Data center water use and cooling WUE field guide

What WUE is, and the water side of the bill

Water usage effectiveness, WUE, is the water a data center uses divided by the energy that reaches its IT equipment, expressed in liters per kilowatt-hour. It is the water counterpart to PUE, and it reads the same way: lower is better, and the IT energy sits in the denominator so the number scales with the work the building actually does. A site at 1.8 L/kWh is drinking far more water per unit of compute than one at 0.2.

The Green Grid published WUE in 2011, after PUE had already become the energy scorecard, precisely because the cheapest way to push PUE down is often to spend water. Evaporative cooling cuts the electricity but drinks water, and a plant chasing a headline PUE could hide a large water draw that no energy metric would ever catch. WUE exists to put that water back on the scoreboard.

Why this matters now and not ten years ago is the buildout. AI training and inference have pushed power and cooling demand up hard, and a lot of the new capacity is going into hot, dry regions where the climate favors evaporative cooling and water is already short. The result is that water has become a public, regulatory, and community question alongside energy, not a line item buried in the facilities budget. The PUE guide covers the energy scorecard; this guide is the water one.

What is the difference between site and source WUE?

Site WUE counts the water the data center uses on the ground, mostly cooling water, divided by IT energy. Source WUE adds the water used off-site to generate the electricity the plant consumes, which is often the larger number. The two answer different questions, and quoting one without saying which is how water reporting gets muddied.

The on-site water is the obvious part: the cooling tower make-up, the adiabatic or evaporative spray, the humidification, anything that crosses the site meter. Source water is less visible and usually bigger, because thermoelectric power generation, coal, gas, and nuclear, uses large volumes of water for its own cooling. The Green Grid handles this with an energy water intensity factor, the liters of water consumed per kilowatt-hour at the power plant, which you multiply by the facility energy to get the source term. A plant on a water-cooled coal grid can have a small site WUE and a heavy source WUE.

The practical upshot is that the total water footprint is site plus source, and the balance between them depends on the grid and the cooling design. An air-cooled data center on a thirsty grid moves its water problem upstream rather than removing it. A dry-grid or renewable-heavy plant can run evaporative cooling on-site and still come out lighter on total water than an air-cooled hall on a thermoelectric grid. Site WUE is what the operator controls directly; source WUE is what siting and procurement decide. Report the basis, because the two are not interchangeable.

Where the water actually goes

On a water-cooled data center the cooling plant is where almost all the water goes, and most of that is evaporation in the cooling tower or the adiabatic stage. Humidification takes a smaller slice, and a little goes to the rest of the building. If you are hunting for the water on a high-WUE site, start at the tower, because that is where the bulk of it leaves.

An evaporative cooling tower loses water three ways, and the sum of the three is the make-up water you have to keep adding. Evaporation is the big one and it is the part doing the cooling: water turns to vapor and carries the heat away with it. Drift is the fine mist that blows off the tower as liquid droplets, usually a small fraction of the flow and held down by drift eliminators. Blowdown, sometimes called bleed, is water deliberately dumped to keep the dissolved minerals from concentrating to the point they scale up the tower. Make-up equals evaporation plus drift plus blowdown.

The split tells you which lever to pull. Evaporation is mostly fixed by the heat you have to reject, so it is hard to cut without changing the cooling approach. Drift is small and controlled by hardware. Blowdown is the one you can manage with water chemistry, and it is the single biggest on-site water lever a tower operator has, which is why the cycles-of-concentration section gets its own treatment. Humidification water matters more in dry climates where the hall would otherwise drift below its low dew-point limit.

Evaporative cooling and the water it drinks

Evaporative cooling rejects heat by evaporating water, and it is efficient because water carries away a great deal of heat per liter when it turns to vapor. That is why the cooling tower has been the standard way to reject heat at scale for decades, and why an evaporative or adiabatic plant can hold a low PUE. You let the water do the work the compressor would otherwise have to do.

