Want to see the numbers for your site? Estimate your recovery and savings →