What Is WUE (Water Usage Effectiveness)

WUE measures how much water a data centre uses per kilowatt-hour of IT energy.

WUE stands for Water Usage Effectiveness. It is the standard metric used by the data centre industry to measure how efficiently a facility uses water relative to the computing work it performs. A lower number is better. The ideal value is zero, meaning no water used at all.

In this article
  1. How WUE Is Calculated
  2. What Water Is Actually Being Measured
  3. Why Data Centres Use Water at All
  4. WUE vs PUE: The Trade-Off
  5. Real-World WUE Figures
  6. The Limitations of WUE
  7. What WUE Means for Hosting Customers
  8. Quick Facts

The metric was created by The Green Grid, an industry consortium founded in 2007 to develop efficiency standards for data centre operations. The Green Grid introduced WUE in 2011 as a companion to its existing Power Usage Effectiveness (PUE) metric. Where PUE measures energy efficiency, WUE measures water efficiency. Both were later formalised as international standards by the ISO: PUE under ISO/IEC 30134-2 and WUE under ISO/IEC 30134-9:2022.

How WUE Is Calculated

WUE = Annual site water consumption (litres) ÷ Annual IT equipment energy consumption (kWh)

The result is expressed in litres per kilowatt-hour (L/kWh). A data centre that uses 120,000 litres of water in a year and consumes 1,000,000 kWh of IT energy has a WUE of 0.12 L/kWh. That happens to be AWS’s reported global figure for 2025, the best reported by any large-scale operator.

The industry average sits around 0.84 L/kWh, according to AWS’s own estimate of the wider market. That’s seven times higher than the best-performing facilities, which gives a sense of how much room for improvement exists across the sector.

What Water Is Actually Being Measured

Not all water that passes through a data centre counts toward WUE in the same way. Two variants appear in industry reporting and they measure different things.

WUE withdrawal counts the total volume of water extracted from any source for data centre use, regardless of whether any of it is returned to the local water supply afterwards.

WUE consumption counts only the water that is permanently lost at the data centre site, typically through evaporation. Water that is used and then returned to the watershed through treatment is not included.

Most operators report WUE consumption rather than WUE withdrawal, which means published figures typically undercount total water impact. Amazon’s widely cited 2.5 billion gallon figure for 2025, for example, represents direct site water withdrawal and excludes the indirect water used by the power plants supplying electricity to its facilities. Water used to generate electricity can be several times higher than what is consumed directly at the data centre itself.

Why Data Centres Use Water at All

Servers generate significant heat during operation. That heat has to be removed continuously or hardware starts to fail. The most common method in warm climates is evaporative cooling: air is drawn in from outside and pushed through water-soaked filters or cooling towers. The water evaporates, pulling heat out of the air as it does, and the cooled air flows into the server halls. The evaporated water has to be constantly replaced, which is where the consumption comes from.

The important detail is that evaporative cooling only kicks in when outdoor temperatures exceed roughly 29°C (85°F). Below that threshold, data centres can run on free air cooling, pulling ambient outside air directly past the server racks without using the water system at all. AWS reports that its facilities use free air cooling for around 90% of the year globally. The remaining 10%, concentrated in summer months and in warm-climate regions, is where most of the water consumption occurs.

This creates a direct relationship between geography and WUE. A data centre in Stockholm uses almost no cooling water because the climate rarely gets warm enough to need it. AWS Stockholm recorded a WUE of 0.02 L/kWh in 2025. A data centre in Jakarta operates in consistently hot, humid conditions and needs water-based cooling most of the time. AWS Jakarta recorded 2.85 L/kWh. Same company, same infrastructure standards, but more than 100 times the water intensity, purely because of location.

WUE vs PUE: The Trade-Off

WUE and PUE are partly in tension with each other. Evaporative cooling improves PUE because it is more energy-efficient than running mechanical chillers. But it increases WUE because it uses more water. Air cooling improves WUE because it uses little or no water, but it increases PUE because it requires more energy to move air effectively.