The physics is the appeal and the problem at once. Evaporating one liter of water absorbs a large amount of heat, far more than warming that same liter a few degrees, so a tower sheds a lot of heat with modest fan and pump energy. The catch is that the liter is gone, evaporated to the atmosphere, and has to be replaced. The cooling you bought cheaply on energy you paid for in water.

This is the core tension of the whole topic. Evaporative cooling is the reason a hall in a hot, dry climate can post a PUE near the best in the world while running a WUE several times the wet-climate average. The water-for-energy trade is not a side effect of the design. It is the design. The free cooling guide covers the adiabatic and evaporative economizer stages that extend the free-cooling hours; the point here is that every one of those hours is bought with water.

Why do data centers use water for cooling?

Data centers use water for cooling because evaporating water rejects heat with far less electricity than moving the same heat with air and a compressor, so spending water buys back energy and lowers PUE. The water-energy trade is the central decision in cooling a hall, and there is no answer that wins on both axes at once.

Run the two extremes. An air-cooled or dry-cooler plant uses little or no water but leans on the compressor and the fans, so it pays in electricity and a higher PUE, especially on hot afternoons when the dry cooler has the least to give. A fully evaporative plant uses water but unloads the compressor, so it pays in water and a higher WUE while holding a low PUE. Water-cooled designs have been reported to cut cooling electricity on the order of 25 to 35 percent against air cooling at peak conditions, though the exact figure depends on the climate and the design.

So the honest framing is a balance, not a winner. You are trading water for energy, and the right point on that line depends on what is scarce and what is dirty at the site. In a water-rich region on a clean grid, leaning evaporative can be the better total-resource answer. In a water-stressed region, the same design can be the wrong one even though its PUE looks great. Optimize PUE and WUE together, because pushing one without watching the other just moves the cost somewhere the metric you are watching cannot see.

What is the difference between water-cooled and air-cooled data centers?

A water-cooled data center rejects its heat through an evaporative path, a cooling tower or an adiabatic stage, so it uses water and less electricity. An air-cooled data center rejects heat to the outdoor air through dry coolers or air-cooled chillers, so it uses little or no water and more electricity. The split is the water-energy trade made physical in the plant.

The water-cooled plant runs condenser water across a cooling tower, evaporates a share of it to carry the heat to the atmosphere, and makes up the loss. Its WUE is real and its PUE is low, because the tower does cooling the compressor would otherwise pay for. The air-cooled plant has no tower and no evaporation, so its on-site WUE is near zero, but the dry coolers and chillers work harder, the fan power is higher, and the PUE climbs, worst on hot days when the air is a poor heat sink.

Neither is simply better. A dry, hot climate with cheap, clean power and tight water can favor air cooling despite the energy penalty. A temperate site with available non-potable water can favor evaporative. Many real plants split the difference with hybrid coolers that run dry most of the year and wet only on the hottest hours, which keeps water low while capping the peak energy. The chilled-water and cooling-tower work behind both paths is its own commissioning scope; the WUE point is that the cooling choice sets the water number before anything else does.

Climate, wet-bulb, and the siting tension

The same plant posts a different WUE in two cities, because evaporative cooling lives on the wet-bulb temperature, and the wet-bulb is set by the climate. A hot, dry climate has a low wet-bulb even when the thermometer is high, which is exactly where evaporative and adiabatic cooling perform best and where free-cooling hours are longest. It is also, very often, where water is scarce.

That is the siting tension in one sentence. The climates that make evaporative cooling most attractive on energy are frequently the climates least able to spare the water. A desert site can free-cool with the tower across most of the year and hold a low PUE, while drawing on an aquifer or a municipal supply that is already stressed. The engineering optimum on energy and the resource reality on water pull in opposite directions.

So the climate analysis has to run on both axes. The free cooling guide covers the bin data that counts the wet-bulb hours an evaporative or waterside economizer can use. The water version of that same analysis asks what those hours cost in liters, and whether the source for those liters can sustain the draw through a drought year, not just an average one. A site picked for its free-cooling hours without a water plan is a site that will be back in front of the regulator when the reservoir drops.

