Water, data centers, and AI

What the figures measure

Withdrawal

Water taken from a river, lake, aquifer, or utility. Some of it returns, for example as cooling-tower discharge or wastewater.

Consumption

Withdrawn water that doesn't return to its source, mostly because it evaporates. Data-center and power-plant figures on this page are consumption unless marked otherwise.

Evaporation

Water lost from a lake or reservoir surface. Gross evaporation counts all of it. Net evaporation subtracts the rain that falls on the lake.

Lifecycle footprint

All the water behind a product, from growing it to making it. For crops, most of it is rain stored in the soil. The irrigation share is the part comparable with cooling water.

Figures of different kinds aren't interchangeable, so the tables keep the kind beside each figure.

AI tasks and water

Water per task is the task's data-center electricity times a water factor per watt-hour: 1.15 mL for on-site cooling only (low), 4.29 mL with U.S. average power-plant water added (central), and 7.48 mL from Li et al.'s Arizona figures (high). One figure, Google's 0.26 mL for a median text prompt, was disclosed by a company. The rest are derived. Energy per task is on the AI energy page.

Published per-prompt figures

Most of the spread comes from three choices: whether power-plant water is counted, the energy assumed per prompt, and the location.

Household reference points

• One person's home use in the U.S., 82 gallons (310 liters) a day, equals the central water estimate for about 240,000 chatbot prompts.
• Twenty chatbot prompts a day at the central estimate come to about 26 mL a day, or 9.4 liters a year.
• A cup of coffee's irrigation water, about 1.3 liters, equals the central estimate for about 1,000 chatbot prompts.

Scale reference

Per day divides a yearly figure by 365.25. Gallons are U.S. gallons, 3.785 liters each. The published figure is under each name. For a sense of size, an Olympic pool at the minimum competition dimensions, 50 by 25 by 2 meters, holds 2,500 cubic meters: 660,000 gallons (2.5 million liters).

United States

1. The 2015 USGS compilation is the latest that covers every use. The 2010 to 2020 figures are newer USGS models of the three largest uses in the lower 48. USGS publishes 2020 estimates for the other uses as separate data sets, and no combined 2020 total was found.
2. No federal national total for reservoir evaporation was found. The 721-reservoir figure is a peer-reviewed estimate. It leaves out smaller reservoirs and natural lakes, and it ends in 2015.
3. The power-plant figure for U.S. data centers includes hydropower reservoir evaporation. Grid water factors that leave hydropower out are much lower.

Regional examples

Figures for one state, city, or lake.

1. Texas reservoir evaporation is a 2014 simulation with 1940s to 1990s hydrology. Newer reservoir-specific data shows evaporation rates rising about 1.1 inches a decade, but no statewide total from it has been published.
2. The two San Antonio figures cover different sites and periods: the reported figure covers two data centers over two calendar years, and Microsoft's covers its own datacenters from July 2024 to June 2025.
3. Lake Mead's volume is a two-year measurement from 2010 to 2012. Lake Powell's is a rounded estimate. Both lakes' areas change with their levels.

Product water footprints

These are lifecycle footprints: all the water used to grow and make one item, as global averages for 1996 to 2005. Most of it is rain stored in the soil where the crop grows. The irrigation share, water pumped or diverted and then consumed, is the part comparable with cooling water.

The water cycle

Evaporation moves water rather than destroying it. The open question is where the water falls again.

• Evaporated water is not destroyed: about 90% of atmospheric moisture comes from evaporation of oceans, seas, lakes and rivers, and it returns as precipitation.
• Once evaporated, a water molecule spends about 10 days in the air (USGS).
• Global average residence time of water in the atmosphere is 8.9 +/- 0.4 days; the distribution is long-tailed with a median around 5 days.
• Water vapour residence time has a mean of 8 to 10 days and a median of 4 to 5 days; differences between estimates come mostly from definitions.
• Longer evaporation residence times often indicate larger distances to areas of high precipitation, i.e., evaporated water can travel far before falling.
• Only about 10% of water evaporated from the oceans falls over land.
• About 40% of precipitation on land comes from land evaporation, and 57% of land evaporation falls back on land; the rest ends up over the ocean.
• The water cycle returns water to Earth, but not always to the same place, or in the same quantity and quality (EPA).
• Water consumption means withdrawn water permanently removed from the immediate water cycle, mainly by evaporation; this is why consumption, not withdrawal, is the figure that matters for a local watershed.

Taken together: water evaporated from a cooling tower or a reservoir returns as rain or snow within days, but mostly not to the basin it left. That is why hydrologists count evaporation as consumption, and why the same volume matters more in a dry basin than in a wet one.

Local context

Consumption matters most where water is scarce. In a basin with water to spare, evaporation from cooling has little local effect. In a water-stressed basin, the same volume is a larger share of what's available, and it can leave the basin as vapor. Figures for a state or a country say little about any one town.

