PUE Calculator — Power Usage Effectiveness
Calculate Power Usage Effectiveness and DCiE from total facility power and IT equipment power, with overhead breakdown.
Inputs
Everything drawn at the utility meter: IT, cooling, lighting, losses.
Servers, storage and network gear, measured at the PDU output.
Results
- PUE (Power Usage Effectiveness)
- 1.50
- DCiE (infrastructure efficiency)
- 66.7%
- Non-IT overhead
- 50.0kW
- Overhead share of total
- 33.3%
Average — typical enterprise room
Share of facility power that reaches IT equipment.
What PUE measures
Power Usage Effectiveness is the ratio of everything a facility draws to the portion that actually reaches IT equipment:
PUE = total facility power ÷ IT equipment power
A PUE of 1.5 means that for every kilowatt delivered to servers, another half kilowatt is spent on cooling, power conversion, lighting and everything else. The theoretical floor is 1.0 — a facility that spends nothing on overhead — which is unreachable in practice because power distribution alone has losses.
The reciprocal, expressed as a percentage, is DCiE (Data Center infrastructure Efficiency). A PUE of 1.5 is a DCiE of 66.7%, meaning two thirds of incoming power does useful work. Both numbers describe the same thing; PUE is simply the one that stuck.
Where to measure
The two figures must be taken at the right boundaries or the ratio is meaningless:
- Total facility power is measured at the utility meter, before anything in the building consumes it.
- IT equipment power is measured at the output of the PDUs, after UPS and transformer losses. Measuring it at the UPS input instead quietly flatters your PUE by counting conversion losses as IT load.
ISO/IEC 30134-2 formalises this and defines measurement categories depending on where and how often you sample. A number quoted without that context is marketing, not engineering.
What counts as a good number
| PUE | DCiE | Interpretation |
|---|---|---|
| 1.0 - 1.2 | 83 - 100% | Hyperscale, purpose-built, often free-cooled |
| 1.2 - 1.5 | 67 - 83% | Efficient modern colocation facility |
| 1.5 - 2.0 | 50 - 67% | Typical enterprise data center |
| Above 2.0 | Below 50% | Legacy room, often a converted office space |
Why annual PUE is the only honest PUE
PUE moves with the weather and with load. A facility using outside-air economisers might hit 1.15 on a cold January night and 1.6 during a July afternoon. Measuring once and quoting the good number is the most common abuse of the metric.
It also degrades badly at low load. A room designed for 500 kW but running 50 kW still pays most of its fixed overhead — fans, lighting, UPS idle losses — so its PUE looks terrible even though nothing is wrong. This is why a half-empty data center almost always reports a worse PUE than a full one with identical equipment.
What PUE does not tell you
- Whether the IT load itself is efficient. Replacing idle servers with fewer, busier ones reduces energy use but can make PUE worse, because IT load falls while overhead stays fixed.
- Anything about carbon. A coal-powered facility and a hydro-powered one can post identical PUE. CUE (Carbon Usage Effectiveness) covers that.
- Water consumption. Evaporative cooling improves PUE precisely by trading electricity for water. WUE is the counterpart metric.
Use PUE to track your own facility over time, where the comparison is like-for-like. Use it cautiously to compare facilities, and never without knowing the measurement boundary and the period.