This trade-off means a data centre cannot be evaluated on either metric alone. A facility with an excellent PUE of 1.1 might be achieving that partly by running water-intensive evaporative cooling towers. A facility with a WUE near zero might be doing so by running power-hungry air-based systems that push its PUE higher. Evaluating a host’s environmental credentials properly requires looking at both figures together.

Metric Measures Formula Ideal value
PUE Energy efficiency Total facility energy ÷ IT energy 1.0
WUE Water efficiency Annual water used (L) ÷ IT energy (kWh) 0.0

Real-World WUE Figures

Disclosure of WUE figures is still less common than PUE disclosure across the industry, but the major cloud operators now publish their figures annually. The 2024-2025 figures for the largest operators:

Operator WUE (L/kWh) Notes
AWS 0.12 2025 global average, first public disclosure
Meta 0.19 2024 figure
Microsoft 0.27 FY2025
Google 0.31 2025
Digital Realty 0.59 2025
Equinix 0.91 2025 global average across all sites
Industry average 0.84 AWS estimate of wider market

Equinix’s figure of 0.91 L/kWh covers all of its sites globally, including many legacy facilities. Its Stockholm expansion achieved 0.42 L/kWh, well below the global average, showing how much location influences the number even within the same company’s portfolio.

The Limitations of WUE

WUE works well for internal benchmarking and year-on-year comparison within a single facility. It’s less reliable as a cross-operator comparison tool, and researchers have raised several specific objections to its use as a regulatory instrument.

It varies seasonally. A data centre in Arizona might report a low annual WUE because it uses free air cooling during mild winter months while consuming large amounts of water in summer. The annual average smooths this out in a way that can be misleading about peak local water demand.

It doesn’t account for local water stress. A WUE of 0.5 L/kWh in water-rich Norway has a completely different real-world impact from 0.5 L/kWh in drought-prone Arizona. The same metric value represents a trivial impact in one location and a serious one in the other. WUE as currently defined gives no weight to where the water comes from.

It excludes indirect water use. The water consumed at power plants generating the electricity that feeds a data centre can be several times greater than what the data centre uses directly. WUE only measures on-site consumption, which leaves a significant portion of total water impact unaccounted for.

It doesn’t distinguish water type. A data centre using potable drinking-quality water has a fundamentally different impact from one using reclaimed municipal wastewater. WUE treats both identically.

These limitations are why researchers at UC Berkeley and others have argued that WUE alone is not sufficient as a regulatory or policy instrument, and why the EU’s forthcoming data centre environmental reporting requirements are expected to require more granular water disclosure than WUE alone provides.

What WUE Means for Hosting Customers

Most web hosting providers don’t publish a WUE figure. The metric is reported at the data centre operator level, which is typically one layer removed from the hosting company you sign up with. A hosting provider using Equinix facilities inherits Equinix’s WUE profile. A provider running its own data centres publishes its own figures if it discloses at all.

Data centre location is the most accessible proxy for WUE when evaluating a hosting provider’s water efficiency. Providers with infrastructure in Norway, Sweden, Finland, Iceland, the Netherlands, or Germany are structurally better positioned on water efficiency than those operating only from warm-climate regions, regardless of any green marketing claims.

ISO 14001 certification covers environmental management systems including water use and is a more reliable signal than carbon-only certifications when evaluating water-related credentials. Some providers also disclose their use of reclaimed water for cooling, which is worth checking alongside any WUE figures they publish.

The guide on the water cost of AI and web hosting covers the broader picture of data centre water consumption with real operator figures. The eco-friendly hosting guide covers the providers with the strongest verified environmental credentials, including data centre location and certification details.

Quick Facts

Detail Info
Full name Water Usage Effectiveness
Created by The Green Grid (2011)
ISO standard ISO/IEC 30134-9:2022
Unit Litres per kilowatt-hour (L/kWh)
Ideal value 0.0 (no water consumption)
Industry average ~0.84 L/kWh
Best reported (2025) 0.02 L/kWh (AWS Stockholm)
Sister metric PUE (Power Usage Effectiveness)