Skip to content
OM SoftwareContact
← Back to the blog

What data centers cost in energy, and why it matters

Data centers use about 1.5% of the world’s electricity. What sits behind the numbers, and why storing less data helps the environment.

3 min read

Sustainability · Infrastructure

Every time we stream a video, save a photo “to the cloud”, or ask an AI assistant a question, that work happens somewhere: inside a data center. These are buildings full of always-on computers, and they hold much of what we call the internet. Convenient and invisible — but not free, especially when it comes to energy.

In short. According to the International Energy Agency (IEA), in 2024 data centers consumed roughly 415 TWh of electricity, about 1.5% of the global total, and demand is set to nearly double by 2030. Efficiency per operation has improved a great deal, but the surest way to save energy is still the same: process and store only the data that truly matters.

How much energy it really takes

A data center is a facility that houses large numbers of servers, storage systems and networking equipment: it is the physical backbone of the cloud. Because it concentrates thousands of always-on machines, it needs substantial electricity to run and, in equal measure, to cool down.

The scale is significant. The IEA estimates that in 2024 data centers drew about 415 terawatt-hours, roughly 1.5% of global electricity consumption, with demand growing about 12% per year over the last five years — more than four times faster than overall electricity use. Projections point toward around 945 TWh by 2030, just under 3% of the world’s electricity. Consumption is also highly concentrated: in 2024 the United States, China and Europe together accounted for most of the demand.

What’s driving the surge

Behind these numbers are everyday activities: cloud storage, streaming, online services and, increasingly, artificial intelligence. AI in particular is bending the curve upward: the IEA estimates that electricity use by the “accelerated” servers behind AI is growing around 30% per year, versus ~9% for conventional servers. In 2025 alone, data center electricity demand rose by 17%, well beyond the growth of global demand overall.

The interesting part is that efficiency and growth pull in opposite directions. Individual machines are getting more efficient, but the amount of computing required is rising faster: a classic rebound effect, in which savings are more than offset by higher volumes.

Measuring efficiency: PUE, cooling and water

To measure how well a data center uses energy, the industry relies on PUE (Power Usage Effectiveness): the ratio of the facility’s total energy to the energy that actually reaches the computers (PUE = total energy ÷ IT energy). The ideal value is 1.0 — every watt going to computing, nothing lost to cooling, lighting or power losses. In practice, older facilities ran near 2.0 (half the energy spent on overhead), while the most efficient large operators approach 1.1 — Google, for instance, has reported an average PUE of around 1.1 across its fleet. PUE is standardised internationally (ISO/IEC 30134-2).

After the servers themselves, the biggest energy draw is cooling: keeping the machines at temperature is a constant challenge, which is why much of the innovation focuses on how to dissipate heat. To this we should add a less visible impact — water: many cooling systems consume significant amounts of water, an increasingly discussed factor in assessing the sustainability of these facilities.

The most sustainable data is the data you don’t centralize

Making data centers more efficient is necessary, but not enough: as long as volumes grow faster than efficiency gains, absolute consumption keeps rising. That’s why the most useful question isn’t only “how do we make a data center greener?”, but also “how much data do we actually need to centralize?”.

It’s the criterion OM Software applies to its own apps: let the device people already own do as much of the work as possible, instead of a server that would have to stay on for them. Less infrastructure to power means a lighter footprint.

Building lean, storing less and centralizing only what truly matters isn’t just good engineering: it’s a concrete way to make technology more sustainable, one application at a time.

Sources

Related insights

3 min read

The safest data is the data you never collect

Data minimisation is a legal duty before it is a product decision: collect less, expose less. What it involves, and how it is applied.

4 min read

Where your data lives: on your device or in the cloud?

When an application processes information, that work happens either on the device or on a remote server. Where it happens decides who can reach the data.