Special offer. See more information about Outbyte and uninstall instructions. Please review EULA and Privacy policy.

Data center water figures are not interchangeable: withdrawal measures water taken in, consumption measures water not returned to the local water system, and water intensity expresses water use relative to an activity such as IT electricity demand. To interpret or compare a figure, check its water boundary, denominator, reporting period, location, and whether it is metered, reported, estimated, or modeled.

Withdrawal and consumption measure different things

Withdrawal is water removed from surface water or groundwater for use. Consumption is the portion not returned to the local water environment and no longer readily available for another local use. Water that evaporates is one example of consumption; some water taken in may instead be discharged back to the environment.

The U.S. Geological Survey notes that “Not all water withdrawn for human use is lost, much of it returns to the environment.” The distinction matters for data centers: an intake or withdrawal total is not automatically the amount consumed. A claim should identify its accounting boundary and describe what happens to water after use. USGS: Water Use in the United States.

What WUE tells you—and what it leaves out

Water Usage Effectiveness (WUE) is an operating intensity KPI: it relates water consumption to IT equipment energy rather than reporting a facility’s raw water volume. The Lawrence Berkeley National Laboratory (LBNL) 2024 U.S. report defines WUE as total data-center water consumption divided by IT electricity demand, expressed in liters per kilowatt-hour (L/kWh). ISO/IEC 30134-9:2022 specifies WUE as a KPI for quantifying use-phase water consumption and reporting operational water intensity. ISO/IEC 30134-9:2022 — Water usage effectiveness (WUE).

Special offer. See more information about Outbyte and uninstall instructions. Please review EULA and Privacy policy.

A WUE value is meaningful only with its category and boundary. WUE (site) concerns direct, on-site water consumption, primarily for cooling and other facility operations. WUE (source) also accounts for water consumed in generating the electricity used by the data center. A site-only figure therefore does not describe that electricity-related water footprint. The ISO standard’s official page identifies the published first edition as March 2022 and notes that a revision is underway; use the standard itself for category-specific calculation and reporting rules.

Water intensity depends on its denominator

Water intensity is a ratio, not a volume. Water per IT kWh, water per unit of floor area, and water per unit of workload answer different questions. Always state the numerator (for example, consumed water rather than total intake), denominator, units, and reporting period. Even two ratios both called “water intensity” cannot be treated as equivalent if they use different denominators or water boundaries.

How to compare two data center water claims

  1. Identify the water quantity. Is the number withdrawal, consumption, or a narrower measure such as cooling-water use?
  2. Check the boundary. Does it cover direct, on-site water only, or include off-site water consumed in electricity generation?
  3. Read the denominator and units. Confirm whether the ratio is per IT energy, facility area, workload, or another activity measure.
  4. Match the reporting context. Compare the same period, geography, units, and measurement category.
  5. Check how the figure was produced. Distinguish a metered value, company-reported aggregate, estimate, or model, and read any stated assumptions.
  6. Consider local conditions. Note the water source, local water availability, climate, cooling design, and whether water is returned after use.

If any of these dimensions differ or are not stated, the figures may still be informative, but they are not like-for-like measurements.

Why data center water use varies

Cooling is a primary source of direct water use, but demand depends on facility type and size, cooling design, outside temperature and humidity, and operating practices. Facilities may combine air, liquid, evaporative, and free cooling. These choices involve trade-offs: the Berkeley Law Center for Law, Energy & the Environment says that “Cooling systems that use less water often do so at the cost of increased energy consumption, and vice versa.” Closed-loop designs and nonpotable water can ease pressure on drinking-water supplies, but their effects depend on local design and conditions. Berkeley Law Center for Law, Energy & the Environment: Regulating Data Center Water Use in California.

Special offer. See more information about Outbyte and uninstall instructions. Please review EULA and Privacy policy.

Local impact cannot be inferred from a single WUE ratio. The ratio does not, by itself, tell you whether a facility draws from a water-stressed source, how much water returns to local systems, or how much water is consumed to generate its electricity. Cooling technology, climate, water availability, return flows, operating practices, and the electricity generation mix all affect the interpretation.

What published U.S. and company figures show

These examples illustrate why boundary and attribution matter; they are not interchangeable measurements or universal facility benchmarks.

Figure What it describes Source and qualification
21.2 billion liters in 2014; 66 billion liters in 2023 Direct water consumption by U.S. data centers LBNL’s 2024 report provides modeled national estimates; hyperscale and colocation facilities accounted for 84% of the modeled 2023 total under the report’s assumptions.
Nearly 800 billion liters in 2023 Indirect water consumption associated with U.S. data center electricity use LBNL’s 2024 report provides a modeled national estimate using electricity-grid assumptions. This is not site WUE.
4.52 L/kWh in 2023 National average indirect-water intensity for U.S. data-center electricity LBNL’s 2024 report estimate; it uses the report’s electricity-grid modeling and assumptions and is not a site-only WUE value.
Approximately 40% of global data center water consumption Estimated share occurring on site International Energy Agency estimate as reported by Berkeley Law Center for Law, Energy & the Environment in 2026. The report also notes other U.S. and European on-site estimates of roughly 17%–50%; these are estimates, not universal constants.
About 76% Share of on-site water used across Equinix’s global data-center operations that is consumed Company-level aggregate data summarized by Berkeley Law Center for Law, Energy & the Environment in 2026; the remaining 24% was reported as discharged to municipal sewers or surface waters.
About 79% Share of water withdrawn globally by Google data centers that is consumed Company-level aggregate data summarized by Berkeley Law Center for Law, Energy & the Environment in 2026; not a sector average.

LBNL’s figures are model-based national estimates, not direct readings from every U.S. facility. Facility-level detail can also be limited: reporting may be voluntary and aggregated, and methods may differ. A national estimate or company-wide aggregate cannot establish the water impact of a specific site.

Independent reader supportYour contribution helps us test, update, and keep practical guides available for everyone.Support on Ko-Fi

What to look for in a useful water disclosure

A clear disclosure should identify the metric and boundary rather than presenting an isolated number. Look for the following details:

Special offer. See more information about Outbyte and uninstall instructions. Please review EULA and Privacy policy.
  • Whether the reported volume is withdrawn or consumed, and how return flows are treated.
  • Whether water use is direct/site, indirect/source, or a combined measure.
  • The numerator, denominator, units, and period for any intensity ratio.
  • The location or geography, water source, and measurement category.
  • Whether the figure is metered, company-reported, estimated, or modeled, along with the method and assumptions.

Without those details, a figure may not support a fair comparison, even if it is reported in familiar units.

Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.