Skip to content

Data Center PUE, WUE, CUE & ERE: Formulas and Boundaries

Compare data center PUE, WUE, CUE, and ERE with formulas, units, measurement boundaries, worked examples, and double-counting checks.

PUE measures facility energy relative to IT energy. WUE measures water use per unit of IT energy. CUE measures operational greenhouse gas emissions per unit of IT energy. ERE adjusts facility energy for energy beneficially reused outside the data center boundary. These metrics describe different resource relationships—not interchangeable measures of overall sustainability. Their usefulness depends on consistent boundaries, reporting periods, and IT-energy measurements.

PUE, WUE, CUE, and ERE Reference Table

The table uses the operational definitions developed in The Green Grid’s metric papers. Site-based WUE and source-based WUE are separated because they include different water flows. None of these metrics, as defined here, constitutes a lifecycle assessment.

Metric Full name Numerator Denominator Resource boundary Unit Operational or lifecycle? Double-counting warning
PUE Power Usage Effectiveness Total data center facility energy IT equipment energy IT equipment plus supporting power, cooling, and other data center infrastructure Dimensionless; kWh/kWh Operational Facility energy already includes IT energy. Do not add IT energy again or subtract exported waste heat.
WUE Water Usage Effectiveness Annual site water use Annual IT equipment energy Water used for data center operations, including cooling, humidification, and associated on-site energy production L/kWh of IT energy Operational Do not add tower evaporation and blowdown to makeup water that already supplies those flows.
WUEsource Source Water Usage Effectiveness Annual site water use plus off-site water use associated with energy production Annual IT equipment energy Site operations plus water associated with producing the energy supplied to the facility L/kWh of IT energy Operational, including energy-supply water Site water is already included. Do not add site WUE to a source-WUE result again.
CUE Carbon Usage Effectiveness Operational greenhouse gas emissions within the stated accounting boundary IT equipment energy for the same period Direct operational emissions and emissions associated with purchased energy; excludes embodied lifecycle emissions kg CO₂e/kWh of IT energy under the Green Grid definition used here Operational Do not add electricity-related emissions again after calculating them from total facility electricity.
ERE Energy Reuse Effectiveness Total facility energy minus qualifying energy reused outside the data center boundary IT equipment energy The PUE facility boundary, with a separately measured beneficial energy export Dimensionless; kWh/kWh Operational Internal heat reuse may already reduce facility energy. Do not subtract it again as an external reuse credit.

Sources: The Green Grid’s PUE White Paper #49, WUE White Paper #35, CUE White Paper #32, and ERE White Paper #29. These establish the terminology and calculation relationships used below; they do not establish that a particular project complies with a reporting standard.

Establish the Measurement Boundary Before Calculating

Use one defined reporting period and the same IT-energy denominator across the metrics. Annual totals are appropriate for capturing seasonal cooling, water use, and heat-reuse conditions. A short-period result should retain its actual period label rather than being presented as measured annual performance. The Green Grid’s WUE and CUE papers explicitly link their denominators to the IT energy used for PUE.

Record the following alongside the results:

  • Reporting dates and whether inputs are measured, estimated, or projected.
  • Facility boundary, including the allocation of shared infrastructure.
  • IT meter location and any non-IT equipment supplied through that meter.
  • Energy sources and the treatment of on-site generation.
  • Water sources, meter locations, and whether the numerator represents supplied water or net consumptive use.
  • Emissions-factor source, geographic coverage, reporting year, and accounting basis.
  • For ERE, the energy-export measurement point and the external use receiving that energy.

These details determine whether results are comparable. The Green Grid’s PUE guidance distinguishes measurements at UPS outputs, PDU outputs, and IT equipment inputs; changing the measurement point can change which distribution losses appear in the denominator. Mixed-use buildings also require separation or allocation of data center energy from unrelated building loads.

For a straightforward all-electric facility, total facility electricity can be used directly with IT electricity. A facility receiving fuel, purchased chilled water, or other energy requires the applicable energy-accounting method—not merely the addition of unlike meter readings. The Green Grid’s PUE guidance addresses source-energy treatment when non-electrical energy is supplied. Do not present an electricity-only ratio as complete facility PUE when material energy inputs are omitted.

PUE: Facility Energy Includes the IT Load

\mathrm{PUE}=\frac{E_{\mathrm{facility}}}{E_{\mathrm{IT}}}

Where:

  • (E_{\mathrm{facility}}) is total energy attributable to the data center and its supporting infrastructure.
  • (E_{\mathrm{IT}}) is energy delivered to the equipment that processes, stores, manages, or routes data.
  • Both quantities cover the same period and use a consistent energy basis.

Total facility energy includes IT energy, cooling energy, power-distribution losses, and other supporting loads. It is not infrastructure overhead alone. This definition appears in The Green Grid’s PUE guidance and DOE’s data center cooling-water guidance.