How much water does a data center use?

A large evaporative-cooled data center can use on the order of millions of liters of water a day, comparable to a small town, while an air-cooled or closed-loop plant of the same size can use almost none on-site. The honest answer is that it varies by more than an order of magnitude with the cooling design, the climate, and whether you count source water.

WUE is how the industry normalizes that spread. Reported site WUE ranges widely, with efficient and closed-loop sites well under 0.5 L/kWh, a frequently cited industry average somewhere in the range of roughly 0.5 to 1.9 L/kWh depending on whose fleet and which year, and older evaporative plants in hot, dry climates running 2 L/kWh or higher. The best hyperscale operators have reported fleet averages in the low tenths, on the order of 0.3 L/kWh, driven by closed-loop and air-cooled designs in their newest builds. Treat all of these as moving figures tied to a specific fleet and year.

The number that matters for any given site is its own metered water over its own IT energy, on a stated site-or-source basis, over a full year. A figure pulled from a press release is for banding and context, not for your permit application. And the spread is the real point: the difference between a high-WUE and a low-WUE design on the same load is the difference between a daily draw a community notices and one it does not.

Recycled and non-potable water

Using reclaimed, grey, non-potable, or harvested rainwater instead of drinking water is the largest single move an operator can make on the water question, because it cuts the demand on potable supply without necessarily changing the cooling design. The cooling tower does not care whether the make-up water started as drinking water or treated municipal wastewater, as long as the chemistry is managed.

This is the community win as much as the engineering one. The objection a town raises is usually about potable water, the same supply households and farms draw on, going to evaporate over a server hall. Switching the make-up to reclaimed municipal effluent, industrial process water, or captured stormwater answers that objection directly. Several large operators have moved a meaningful share of their cooling water to reclaimed or non-potable sources, with some campuses running entirely on recycled municipal wastewater, and others reporting a quarter or more of their fleet on non-potable supply. The numbers are operator-specific, but the direction is consistent.

The cost is treatment and infrastructure. Reclaimed water carries more dissolved solids and biological load than potable, so it needs more aggressive treatment, and it usually requires a dedicated supply line, a purple-pipe connection, or an on-site reclamation plant. Those are real capital items. But against the alternative, fighting a community over potable draw or losing a permit, the reclaimed-water path is often the one that keeps the project alive. It does nothing for source water, which still tracks the grid, but on the on-site number it is the strongest lever there is.

Cycles of concentration and blowdown

Running the cooling tower at higher cycles of concentration cuts blowdown, which cuts make-up water, and it is the cheapest on-site water lever a tower operator has. Cycles of concentration is the ratio of dissolved minerals in the circulating water to the minerals in the make-up water. Two cycles means the circulating water is twice as concentrated as what you feed in. Push the cycles up and you dump less water to blowdown for the same evaporation.

The math works because make-up is evaporation plus drift plus blowdown. Evaporation is fixed by the heat load, drift is small, so blowdown is the part you can shrink. Going from, say, three cycles to six roughly halves the blowdown for the same evaporation, which is real water off the meter every hour. The Department of Energy and water-efficiency programs treat blowdown management as the single highest-impact cooling-tower conservation measure for exactly this reason.

The limit is the water chemistry, and this is where it bites back. Concentrate the minerals too far and you scale the heat-transfer surfaces, foul the fill, and corrode the metal, which costs you capacity and equipment. How high you can cycle depends on the hardness, alkalinity, silica, and conductivity of the make-up water, and reclaimed water often caps the cycles lower because it starts more loaded. The cooling-tower and water-treatment scope owns the chemistry that sets the ceiling; the WUE point is that every cycle you can safely add is make-up water you stop buying.

