PUE

in English: power usage effectiveness

advancedwidespread

PUE shows how many times the total power consumption of a data center exceeds what the computers themselves use.

In plain terms

In a data center, not only the servers consume electricity. Cooling, ventilation, uninterruptible power supplies, conversion losses, and lighting also use energy. The PUE tells you in a single number how large this 'utility bill' is compared to useful work.

It is like a restaurant's electricity bill. The oven and stove prepare the food, but the cold storage, extractor hood, and lights also consume power. If the total bill is one and a half times the cost of cooking, the restaurant's PUE is 1.5.

The theoretical perfect value is 1.0: in this case, all power would go to the computers. In reality, the best giant data centers are around 1.1, the average is closer to 1.5, and old, poorly cooled server rooms can be above 2.

It became a headline figure due to AI. Training and running large language models require vast amounts of electricity, so it matters how much of this is used purely for cooling.

How it works

They measure the total power consumption of the data center, typically at the facility's main power feed, preferably over a full year.
They separately measure how much electricity actually reaches the IT devices: the servers, the data storage units, and the network equipment.
Total consumption is divided by IT consumption. If the building consumes 1500 kWh and 1000 kWh of that goes to the machines, the PUE is 1.5.
The portion above 1 represents loss and servicing: at 1.5, for every kilowatt-hour spent on machines, an additional half kilowatt-hour goes to cooling and other operations.
Operators try to lower this number, for example with outside-air or liquid cooling, better airflow, and more efficient power supplies.

An everyday example

You read in the news that a tech company is building a new AI data center nearby and promises a 1.2 PUE. This means that for every unit of electricity consumed by the machines, only 20 percent goes to cooling and servicing. However, it does not reveal how much electricity the facility consumes in total.

Who is affected

entrepreneur, executiveWhen choosing a cloud provider or data center, the PUE affects both your costs and your company's sustainability report.

industrial worker, engineerIn the design and operation of data centers, this is one of the most important target figures used to size cooling and power supply. → Industry

lawyer, citizenIn the EU, data centres with an IT power demand of at least 500 kW must report their energy metrics, including PUE, every year, so this is data that can be demanded in debates about local investments. → Law

everyoneBehind every question you ask an AI, a data center is at work, and its power demand strains both the grid and the climate.

A common misconception

MythA low PUE means the data center is green and consumes little power.

In realityPUE only shows a ratio: a giant facility with a value of 1.1 can consume as much electricity as all households in a large city. It also says nothing about whether the electricity comes from a coal power plant or solar panels, or whether the machines are doing useful work.

Limits and risks

It does not measure how efficient the servers themselves are: wasteful, partially idle machines can still achieve a good PUE.
The value depends on the climate, the season, and the measurement method, so figures from two data centers can only be compared cautiously.
Companies often report the planned or best-moment measured value, not the annual average, so always ask which one you are seeing.
It says nothing about water consumption, even though evaporative cooling can improve the PUE while using large amounts of water.

Similar, but not the same

Water usage effectiveness (WUE) – WUE measures how much water a data center uses relative to the machines' consumption, while PUE only accounts for electricity.
Carbon Usage Effectiveness (CUE) – CUE measures emissions, so the power source matters; with PUE, the source is irrelevant.
Carbon footprint – Carbon footprint is the total amount of emissions; PUE is merely an efficiency ratio.

Related concepts

Broader

Data center, Energy efficiency, Sustainable IT

Narrower

Liquid Cooling, Free cooling, Waste heat recovery

On the edge

Water usage effectiveness (WUE), Carbon Usage Effectiveness (CUE), AI energy demand, Hyperscale data center, computing capacity

A brief history

The metric was popularized in 2007 by the industry alliance called The Green Grid. In 2016, it became an international standard with the designation ISO/IEC 30134-2.

Where things stand

By the end of 2026, PUE is the best-known efficiency metric for data centers and is also part of the EU reporting obligation. With the spread of AI data centers, it is increasingly required together with water-consumption and emissions metrics, because on its own it is not enough.

Status: 2026 November

Also known as

PUE · PUE value · power usage effectiveness · energy efficiency · datacenter energy efficiency · PUE metric

Sources

Wikipedia: Power usage effectiveness

This entry was machine-edited and has undergone machine expert review; the “In the news” section updates automatically from Agent Age stories.