For an all-electric facility, the relationship can also be written as:

E_{\mathrm{facility}}
=
E_{\mathrm{IT}}+E_{\mathrm{overhead}}
\mathrm{PUE}
=
1+\frac{E_{\mathrm{overhead}}}{E_{\mathrm{IT}}}

A PUE of 1.40 means the facility uses 1.40 kWh for every 1.00 kWh delivered to IT equipment. It does not mean the IT equipment converts electricity into useful computing work at 71% efficiency. PUE measures infrastructure energy allocation, not computing productivity.

Do not subtract exported heat from the PUE numerator. That adjustment belongs in ERE. A PUE below 1.0 under this boundary indicates a boundary, metering, or accounting problem rather than a heat-reuse achievement.

WUE: Identify What the Water Meter Counts

Site-based WUE

\mathrm{WUE}=\frac{V_{\mathrm{site}}}{E_{\mathrm{IT}}}

Here, (V_{\mathrm{site}}) is annual site operational water use in liters, and (E_{\mathrm{IT}}) is annual IT energy in kWh. The result is L/kWh of IT energy—not L/kWh of total facility energy. DOE uses this site-based relationship in its cooling-water guidance.

The Green Grid’s WUE paper includes humidification and cooling-related water, including cooling-tower evaporation, blowdown, and drift. Its operational boundary also includes water used for associated on-site energy production.

Water terminology requires particular care:

  • Makeup water is water supplied to replace water leaving a cooling system.
  • Evaporation transfers water to the atmosphere.
  • Blowdown removes water to control dissolved-solids concentration.
  • Circulating water repeatedly passes through the system; its cumulative flow is not new water use.

DOE describes makeup and blowdown as distinct flows in cooling-tower operation. If a makeup meter captures water replacing evaporation, blowdown, and drift, adding those estimated losses to the metered makeup would count the same supply twice.

A WUE based on supplied makeup water is not automatically equivalent to a net consumptive-water indicator that subtracts qualifying return flows. State the numerator convention. Also identify reclaimed and potable supplies separately where relevant: an undifferentiated L/kWh result does not explain water quality, availability, or local water stress. The Green Grid’s WUE paper identifies water source, quality, and locality as limitations requiring further consideration.

Source-based WUE

\mathrm{WUE}_{\mathrm{source}}
=
\frac{V_{\mathrm{site}}+V_{\mathrm{offsite,\ energy}}}
{E_{\mathrm{IT}}}

(V_{\mathrm{offsite,\ energy}}) is water associated with producing energy supplied to the facility, excluding site water already counted in (V_{\mathrm{site}}). The Green Grid distinguishes this expanded operational boundary from site WUE; it still excludes water embodied in construction and IT-equipment manufacturing.

For an all-electric facility with a compatible electricity-water intensity factor:

\mathrm{WUE}_{\mathrm{source}}
=
\mathrm{WUE}
+
\mathrm{PUE}\times f_{\mathrm{water}}

Here, (f_{\mathrm{water}}) is expressed in liters per kWh of facility electricity. This simplified relationship requires the factor to match the supplied electricity and the water-accounting convention. A withdrawal factor and a consumptive-use factor are not interchangeable. Use geographically and temporally appropriate data rather than treating the historical factors in the original WUE paper as universal defaults.

CUE: Count Emissions Before Normalizing Them

\mathrm{CUE}=\frac{C_{\mathrm{operational}}}{E_{\mathrm{IT}}}

(C_{\mathrm{operational}}) is operational greenhouse gas emissions in kilograms of CO₂ equivalent within the declared accounting boundary. (E_{\mathrm{IT}}) is IT energy in kWh over the same period. The Green Grid’s original CUE paper uses kg CO₂e/kWh and excludes lifecycle burdens such as IT-equipment manufacturing and shipping.

For a grid-electricity-only case with no additional direct emissions:

C_{\mathrm{operational}}
=
E_{\mathrm{facility}}\times f_{\mathrm{electricity}}
\mathrm{CUE}
=
\mathrm{PUE}\times f_{\mathrm{electricity}}

The electricity factor is in kg CO₂e/kWh of facility electricity. The Green Grid presents this relationship and requires the factor to reflect the relevant energy supply, location, and reporting period.

The shortcut is incomplete when the site also has generator fuel use or other direct operational emissions. In that case, calculate purchased-energy emissions and direct emissions separately, then combine them without counting the same source twice. Generator testing matters even when generators provide little useful annual electricity; the original CUE paper explicitly identifies maintenance operation of diesel generators as a source requiring inclusion.

Record whether the emissions factors represent CO₂ alone or CO₂ equivalent. Where results use different electricity-accounting methods, retain those labels rather than presenting them as directly comparable CUE values.

Do not replace PUE with ERE in the electricity-emissions calculation. Exported heat does not erase the emissions associated with electricity already consumed. Any estimate of emissions displaced at the receiving building requires a separate counterfactual analysis; ERE does not calculate that benefit.