Warmer water and more dry cooling

Running the cooling loop warmer cuts water two ways: it opens more hours where dry or free cooling can carry the load without evaporation, and it reduces the share of the year the evaporative stage has to run. The ASHRAE thermal guidelines allow a wider server-inlet and chilled-water range than most halls actually use, and every degree warmer you can safely run is a degree the outdoor air can handle on its own.

The mechanism is the same one the free cooling guide leans on for energy, read for water. A plant that has to make 7 C chilled water needs the evaporative tower far more of the year than one that can run a 20 to 30 C loop, because the warmer loop can reject heat to the dry outdoor air across many more hours. Warm-water cooling and water savings pull in the same direction, the way warm-water cooling and energy savings do.

The trade is thermal margin. Running warmer means the hall starts closer to its limit and heats up faster when cooling drops, so the ride-through window shrinks, and that is a real design constraint, not a free lunch. Set the warmest supply the equipment class and the warranty allow, confirm it against the ASHRAE envelope and the project documents, and you buy both the free-cooling hours and the lower water draw with the same setting.

Liquid cooling and the facility water

Liquid cooling, direct-to-chip cold plates or full immersion, moves the heat from the chip into a liquid loop, but it does not by itself decide whether the site uses water. The question is what the facility side of that loop rejects its heat into. A closed liquid loop coupled to a dry cooler uses almost no water. The same loop coupled to a cooling tower evaporates water like any other water-cooled plant.

This is the part that gets glossed over in the marketing. Liquid cooling is sold as efficient, and it is, on energy: bringing the coolant to the chip sheds the fan power and lets the plant run a warm loop. But the technology-cooling distribution unit hands the heat off to a facility loop, and the facility loop's heat rejection is where the water story actually lives. Direct-to-chip and immersion can reach a WUE near zero with a closed dry-cooled facility loop, or they can carry a meaningful WUE if the facility side is evaporative.

What liquid cooling does change is that the warm loop makes dry and closed-loop rejection practical at densities air never could. Because the chip heat comes out in warm water, the facility side can often reject it to the air without a tower, which is why liquid cooling is one of the routes to a low WUE and a low PUE at once. The trick is to design the facility loop deliberately, not to assume the liquid on the chip made the water question go away. The liquid-cooling commissioning scope owns the loop chemistry and the leak detection; the WUE point is to look past the cold plate to where the heat finally leaves the building.

Heat reuse and the net footprint

Reusing the waste heat instead of rejecting it cuts the net energy and, indirectly, the water, because heat sent to a district loop or a greenhouse is heat the cooling plant does not have to evaporate water to throw away. It is a smaller lever than the cooling choice on most sites, but on the right site it changes the whole resource picture.

The heat a data center makes is low-grade, warm rather than hot, which has historically limited reuse. A warm-water liquid-cooled loop changes that, because it delivers heat at a temperature a district heating network or an industrial process can actually use without a large heat pump. Where there is a customer for the heat next door, a greenhouse, an aquaculture operation, a municipal hot-water loop, the heat that would have gone up an evaporative tower goes into productive use instead.

On the metrics, reuse shows up in energy reuse effectiveness, ERE, more than in WUE directly, but the water benefit is real: every unit of heat carried away by the reuse loop is a unit the evaporative plant did not have to handle. The catch is that reuse needs a heat customer with matched timing and proximity, which most sites do not have, so it is a siting and partnership question as much as an engineering one.

Zero-water and air-cooled designs

A zero-water or near-zero-water design rejects all its heat to the air, using air-cooled chillers, dry coolers, or a refrigerant economizer, so it draws essentially no water on-site. In water-stressed regions this has gone from a niche choice to the default for several large operators, who would rather pay the energy penalty than fight for water rights or face a moratorium.

The way you get to zero water is to take evaporation out of the plant entirely. Air-cooled chillers and dry coolers reject heat across a coil to the outdoor air with no tower and no spray. A pumped-refrigerant or refrigerant-side economizer can carry free cooling in cold weather without water. The result is a site WUE at or near zero, which is the cleanest possible answer to a community water objection.

