← AI Data Center Engineering Studio
Concept Explainer · AI Data Centers

PUE vs. WUE

A data center can post a great PUE while consuming millions of gallons of water a year — because Power Usage Effectiveness and Water Usage Effectiveness measure two completely different resources, and optimizing one can actively work against the other.

PUE (Power Usage Effectiveness) gets most of the attention because it's the metric that made data center efficiency reporting mainstream. But it only measures electricity — it says nothing about water. Many of the cooling techniques that produce excellent PUE numbers, like evaporative and adiabatic cooling, get that efficiency specifically by evaporating water to reject heat instead of running energy-hungry mechanical chillers. That trade is invisible to PUE and is exactly what WUE (Water Usage Effectiveness) was created to capture. As AI data centers scale into the hundreds of megawatts, water consumption has become a first-order siting and permitting constraint in water-stressed regions — which is why operators increasingly report both metrics side by side, not PUE alone.

PUE: total facility power ÷ IT power

Electricity Metric
TOTAL FACILITY POWERIT/GPU load+ cooling power+ distribution losses+ auxiliary/lighting(all metered in kWh)÷IT EQUIPMENTPOWERservers, GPUs, storage=PUEe.g. 1.2 — 2.0measures electricity only — water use is invisible here
Resource measured
Electricity (kWh)
Lower is better; 1.0 is the theoretical (unreachable) minimum.
Defined by
The Green Grid
Published 2007; the industry-standard efficiency metric ever since.

WUE: annual site water use ÷ IT power

Water Metric
ANNUAL SITE WATER USEevaporative cooling towersadiabatic pre-coolinghumidificationon-site power generation(measured in liters)÷IT EQUIPMENTENERGYmeasured in kWh=WUEL/kWh, e.g. 0.2 — 2.0+a chiller-only (no evaporation) design can post WUE ≈ 0
Resource measured
Water (liters)
Lower is better; 0 is achievable with a fully non-evaporative design.
Defined by
The Green Grid
Published 2011 as a companion metric alongside PUE, not a replacement for it.
Why this works

The two metrics can move in opposite directions from the same design choice.

Evaporative and adiabatic cooling reject heat by evaporating water, which is thermodynamically very energy-efficient — it's a major reason many facilities that lean on it post PUE numbers as low as 1.1-1.2. But every liter evaporated is a liter of WUE. A facility that instead uses mechanical chillers with no evaporative assist can drive WUE toward zero, but at the cost of higher electrical energy use for compression cooling, which pushes PUE higher. Neither metric alone tells you whether a design is actually "efficient" — you need both, because they represent a genuine engineering trade-off between two different scarce resources: electricity and water.

Common misconception
"A great PUE means the data center is environmentally efficient, full stop."

PUE only accounts for one resource — electricity — and it's entirely possible to achieve an excellent PUE by shifting the cooling burden onto water evaporation rather than reducing it. A facility with a PUE of 1.1 that evaporates millions of gallons of water annually is not automatically the "better" design from a sustainability or community-impact standpoint, especially if it's sited in a water-stressed region. This is exactly why operators reporting sustainability metrics increasingly publish WUE (and sometimes carbon-based metrics like CUE) alongside PUE — a single-number efficiency story is incomplete without knowing which resource was traded to get there.

Related Concept Explainers
Liquid Cooling vs. Air Cooling
The cooling technology choice that drives both of these metrics
PUE Calculator
Calculate PUE from your own facility power breakdown

PUE vs. WUE — Concept Explainer

Explains why Power Usage Effectiveness (PUE) and Water Usage Effectiveness (WUE) are companion metrics, not substitutes for each other — the cooling design choices that produce a great PUE (like evaporative cooling) often increase WUE, and vice versa. Illustrated with the PUE and WUE formulas side by side.

Why This Is Commonly Confused

PUE became the industry's headline efficiency number well before WUE existed and remains the most widely reported metric, so it's easy to treat a low PUE as a complete efficiency story. It isn't — PUE is silent on water because it was designed to measure electrical overhead specifically. The Green Grid published WUE in 2011, four years after PUE, precisely because operators and researchers recognized the electrical efficiency metric alone didn't capture a real environmental cost that certain cooling designs (particularly evaporative and adiabatic systems) were incurring.

What Each Metric Actually Measures

PUE = Total Facility Power ÷ IT Equipment Power, both measured in kWh over the same period. It captures every source of electrical overhead: cooling equipment power, UPS and power distribution losses, and auxiliary loads like lighting. A PUE of 1.0 is the unreachable theoretical minimum where 100% of power reaches IT equipment.

WUE = Annual Site Water Usage (liters) ÷ IT Equipment Energy (kWh), so it's not even in the same units as PUE — WUE is a rate (liters per kilowatt-hour), not a dimensionless ratio. Site water usage includes water consumed by evaporative cooling towers, adiabatic pre-cooling (evaporatively cooling incoming air before it reaches a dry cooler), humidification systems, and any on-site power generation with water cooling needs. A facility using purely mechanical (chiller-based, no evaporation) cooling can post a WUE near zero.

Where This Matters in AI Data Center Siting and Design

AI data center campuses drawing hundreds of megawatts have made water consumption a front-page issue in several proposed and built projects, particularly in drought-prone regions — permitting authorities and local communities increasingly scrutinize projected annual water withdrawal alongside power demand. This has pushed some operators toward closed-loop, non-evaporative cooling designs (accepting a modestly higher PUE) specifically to minimize WUE and ease permitting in water-stressed locations, while others in water-abundant regions continue to favor evaporative cooling for its electrical efficiency. Understanding both metrics — and that they trade against each other — is essential for site selection and cooling system design, not an academic footnote.

Frequently asked questions

Can a data center have both a low PUE and a low WUE?

Yes, but it typically requires more capital-intensive engineering — examples include using free/economizer cooling during favorable outdoor conditions (low energy cost, no water use) combined with closed-loop mechanical cooling as backup, or liquid-to-chip cooling systems that reject heat efficiently without evaporation. It is a harder design target than optimizing for either metric alone.

What are typical WUE values for AI/hyperscale data centers?

Values vary widely by design and climate — facilities disclosing figures have reported WUE roughly in the 0.1-2.0+ L/kWh range. Non-evaporative, closed-loop designs report WUE near zero; heavily evaporative-cooled facilities in hot, dry climates (where evaporative cooling is thermodynamically most attractive) tend to report the higher end of that range.

Does liquid cooling for GPUs increase or decrease water use compared to air cooling?

It depends on the facility's overall cooling architecture, not the direct-to-chip loop itself. The direct-to-chip/CDU loop is typically a closed loop with no water loss. Whether the facility's WUE goes up or down depends on how the facility rejects heat from that loop to the outdoors — a cooling tower (evaporative) versus a dry cooler or chiller (non-evaporative) — which is a separate design decision from choosing liquid vs. air cooling at the rack level.

Is there a single combined metric that captures both power and water?

Not one that has achieved PUE's level of industry standardization. Some operators report Carbon Usage Effectiveness (CUE) alongside PUE and WUE to capture emissions, but there is no single widely adopted ratio that combines electrical and water efficiency into one number — which is exactly why both PUE and WUE are typically reported separately, side by side.

Related tools & guides

PUE CalculatorData Center Cooling Load CalculatorLiquid Cooling vs. Air Cooling Concept ExplainerAI Data Center System Architecture