Overall, data centers (including the AI-driven portion) remain a small share of total U.S. water use—even under growth projections—while homes and other facilities (hospitals, businesses, universities, etc.) account for far larger absolute volumes when viewed nationally.
Current Scale (≈2023 data)
- Data centers (direct on-site cooling): ~17–17.4 billion gallons per year. This is the evaporative/cooling water used at the facilities themselves.
- Including indirect water (used to generate the electricity that powers them): often estimated around 200+ billion gallons total footprint in some analyses, though methodologies vary and some argue the indirect figures overstate the impact.
- This equates to roughly the annual water use of ~160,000 average U.S. households (or a small city).
- As a percentage of total U.S. freshwater use (~322 billion gallons per day, or well over 100 trillion gallons per year): data centers are well under 1% (often cited around 0.04–0.5% depending on whether only direct use is counted).
Homes (residential): Public-supply domestic deliveries are a major part of the ~39 billion gallons per day in public supply. Average indoor use has dropped significantly (to ~38–40 gallons per person per day indoors), but outdoor/irrigation use remains substantial in many areas. Residential use dwarfs data centers nationally.
Other facilities (commercial + institutional — hospitals, businesses, universities, offices, schools, etc.): These make up a large share of public-supply non-residential use (historically ~17% or more of public-supply withdrawals in EPA framing). Hospitals and campuses with cooling towers can be intensive locally, but the sector as a whole is spread across tens of thousands of buildings and is still much larger in aggregate than data centers.
Projections (to ~2028–2030)
Growth is driven heavily by AI and hyperscale computing:
- Direct data-center water use is commonly projected to rise to 38–73 billion gallons per year by ~2028 (roughly 2–4× the 2023 level) under various LBNL and related scenarios.
- Some higher-growth or longer-horizon analyses (to 2030) show continued increases, with peak daily demands potentially requiring hundreds of millions of gallons per day of new capacity in aggressive cases. AI servers specifically are modeled in some studies as contributing hundreds of millions of cubic meters annually (hundreds of billions of gallons range across scenarios).
Even at the high end of these projections:
- Data centers would still represent only a low single-digit percentage of total public water withdrawals or consumptive use in most national analyses (e.g., 0.6–1.1% of public withdrawals or up to ~4–7% of public consumptive use in some 2030 baseline scenarios, depending on assumptions about efficiency and growth).
- Homes + commercial/institutional facilities would continue to dominate the non-agricultural, non-power-plant portion of water use.
Key Caveats
- Local vs. national: Nationally the percentage is small. Locally (especially in arid Southwest or stressed aquifers where many new data centers are proposed), a single large facility can be a noticeable fraction of a city’s or county’s water budget—sometimes comparable to thousands of homes or a significant commercial user.
- Direct vs. indirect: Most public debate focuses on on-site evaporative cooling. The much larger indirect component (power-plant cooling) is tied to the broader electricity grid and is not unique to AI/data centers.
- AI share within data centers: AI workloads are the main growth driver, so a rising share of future data-center water use will be AI-related, but exact percentages are scenario-dependent and not fixed.
- Efficiency trends (better cooling designs, more air/liquid cooling, reclaimed water) could moderate the growth, while faster AI expansion would increase it.
Bottom line: Projected AI/data-center water use is growing quickly and can be locally significant, but it is still expected to remain a small single-digit (or lower) percentage of overall U.S. water use compared with homes, businesses, hospitals, universities, agriculture, and power generation. The bigger water story nationally remains irrigation and thermoelectric power plants.
