PUE calculator
PUE is one division. The work is deciding what belongs on each side of it.
Power usage effectiveness is total facility energy divided by IT energy. Enter the annual-average power of each supporting system below and the ratio falls out, along with the absolute energy those systems consume — which is the number that actually shows up on an electricity bill.
The ratio is reported everywhere and understood almost nowhere, because a PUE only means something next to the boundary it was measured on. A facility that excludes its office block, or that measures on the coldest week of the year, can publish a number it could not sustain for an hour.
Worked example — a 250 kW hall
A single hall with chilled-water cooling, a double-conversion UPS and no free cooling credited.
- IT load — 250 kW. Annual-average power drawn by servers, storage and network equipment, measured at the IT boundary.
- Cooling and heat rejection — 95 kW. Chillers, pumps, CRAC and CRAH fans, dry coolers and condensers.
- Power conversion losses — 22 kW. UPS double-conversion losses, transformer losses and distribution losses upstream of the rack.
- Lighting and other support — 8 kW. Lighting, fire systems, security, BMS and anything else inside the declared boundary.
- Electricity price — 0.14 $/kWh. Blended delivered rate including transmission and levies.
PUE: 1.500 — 66.7% DCiE — the share of facility energy reaching IT equipment
| Total facility power | 375.0 kW |
|---|---|
| Support power 33.3% of the facility | 125.0 kW |
| IT energy per year | 2,190 MWh |
| Facility energy per year | 3,285 MWh |
| Overhead energy per year Everything the building spends to keep the IT load running | 1,095 MWh |
| Annual electricity cost | $459,900 |
| Cost of the overhead alone | $153,300 |
Formula
PUE = total facility energy ÷ IT equipment energy
DCiE is the reciprocal expressed as a percentage. Both are ratios of energy over a period, not of power at an instant — an instantaneous PUE read at 03:00 in February is not the facility PUE.
What the ratio tends to mean
Directional bands only. A published figure without its boundary, its measurement period and its instrumentation is not comparable with any of these.
| PUE | DCiE | Typically describes |
|---|---|---|
| 1.05 – 1.20 | 83 – 95% | Purpose-built hyperscale with extensive free cooling and a favourable climate |
| 1.20 – 1.40 | 71 – 83% | Modern colocation with containment, EC fans and elevated supply temperatures |
| 1.40 – 1.80 | 56 – 71% | Well-run enterprise room, mechanical cooling most of the year |
| 1.80 – 2.50 | 40 – 56% | Legacy open room, poor airflow hygiene or a heavily part-loaded facility |
| Above 2.50 | Below 40% | Support plant sized for a load the room does not yet carry |
What counts as IT and what counts as overhead
The denominator is the energy delivered to IT equipment. In practice it is measured at the UPS output, at the PDU output, or at the rack — and each of those choices moves the number, because every metre of distribution between them has losses that migrate from the numerator to the denominator depending on where the meter sits. Measuring at the rack gives the most honest answer and the worst-looking ratio.
The numerator is everything inside the declared boundary: cooling and heat rejection, conversion and distribution losses, lighting, fire and security systems, and the BMS itself. Buildings that share plant with offices, labs or a trading floor have to apportion that plant, and how they apportion it is a judgement rather than a measurement.
Why absolute energy belongs next to the ratio
PUE is a ratio, so it can be improved by making the denominator bigger. Adding IT load to a facility whose support plant is already running improves the ratio without saving a single kilowatt-hour, and switching off an idle server worsens it. This is the single most common way the metric is misread.
The calculator reports facility MWh per year and the cost of the overhead alongside the ratio for exactly that reason. If a change improves PUE while total energy rises, the facility got worse and the metric got better.
What this calculator does not tell you
- Carbon. Two facilities with identical PUE can differ by an order of magnitude in emissions depending on grid mix and procurement — that is CUE, a separate metric.
- Water. Evaporative cooling trades electricity for water and improves PUE while consuming a resource the ratio cannot see — that is WUE.
- Useful work. A facility full of idle servers can hold an excellent PUE indefinitely.
- Part-load behaviour. Support plant is rarely linear, so a ratio measured at 40% occupancy will not hold at 90%.
- Seasonality. An annual figure and a January figure are different measurements of different things.
Frequently asked questions
What is a good PUE for a data center?
There is no single answer without a boundary and a climate. A well-run enterprise room in a temperate climate typically lands between 1.4 and 1.8; modern colocation with containment lands between 1.2 and 1.4; purpose-built hyperscale facilities with extensive free cooling report 1.1 and below. Comparing your number with a published one is only meaningful if both were measured at the same boundary over the same period.
How do you calculate PUE?
Divide total facility energy by IT equipment energy over the same period. If the facility consumes 375 kW on average and the IT equipment consumes 250 kW, the PUE is 375 divided by 250, or 1.50. Use energy over a full year rather than power at an instant, because support plant varies with weather and load.
Can PUE be below 1.0?
Not as defined. A value below 1.0 means energy is being credited to the facility that the boundary does not account for — most often heat reuse exported to a district system, or on-site generation counted incorrectly. Operators report a separate energy reuse effectiveness figure for that case rather than pushing PUE below one.
What is the difference between PUE and DCiE?
They carry the same information. DCiE, data center infrastructure efficiency, is the reciprocal of PUE expressed as a percentage: a PUE of 1.5 is a DCiE of 67%, meaning two thirds of the energy entering the building reaches IT equipment. PUE is the more widely used of the two.
Does adding servers improve PUE?
Usually yes, and that is a weakness of the metric rather than a real efficiency gain. Support plant has a large fixed component, so spreading it across more IT load lowers the ratio even when total consumption rises. Always read the ratio beside absolute facility energy.
Should PUE be measured at the UPS or at the rack?
At the rack is the most honest, because losses in the distribution between UPS and rack are genuinely overhead. Measuring at the UPS output moves those losses into the IT side of the division and flatters the result. Whichever you choose, state it beside the number.
Where this calculation stops
This is an arithmetic model of one declared boundary, not a metering strategy. Real measurement requires instrumentation at defined points, an agreed averaging period and a documented treatment of shared plant.
Nothing here is a sustainability result. PUE says how much energy the building spends supporting IT; it says nothing about where that energy came from, how much water it used, or whether the IT load was doing anything worth doing.