Where data centers are built

• About two-thirds of US data centers built or in development since 2022 are in places with high water stress (Bloomberg analysis of WRI Aqueduct and DC Byte data).
• Microsoft reports 48% of its FY25 water consumption came from water-stressed areas; its Phoenix datacenters withdrew 981 ML in FY25 against 2,675 ML of local replenishment.
• Google reports 13% of its 2025 freshwater withdrawal came from sources at high risk of depletion or scarcity and 15% at medium risk.
• Meta reports 748 ML of its 2024 water consumption came from high or extremely high water-stress areas.

Cooling trade-offs

• Evaporative (water-cooled) systems are generally more energy-efficient; air-cooled chillers use no water but more energy (LBNL).
• Microsoft's zero-water chip-level cooling design avoids more than 125 million liters a year per datacenter, at a 'nominal increase' in annual energy use because mechanical cooling raises PUE.
• Closed loops are not automatically water-free; closed loops that use water to chill the coolant can still use up to half a billion gallons a year (HARC estimate).
• More electricity means more power-plant water where the grid is water-intensive. Grid water factors range from 1.29 L/kWh (ERCOT) to 9.50 L/kWh (Northwest, hydro-heavy) in WRI's data, so moving load to dry cooling in Texas shifts little water to power plants, while the same shift on a hydro-heavy grid shifts more.
• Texas's grid has a low water factor (ERCOT 1.29 L/kWh), which is why Li et al.'s Texas per-request figure (7.6 mL) is the lowest US location in their table, while Arizona (29.9 mL) is the highest.

Growth projections

• Reported in July and August 2025 as a forthcoming HARC white paper: 49 billion gallons for Texas data centers in 2025, and up to 399 billion gallons a year by 2030, about 6.6% of state water use. HARC's published paper (January 2026) gives 25 billion gallons for 2025 and 29 to 161 billion by 2030, 0.5% to 2.7%.
• LBNL projects on-site water for U.S. hyperscale data centers alone at 60 to 124 billion liters a year by 2028, against 66 billion liters for all U.S. data centers in 2023.

Specific sites

• Uruguay: Google's Canelones proposal initially called for up to 7.6 million liters of potable water a day during the country's worst drought in decades; the project was downsized and switched to air cooling before approval in 2024.
• Chile: residents of Cerrillos opposed a Google data center whose cooling towers could draw 169 L/s; Google moved to a less water-intensive design. A widely repeated book comparison of that site overstated it about 1,000x because of a unit error in a government document; corrected, 169 L/s is about 104.5% of Cerrillos residential use in 2019.
• Spain: Amazon's three proposed data centres in drought-hit Aragon are licensed for about 755,720 m3 a year, and Amazon asked to raise water consumption at its three existing sites by 48%.
• Arizona: Google's first Mesa data centre holds a permit for 5.5 million m3 a year, while Google reports that its Mesa site consumed 9.7 million gallons (about 36,700 m3) in 2025. Permits are caps, not measured use.
• Tucson's city council voted in August 2025 to end the Project Blue data center proposal after water concerns.
• San Antonio: two data centers run by Microsoft and the Army Corps used a combined 463 million gallons of water in 2023 and 2024, while SAWS customers were under Stage 3 rules limiting lawn watering to once a week (as reported; the SAWS data is not published).
• Microsoft reports its San Antonio datacenters withdrew 420 ML (about 111 million gallons) in FY25, 79% of it non-potable water.

Texas

• Texas data centers use an estimated 25 billion gallons a year including power-plant water, about 0.4% of state water use; 29 to 161 billion gallons a year by 2030 (up to 2.7%). State planners lack data: fewer than a third of 341 data centers answered a 2026 PUC survey.
• HARC's researcher: small-to-mid-size data centers need not use more water than a large subdivision or golf course; impact depends mainly on size and cooling technology, and on whether a small community is the supplier.
• In 2011, Texas's most intense drought year on record, evaporation from Texas reservoirs was higher than municipal water use; in 2023 evaporation was the largest water use from the Highland Lakes.
• SAWS declined to release water records for 36 Bexar County data centers from after its smart-meter rollout (about 2023), citing a Texas law on smart-meter data that a February 2026 Attorney General ruling applied to commercial accounts.
• HARC's January 2026 white paper splits its 25 billion gallons for Texas data centers into about 8 billion on site for cooling and about 17 billion consumed at the power plants that supply them.

Put a number in context

Enter a volume of water and the period it covers. The tool restates it in other units, converts it to a yearly amount, and compares that with reference values from this page: national ones first, then Texas ones, since both examples are in San Antonio. Each comparison uses the reference's own measure, so read the Reference column before comparing across rows.