The bill comes due on energy, and you cannot pretend otherwise. Air cooling pays a higher PUE, worst on the hottest days when the air is the poorest heat sink and the dry coolers struggle, which is also when the grid is most stressed. So the zero-water design can move the resource problem to electricity and, through the grid, to source water. Whether it is the right call is a site-by-site judgment between scarce water and scarce or dirty power, and it should be made on the total footprint, not on the on-site WUE alone.

Water sourcing and the drought risk

Where the water comes from is a risk question, not just an efficiency one, and it should be settled before the cooling design is locked. Potable municipal supply, non-potable reclaimed water, groundwater from a well, and surface water each carry a different reliability, a different permit, and a different exposure to a dry year. A plant designed around water it cannot reliably get is a plant with a single point of failure nobody metered.

The acute risk is drought and restriction. A site that runs comfortably on its allocation in an average year can hit a curtailment or a use restriction in a dry one, and an evaporative plant that has no fallback has to either break the restriction or lose cooling capacity. Water rights and withdrawal permits set hard ceilings in many jurisdictions, and those ceilings can be cut when the basin is stressed. The plant inherits whatever the water authority decides, the same way it inherits the adopted code.

The defensive design carries a fallback. A hybrid plant that can run dry through a restriction, on-site storage that rides through a short interruption, or a non-potable source that is not subject to the same potable restrictions all reduce the exposure. The point is to treat the water supply as a resource with a permit, a ceiling, and a drought scenario, and to size the cooling so the plant survives the dry year, not just the typical one.

The community and the regulator

Water has become the most visible point of friction between data centers and the communities that host them, and it is now a permitting and reputational risk, not just an operating cost. A plant that evaporates millions of liters a day of drinking water in a region worried about its supply is a headline waiting to happen, and several projects have been delayed or blocked over exactly this.

The objection is usually specific and local: potable water, the same supply households and farms use, being consumed by a facility that the town sees as benefiting someone else. That framing is why the reclaimed-water and zero-water moves matter beyond their effect on the metric. They are the answer to the question the community is actually asking. Transparency helps too. Operators that report their water use, on a clear site-or-source basis, and that can show a non-potable source and a drought plan, tend to fare better in front of a planning board than those that decline to say.

On the regulatory side, the trend is toward disclosure and limits. Jurisdictions are starting to require water reporting, to condition permits on water sourcing, and in some cases to pause new data center water connections entirely. This is the ESG and community angle that sustainability reporting frameworks now expect a plant to address. Treat the community water question as a design input from the start, because retrofitting a water answer onto a plant already fighting a moratorium is the most expensive way to solve it.

How do you measure and report WUE?

You measure WUE by metering the water the facility uses over a period and dividing by the IT energy over the same period, in liters per kilowatt-hour, then stating whether the figure is site or source and annual or instantaneous. As with PUE, the arithmetic is trivial and the honesty is in the boundary, the period, and the basis.

Site WUE needs a water meter on the make-up supply to the cooling plant, plus any humidification and other process water, and the same IT energy figure the PUE uses as its denominator. Source WUE adds the facility energy multiplied by an energy water intensity factor for the grid that serves the site, which the operator usually has to take from a published regional factor rather than measure directly. Report the period, because a WUE quoted off the wettest month of the year is as misleading as a PUE quoted off the coldest night. Annual is the figure that matches the resource draw.

The reporting then lives in the sustainability disclosure alongside PUE, and increasingly in a permit filing. The Green Grid defines the metric and the site-versus-source split; the project's chosen reporting framework decides exactly what gets disclosed and how. State the basis every time. A WUE with no site-or-source label and no period is a number a reviewer cannot use and a regulator will not accept.

Optimizing WUE and PUE together

PUE and WUE pull against each other through the cooling design, so optimizing one in isolation usually worsens the other, and the only honest target is the pair. A plant tuned to the lowest possible PUE by leaning hard on evaporative cooling can post a WUE several times the average. A plant tuned to zero water by going air-cooled can carry a PUE the energy team would never accept. The design lives between them.