ERE: Credit Beneficial Energy Export, Not Available Waste Heat

\mathrm{ERE}
=
\frac{E_{\mathrm{facility}}-E_{\mathrm{reuse}}}
{E_{\mathrm{IT}}}

(E_{\mathrm{reuse}}) is energy crossing the data center boundary for beneficial use outside that boundary during the reporting period. Use energy quantities in consistent units; exported thermal energy may be expressed in kWh thermal. The Green Grid’s ERE paper defines this boundary and subtraction method.

The equivalent energy reuse factor relationship is:

\mathrm{ERF}
=
\frac{E_{\mathrm{reuse}}}{E_{\mathrm{facility}}}
\mathrm{ERE}
=
(1-\mathrm{ERF})\times\mathrm{PUE}

With no qualifying reuse, ERE equals PUE. With qualifying export, ERE can fall below 1.0 without implying PUE below 1.0. Report PUE and ERE together because the same ERE can result from efficient infrastructure with modest reuse or less-efficient infrastructure with extensive reuse.

Count the beneficial export, not the heat-recovery equipment’s rated capacity or all heat potentially available from IT equipment. For projected annual ERE, the reuse estimate must reflect when an external demand can accept that energy.

Heat reused within the data center boundary—for example, to heat a battery room—must not receive an additional ERE subtraction when its benefit is already captured in reduced facility energy. Conversely, heat supplied to an unrelated building can qualify as an external reuse flow. The Green Grid gives both cases in White Paper #29, pages 11–13.

ERE measures an energy relationship, not heat value or avoided emissions. It does not establish that low-temperature exported heat replaces an equal quantity of boiler fuel or electricity at the receiving site.

Worked Example: One Denominator, Four Different Results

Consider two hypothetical annual operating scenarios for the same all-electric data center. These are calculation assumptions, not measured facility performance or benchmark targets.

Assume no additional direct operational emissions. Both scenarios use an illustrative electricity factor of 0.40 kg CO₂e/kWh. Site water means metered operational supply, and exported heat means qualifying beneficial reuse measured at the data center boundary.

Annual input Scenario A Scenario B
IT electricity 10,000,000 kWh 10,000,000 kWh
Total facility electricity, including IT 14,000,000 kWh 15,000,000 kWh
Site operational water 12,000,000 L 8,000,000 L
Electricity emissions factor 0.40 kg CO₂e/kWh 0.40 kg CO₂e/kWh
Qualifying exported heat 5,000,000 kWh thermal 5,000,000 kWh thermal

Applying the definitions above:

Metric Scenario A calculation Scenario A result Scenario B result
PUE 14,000,000 ÷ 10,000,000 1.40 1.50
Site WUE 12,000,000 ÷ 10,000,000 1.20 L/kWh 0.80 L/kWh
Operational emissions 14,000,000 × 0.40 5,600,000 kg CO₂e 6,000,000 kg CO₂e
CUE 5,600,000 ÷ 10,000,000 0.56 kg CO₂e/kWh 0.60 kg CO₂e/kWh
ERE (14,000,000 − 5,000,000) ÷ 10,000,000 0.90 1.00

Scenario B uses 4 million fewer liters of site water but 1 million more kWh of electricity. Under the assumed emissions factor, its operational emissions increase by 400,000 kg CO₂e. Its lower site WUE therefore does not establish lower overall environmental impact.

Source WUE cannot be calculated from these inputs because the electricity-supply water factor is missing. Likewise, the heat export alone does not establish avoided emissions at the receiving building. Those conclusions require additional inputs, not a different interpretation of the table.

Compare Like-for-Like Results, Not Universal Rankings

Use these metrics to examine a defined operating change or compare planning scenarios under disclosed assumptions:

  • For infrastructure-energy changes, compare PUE together with absolute facility and IT energy. A smaller IT denominator can increase PUE even when total energy falls.
  • For cooling-water decisions, compare site water, WUE, and energy use together. Lower site water use can shift water demand to electricity generation rather than eliminate it.
  • For emissions decisions, retain the emissions-factor source and accounting basis. A lower CUE can reflect a different energy supply as well as a different PUE.
  • For heat-reuse decisions, report PUE, ERE, and the actual or projected reuse quantity together. ERE alone cannot distinguish infrastructure efficiency from the scale of external reuse.

Calculate period-wide metrics from cumulative numerators and denominators. For example, annual PUE is total annual facility energy divided by total annual IT energy—not the unweighted average of monthly PUE values. Apply the same ratio-of-totals approach to water and emissions intensities.

If IT energy is zero, these metrics are undefined. If it is very small, investigate the denominator before interpreting a large ratio. For projected results, disclose load, weather, water-use, emissions-factor, and heat-demand assumptions near the values they affect.

Keep lifecycle carbon separate from operational CUE. A lifecycle comparison needs its own assessment period and inventory of construction, equipment, replacements, operations, and end-of-life activities. Adding embodied emissions to annual operational emissions and dividing by one year of IT energy does not produce a comparable operational CUE. The Green Grid’s CUE and WUE papers explicitly exclude full lifecycle burdens from these operational metrics.

Continue reading