Computing, and AI in particular, is a large and fast-growing source of electricity demand and carbon emissions

Observable evidence

Data centres consumed about 415 TWh in 2024, roughly 1.5% of global electricity, and that demand has grown around 12% per year over the last five years. The IEA base case projects it more than doubling to about 945 TWh by 2030, close to 3% of global electricity, and rising toward 1,200 TWh by 2035, with AI the most important driver (IEA, Energy and AI, 2025). The United States, China and Europe together accounted for around 85% of 2024 consumption, and the US and China for close to 80% of the projected growth to 2030.

Why it matters for climate

In emissions terms the near-term share is modest but moving against the trend. The IEA estimates data-centre emissions reach about 1% of global CO2 by 2030 in the central case, or 1.4% in a faster-growth case, and notes this is one of the few sectors where emissions are set to grow while most others decline (Carbon Brief, 2025). Concentration compounds the strain: AI data centres cluster geographically, so local grid impact far exceeds the global average, and in several advanced economies data centres drive 20% or more of electricity-demand growth to 2030. Roughly a fifth of planned data-centre projects risk grid-connection delays.

The efficiency paradox

Per-task efficiency is improving quickly, in some measures by around an order of magnitude per year, so a simple text query now uses less electricity than running a television for the same time. Total demand still climbs, because query volume and new energy-intensive uses (video generation, reasoning, agents) grow faster than efficiency, and those heavier tasks can use hundreds to thousands of times more energy than a simple query (IEA, Key Questions on Energy and AI, 2025).

Who is affected

Developers and operators; the organisations accountable for the resulting emissions; grid operators and communities near data-centre clusters; and everyone downstream of the climate impact and of the fossil capacity being added to serve this load.

Scope

This issue is the footprint of computing itself, spanning both operational energy and the embodied impact of the hardware. It excludes the separate debate over whether particular AI applications are worthwhile. Related but distinct facets (measurement, grid and water strain, hardware lifecycle) are handled as sub-issues.

Impact if unaddressed

Rising, geographically concentrated demand risks extending reliance on fossil generation, delaying grid decarbonisation, and making computing a growing rather than shrinking contributor to warming.