This is the most common high-level error in data center sustainability: reporting a great PUE and staying silent on water, which is water-blind optimization. The PUE looks like a win, and the water cost is real but invisible because nobody is measuring it. That is the exact failure WUE was created to expose, and it is why the guidance has settled on reporting WUE alongside PUE on any site that uses water for cooling.

The way through is to optimize on the total resource picture for the actual site. Weigh the energy, the on-site water, and the source water together, weighted by what is scarce and what is dirty where the plant sits. In a water-rich region on a clean grid, a low PUE via evaporative cooling can be the right total answer. In a water-stressed region, a slightly higher PUE with little or no water often is. Neither metric alone tells you which, and the plant that reports only the flattering one is hiding the cost, not avoiding it.

Water treatment, scale, and Legionella

Make-up water for an evaporative plant has to be treated, and the treatment program sets how high you can cycle the tower, how long the equipment lasts, and whether the plant grows a health problem. Skip it or run it loose and you lose water efficiency to scaling and inherit a Legionella risk, so the chemistry and the WUE are tied together.

Three failure modes drive the program. Scale forms when the concentrated minerals drop out on the warm heat-transfer surfaces, which cuts capacity and forces lower cycles and more blowdown, raising water use. Corrosion attacks the metal when the chemistry runs aggressive, especially on reclaimed water with a heavier dissolved load. And biological growth, including the Legionella bacteria that causes Legionnaires' disease, thrives in warm tower water and aerosolizes in the drift, which is a genuine public-health hazard, not a maintenance nicety. The treatment balances scale inhibitors, corrosion inhibitors, and biocide against the make-up chemistry.

Reclaimed and non-potable water make the treatment harder, which is the quiet cost of the water-source win. Water that started as municipal effluent carries more nutrients and solids, so it caps the cycles lower and needs a more careful biocide regime to hold the biological load down. The cooling-tower and water-treatment scope owns the detailed chemistry and the Legionella control plan; the WUE point is that the water number and the treatment program are the same conversation, and a plant that lets the treatment slip pays in both lost capacity and risk.

AI, hyperscale, and the water scrutiny

The AI buildout has put data center water under a spotlight, because the new capacity is large, fast, and often sited in regions where water is already contested. Annual on-site water use across the industry has been projected to rise several times over across the second half of the decade as hyperscale and AI capacity comes online, which is what turned a facilities metric into a public issue.

The pressure cuts two ways. The density of AI racks pushes toward liquid cooling, which, designed with a dry or closed-loop facility side, can drop both WUE and PUE and break the old trade. But liquid cooling on an evaporative facility loop still drinks water, and the sheer scale of the load means even a low WUE multiplied by a very large energy figure is a large absolute draw. A great per-kilowatt number on a gigawatt campus is still a lot of water.

So the efficiency push and the scrutiny are running at the same time. Operators are reporting water more openly, moving to non-potable and air-cooled designs in stressed regions, and being held to it by communities and regulators who now ask the water question first. The teams that treat water as a first-class design constraint, measured, sourced, and reported alongside energy, are the ones whose projects are clearing review. The ones still optimizing PUE and staying quiet on water are the ones meeting the moratorium.

What to document

A WUE figure with no method behind it is a number a regulator will not accept and a successor cannot reproduce. The record is what lets a reviewer check the figure against the meters, and what lets the next engineer understand which water lever was pulled and what it cost on energy.

Capture the WUE basis, site or source, the measurement period, the metered facility water and the IT energy with their meter points, the water source and its permit or allocation, the cooling type, and the energy water intensity factor if you reported source WUE. The table below frames the main reduction levers against what each one does to water and what it costs on energy or capital, which is the trade every one of these decisions turns on. The PUE guide covers the energy side of the same record; document them together, because they move against each other.

