Water Usage and Heat Reuse in Data Centers

PUE tells you how much extra power a facility burns to deliver one kilowatt hour to your servers, and for years it has been the headline number in every sustainability report. Yet power overhead alone never tells the sustainability story. A facility that optimizes PUE by leaning on evaporative cooling can trade electricity for water in a region where water is the scarcer resource, and a facility that vents every watt of server heat into the atmosphere can look efficient on paper while wasting thermal energy that a neighboring building would gladly buy. If you buy, operate, or report on data center capacity, the metrics that sit beside PUE are no longer optional detail, they are the difference between an honest sustainability claim and a narrow one.

This guide moves beyond PUE to the two metrics most often missing from procurement conversations: Water Usage Effectiveness and heat reuse captured through the Energy Reuse Factor. It explains what each one measures, what drives the number up or down, how heat reuse reshapes the efficiency math, and how to read all three alongside Carbon Usage Effectiveness so you can evaluate a facility or plan your own improvements without trading one resource for another.

Why PUE Alone Misleads

PUE is a ratio of power to power, which makes it deliberately blind to everything else the facility consumes. Two halls can report an identical 1.3 and impose completely different costs on the environment around them, because one achieves that number with dry coolers and a larger electric bill while the other uses evaporative towers that consume millions of liters per year. The metric rewards lower electricity overhead no matter what resource pays for the saving, so a site that trades a modest power improvement for a large water increase will look better on the one number you publish and worse on the one you omit.

The industry learned this trade the hard way. As energy prices pushed operators toward evaporative and adiabatic cooling, aggregate water consumption rose in markets where water scarcity was already a planning constraint. Regulators noticed, and voluntary reporting frameworks followed. ISO/IEC 30134 now defines a family of metrics that sit beside PUE precisely to close this loophole: WUE for water, CUE for carbon, and ERF for energy reuse, each measured with the same boundary discipline and annual accounting that make PUE comparable. A procurement team that asks only for PUE is asking for one quarter of the sustainability picture.

What WUE Actually Measures

Water Usage Effectiveness is defined in ISO/IEC 30134-9 and it mirrors PUE in structure: liters of water in the numerator, kilowatt hours of IT energy in the denominator. The numerators make all the difference. WUE counts water consumed, not water withdrawn, so water that enters and leaves the facility without evaporation is not the point. What matters is the water that does not return to the source, primarily the evaporation in cooling towers and adiabatic coolers, plus blowdown that keeps dissolved solids in check and the smaller volume used for humidification. A WUE of 1.8 means the facility consumes 1.8 liters for every kilowatt hour that reaches your equipment, and the unit itself tells you which resource is being spent.

There are two scopes you will encounter and they answer different questions. Site WUE, often labeled WUE category 1, counts only water consumed inside the facility boundary, which is the number a site operator controls directly. Source WUE, categories 2 and 3 in the standard, adds the off site water consumed to generate the electricity the facility draws, which depends on the grid mix and can be far larger than the on site number for thermoelectric heavy grids. Both have uses: site WUE is the operational metric you manage week to week, while source WUE connects water and power choices so an apparent water saving does not simply outsource the consumption to the power plant upstream. When someone quotes a WUE, the honest follow up is which category they mean, over what year, and whether the boundary includes or excludes that upstream water.

What Drives Your WUE

The largest driver is the cooling system your facility chose years before you arrived. Open evaporative towers trade electricity for water and produce the highest site WUE, often 1.5 to 2.5 liters per kilowatt hour at the site level. Hybrid adiabatic systems that wet the media only on hot days sit materially lower, typically 0.5 to 1.0, because they evaporate only when dry cooling alone cannot reject the heat. Fully dry systems that rely on air cooled chillers or direct liquid to ambient heat rejection can approach a site WUE near zero, at the cost of higher electrical consumption and in hot climates a higher PUE to compensate. Each design moves the same thermal load with a different mix of water and watts, and the optimum mix depends on the local price and availability of each.

Climate and operating discipline amplify the design choice. A temperate site that can run dry or free cooling for most of the year consumes far less water than an identical plant in a hot arid market, and a site that raises its chilled water set points within the ASHRAE allowable range evaporates less while drawing slightly more fan or pump power, the same trade playing out at the set point level. Water quality matters too, because higher cycles of concentration reduce blowdown but demand tighter chemistry control. This is also where scarcity reframes the metric: a liter in a water stressed basin carries a different planning weight than a liter where rainfall is abundant, and a growing number of corporate water stewardship frameworks ask not just for the ratio but for the absolute volume mapped against local basin risk. A low WUE in a drought prone region may still represent a material business risk even when the ratio itself looks responsible.

Heat Reuse and the Energy Reuse Factor

If WUE captures a cost that PUE hides, heat reuse captures a value that PUE ignores. Every watt that reaches your servers becomes heat, and in a conventional facility that heat is rejected to the atmosphere as waste. In a reuse design that same heat is captured at a usable temperature and exported to a district heating network, a greenhouse, a swimming complex, or an adjacent industrial process that would otherwise burn gas to make the same heat. The Green Grid defined the Energy Reuse Factor to quantify this: ERF is reused energy divided by total facility energy, a number between 0 and 1.0, and it is designed to be reported alongside PUE so the two numbers adjust each other.

