IEA publishes the Energy and AI special report on data centre electricity demand

Background

In April 2025 the IEA (International Energy Agency) published a special report, Energy and AI, which the agency described as the most comprehensive data-driven global analysis of the energy-AI nexus to date.^[3] The report set out how much electricity data centres use today, how much they are projected to use over the following decade, and where that power is expected to come from.

The starting point: about 1.5% of world electricity

In 2024, data centres accounted for around 1.5% of the world's electricity consumption, or about 415 terawatt-hours.^[1] A terawatt-hour is a billion kilowatt-hours, so this is a national-scale quantity of power. Behind the modest-looking share sits a steep trend: global data centre electricity consumption had grown by around 12% per year since 2017, more than four times faster than total electricity consumption.^[1]

The load is also concentrated geographically. The United States accounted for the largest share of global data centre electricity consumption in 2024, at 45%, followed by China at 25% and Europe at 15%.^[1]

The projection: doubling by 2030

The IEA projects that data centre electricity consumption will more than double to around 945 TWh by 2030 — slightly more than Japan's total electricity consumption today — and reach around 1,200 TWh by 2035.^[1] The main driver is AI hardware: electricity consumption from accelerated servers, driven mainly by AI adoption, is projected to grow about 30% annually in the report's Base Case and to account for almost half of the net increase in global data centre electricity consumption.^[2] Accelerated servers are machines built around specialist chips for training and running AI models rather than around general-purpose processors alone; a Base Case is a scenario that extends current trends and announced policies, not a claim about a single outcome.

The scale is easier to picture one facility at a time. A typical AI-focused data centre consumes as much electricity as 100,000 households, and the largest ones under construction will consume 20 times as much.^[1]

For the United States the numbers are large relative to the rest of the power system. Data centres account for nearly half of electricity demand growth to 2030, and by then the country will consume more electricity for data centres than for producing aluminium, steel, cement, chemicals and all other energy-intensive goods combined.^[1]

Investment and where the power comes from

The demand is backed by money: global investment in data centres nearly doubled since 2022 and amounted to half a trillion dollars in 2024.^[1]

On the supply side, the report projects renewables meeting about half of data centre demand growth to 2035, with natural gas expanding 175 TWh and nuclear contributing a similar amount; the first small modular reactors (SMRs) come online around 2030.^[1] Small modular reactors are reactors of smaller capacity, designed to be factory-built and assembled on site.

Why it matters

Numbers like these turn a vague argument about AI's appetite for power into something that can be checked. A load that was around 1.5% of world electricity in 2024 and is projected to more than double by 2030 is no longer a minor line in electricity planning: it is a demand segment that grids, generators and governments have to count separately from households and industry.^[1] The report also shows that the growth is unevenly distributed, concentrated in the United States, China and Europe, so pressure on grids will be felt most in those places.^[1]

It is worth keeping in mind that these figures are projections under stated assumptions rather than measured outcomes.