Common mistakes

• Optimizing PUE while never measuring WUE, so the water cost of evaporative cooling stays invisible.
• Quoting a WUE with no site-or-source basis and no measurement period.
• Drawing potable drinking water where reclaimed or non-potable supply was available.
• Running the cooling tower at low cycles of concentration and dumping make-up water to blowdown for no reason.
• Siting an evaporative plant in a water-stressed region with no drought fallback or storage.
• Treating a liquid-cooled plant as low-water without checking what the facility loop rejects heat into.
• Ignoring the community and regulatory water risk until a permit or a moratorium stops the project.
• Letting the water treatment slip, losing capacity to scale and inheriting a Legionella risk.
• Judging the zero-water design a winner on on-site WUE while ignoring the source water its higher energy draws.

Field checklist

Standards and references

The defining reference for WUE is The Green Grid, which created the metric, the liters-per-kilowatt-hour basis, and the site-versus-source split, including the energy water intensity factor used to compute source WUE. WUE is the water companion to the PUE metric, which The Green Grid originated and which ISO/IEC 30134-2 standardizes; WUE itself is now formalized as ISO/IEC 30134-9, part of the 30134 series that standardizes the family of data center resource-efficiency metrics, so confirm the current part and edition before citing a method on a report.

The ASHRAE TC 9.9 thermal guidelines set the temperature and humidity envelope that decides how warm the hall and the loop can run, which drives both the free-cooling hours and the evaporation, and ASHRAE also publishes guidance on water-cooled equipment and Legionella risk management relevant to the treatment program. Cooling-tower water efficiency, cycles of concentration, and blowdown management are addressed in Department of Energy and federal water-efficiency best-practice guidance and in water-utility programs, which give the practical targets for tower conservation. The exact metric definitions, factors, and guideline numbers revise, so verify them against the current published editions and the project's reporting framework.

Where a jurisdiction or a water authority mandates water reporting, a sourcing condition, or a withdrawal limit, that adopted requirement controls over any rule of thumb. The water rights, the withdrawal permit, the local water authority's rules, and the owner's chosen sustainability framework decide what the plant may draw and what it must report. Confirm them with the authority having jurisdiction, the same way you would the energy code.

Units, terms, and conversions

WUE carries units, unlike the dimensionless PUE, and the same water flows show up under several names across an energy model, a tower submittal, and a water permit. The terms below travel across the whole water scope.

WUE is liters of water per kilowatt-hour of IT energy in the standard form, though some reports use gallons per kilowatt-hour, so confirm the unit before comparing figures. Water flows are given in liters, cubic meters, or gallons per day or per year. The make-up balance, evaporation plus drift plus blowdown, is in those same volume-per-time units. Source WUE folds in an energy water intensity factor in liters of water per kilowatt-hour of generated electricity.

WUE

Water usage effectiveness, water used per unit of IT energy, in liters per kilowatt-hour; lower is better

Site WUE

WUE counting only the water used on-site, mostly cooling water, divided by IT energy

Source WUE

WUE that adds the off-site water used to generate the electricity the facility consumes

EWIF

Energy water intensity factor, liters of water per kWh of generated electricity, used to compute source WUE

Make-up water

Water added to the cooling tower to replace what is lost: evaporation plus drift plus blowdown

Blowdown / bleed

Water deliberately discharged from the tower to keep dissolved minerals from concentrating and scaling

Cycles of concentration

Ratio of dissolved minerals in the circulating water to the make-up water; higher cycles cut blowdown

Drift

Fine liquid droplets carried out of the tower by the airflow, limited by drift eliminators

PUE

Power usage effectiveness, total facility energy over IT energy; WUE is its water counterpart

Related tools

Calculators and readiness checks for this work

Compare your options

FAQ

People also ask

• What is free cooling in a data center?
• What is PUE?
• What is adiabatic cooling in a data center?
• How does a cooling tower work?
• What is a CDU in a data center?