The arithmetic is straightforward. Effective PUE subtracts reused energy from the numerator before dividing by IT energy, which is equivalent to PUE multiplied by one minus ERF. A hall that draws 1.4 kilowatt hours per kilowatt hour of IT energy and reuses 30 percent of its total energy has a conventional PUE of 1.4, an ERF of 0.3, and an effective PUE near 0.98 to 1.0 depending on boundary, which reframes the efficiency story from overhead to circular energy use. Reported examples show the range: Nordic colocation sites coupled to municipal heating networks routinely export enough heat to cover hundreds of homes, continental facilities have wrapped nearby pools and office parks into their rejection loop, and hyperscale campuses in cooler markets preheat ventilation air for adjacent buildings. Each project follows the same physical ladder, and temperature decides how far up it goes.

That ladder starts with what your servers actually produce. Air cooled exhaust at 25 to 35 degrees is usable for preheating but needs a heat pump to reach the 65 to 90 degrees a district network expects, which adds electricity but typically returns three to five units of heat per unit of electricity. Direct liquid cooling that leaves the rack at 40 to 60 degrees is far closer to network temperature and can reuse more heat with less pumping, which is one practical reason density and reuse planning now sit in the same conversation. The constraint is always proximity and temperature: there must be a heat customer close enough to make the pipe economic and a supply temperature high enough to be worth upgrading, which is why the strongest reuse business cases sit in cold climate metros with existing heating grids rather than isolated campuses in warm zones.

CUE and the Full Sustainability Picture

PUE, WUE, and ERF together describe how the facility handles power, water, and waste heat, and CUE closes the loop on carbon. Carbon Usage Effectiveness multiplies total facility energy by the grid emission factor and divides by IT energy, so it reports kilograms of carbon dioxide equivalent per kilowatt hour of IT energy. The grid factor is the lever: a hall at PUE 1.1 on a coal heavy grid can carry a worse CUE than a hall at 1.5 on hydro or nuclear power, and an annual procurement contract for clean power can move CUE without touching any equipment in the building. This is precisely why sustainability reporting is moving toward location based and market based carbon numbers side by side, so efficiency and procurement are visible as separate contributions.

Reading the family together is what keeps any single metric honest. A low PUE alongside a high source WUE hints that water is subsidizing the power number. A low site WUE alongside a high PUE in a hot climate may signal that dry cooling was chosen without enough plant to handle the peak. A strong ERF that pulls effective PUE toward 1.0 still needs a CUE alongside it, because reused heat displaces gas heating and the carbon credit belongs in the carbon accounting, not just the power ratio. No single number can carry a sustainability claim on its own, which is why the ISO 30134 family was designed to be reported as a bundle, each measured over the same year and the same boundary.

What Good Numbers Look Like

Targets only mean something when the boundary and climate are held constant, so treat benchmarks as ranges to situate your own site rather than as a league table. For site WUE, a facility with evaporative towers typically lands between 1.5 and 2.5 liters per kilowatt hour, a hybrid adiabatic plant lands between 0.5 and 1.0, and a fully dry or liquid to ambient design can approach zero on site, with the caveat that it will draw more electricity to get there. A source WUE that includes upstream water for electricity generation can run higher, often 2 to 8 liters per kilowatt hour depending on how much of the local grid comes from thermoelectric generation, which is why a procurement shift toward wind or photovoltaics moves both CUE and source WUE in the right direction at once.

For reuse, ERF spans the full range because it depends on whether there is anyone to sell the heat to. A facility with no reuse customer reports an ERF of 0, which is still the norm globally. A site coupled to a district network or a large neighboring heat load typically reports 0.2 to 0.4, and the best documented Nordic integrations that feed low temperature networks have exceeded 0.5 during heating season. An ERF above 0.3 already represents a material energy contribution to the community around the facility, and when you see an effective PUE below 1.0 it almost always reflects an ERF large enough to offset the entire overhead, not a building that somehow beats thermodynamics.

The honest target is the trend, not the headline. A site that moves site WUE from 1.8 to 1.0 through hybrid retrofits and careful set point management has made a real water saving regardless of what any peer reports, while a site that quotes a single winter month or a single cool year is not showing a durable number. The same discipline that makes PUE useful makes the broader family useful: measure at the same boundary, over the same full year, and compare against yourself.

Measuring Honestly and Planning Around It

Regulation and customer scrutiny are already pushing measurement from voluntary to required. The European Energy Efficiency Directive obliges larger data centers to report energy and sustainability data, Germany’s energy efficiency law tightens PUE limits for new builds and pairs them with reuse obligations for facilities above 1 megawatt of waste heat potential, and voluntary schemes such as EU Code of Conduct and LEED map to the same ISO metrics the standards define. When procurement teams write RFPs they increasingly ask for annual PUE, site WUE, ERF, and CUE together, and the facilities that can produce auditable year long series win the credibility argument. The question is no longer whether you will be asked for these numbers but whether you can produce them with the boundary and period clearly stated.

The practical work for your own footprint starts with the denominator you already control: knowing exactly what is installed, what it draws, and where the power and thermal load sit. Site level capacity records that capture which racks sit in which hall, the nameplate and measured draw of each device, the PDU and circuit behind each rack, and the thermal design point per zone turn an annual ratio into a plan you can act on. A water saving proposal can then be weighed against the room it affects, a reuse study can start from the actual supply temperatures available in each row, and a carbon comparison between sites can reference the specific racks and zones behind the carbon math. For teams that maintain those records in a source of truth such as site management and rack management, the sustainability bundle stops being a reporting exercise and becomes a capacity planning input, visible to the whole team and comparable year over year. The metrics beyond PUE reward the same discipline that makes PUE useful in the first place: document the physical reality, measure consistently, and let the trend tell you whether you are actually getting more sustainable.