JupiteX Get the app
Environment & Climate19 Nov 2026 · about 7 min

Datacentres to create 70% more carbon emissions by 2030, IEA report warns

The brief

The IEA predicts emissions associated with datacentres will rise by 70% by 2030. This is a projection for the sector, not a claim that every datacentre will emit more. It matters because these facilities are expanding quickly as AI use grows, increasing demand for electricity that must be generated somehow. The key mechanism is the electricity supply. Datacentres need power for their servers, and the IEA expects fossil fuels to supply a significant share of the additional demand in the US and China. It forecasts that coal will provide about half of the electricity for China’s AI datacentres, while gas will supply about 60% of US datacentre electricity by 2030. Developers are also major corporate buyers of renewable energy, but the IEA still expects fossil-fuel appetite to grow faster. Its projection reflects that mix of power sources and the sector’s rapid expansion. The report says AI accounts for three-quarters of datacentres’ electricity growth by 2030.

01

What does the IEA predict will happen to datacentre emissions by 2030?

The IEA predicts emissions associated with datacentres will rise by 70% by 2030. This is a projection for the sector, not a claim that every datacentre will emit more. It matters because these facilities are expanding quickly as AI use grows, increasing demand for electricity that must be generated somehow.

The key mechanism is the electricity supply. Datacentres need power for their servers, and the IEA expects fossil fuels to supply a significant share of the additional demand in the US and China. It forecasts that coal will provide about half of the electricity for China’s AI datacentres, while gas will supply about 60% of US datacentre electricity by 2030.

Developers are also major corporate buyers of renewable energy, but the IEA still expects fossil-fuel appetite to grow faster. Its projection reflects that mix of power sources and the sector’s rapid expansion. The report says AI accounts for three-quarters of datacentres’ electricity growth by 2030.

02

How much more electricity will AI datacentres use each year by 2030, and how does that compare with Great Britain’s annual use?

The IEA estimates that AI datacentres will add 350 terawatt-hours to annual electricity use within four years. That is more electricity than Great Britain consumes in a whole year. The comparison gives a sense of scale: growth in one part of the digital economy could require power on the scale of a country’s yearly use.

The increase is tied to the fast expansion of AI computing. The IEA says AI accounts for three-quarters of the growth in datacentre electricity demand by 2030. Its estimates draw on annual shipments of AI servers, which have risen eighteen-fold since 2020, and satellite images tracking 93 large AI factories.

The IEA also forecasts that datacentres overall will consume twice as much electricity in 2030 as they do today. This new demand will need to be met by power supplies that vary by region. The agency expects fossil fuels to play a substantial role, especially in the US and China, even as renewables supply much of the growth elsewhere.

03

Where will datacentres in the US, China, and the EU get most of their electricity by 2030?

The IEA expects the main electricity sources for datacentres to differ sharply by region. China’s AI datacentres are forecast to get about half their electricity from coal. US datacentres are expected to rely chiefly on gas. In the EU, renewables are projected to supply about 70% of datacentre electricity by 2030.

The forecasts reflect planned power generation linked to the fast-growing industry. The IEA identified plans for about 180 gigawatts of new gas-fired plants linked to US datacentres. In China, the remainder of AI datacentre power after coal is expected to come from renewables, with smaller amounts from nuclear and very little from gas.

The EU’s remaining datacentre electricity is expected to come from nuclear power, gas and a very small amount of coal. These are projections, not guarantees about every facility’s supply. Overall, the IEA expects renewables to play a greater role outside the US and China, while fossil-fuel generation grows particularly quickly in those two countries.

04

How can datacentre growth increase emissions even when developers buy renewable electricity?

Datacentre growth can increase emissions because the facilities need large amounts of electricity, and generating that electricity can produce emissions. Developers are major corporate buyers of renewable energy, but the IEA still expects emissions associated with the sector to rise. The important issue is not only who buys renewable power, but also what sources generate the electricity used to meet growing demand.

The article says datacentre servers connect to local power grids. It also forecasts substantial fossil-fuel generation for these facilities: coal is expected to supply about half of China’s AI datacentre power, and gas about 60% of US datacentre power by 2030. These examples show how fossil fuels can remain part of the supply as electricity needs rise.

The IEA projects datacentre emissions will increase 70% by 2030. It expects renewables to have a larger role outside the US and China, including about 70% of EU datacentre electricity. But renewable purchases and expanding fossil-fuel generation can coexist, so the sector’s overall emissions depend on its growth and the power mix supplying it.

05

What is a datacentre, and what does it do?

A datacentre is a facility that contains computer servers and the equipment needed to operate them. Servers are computers that process requests, run software and store or handle data for digital services. This is a general definition from established computing knowledge; the article does not define the term directly.

The article focuses on datacentres built for AI. It describes their servers as connected to local power grids and explains that AI servers are driving much of the industry’s electricity growth. The IEA tracked annual shipments of AI servers and used satellite imagery to follow construction of 93 large AI factories around the world.

The scale of these facilities helps explain why their energy use matters. The IEA estimates existing global datacentre capacity at 120 gigawatts, including AI-optimised servers, with 80% in China, the US and the EU. Under its central scenario, another 120 gigawatts will be built by the end of the decade, three-quarters of it in AI datacentres. The agency expects total datacentre electricity consumption to double by 2030.

06

Why do AI servers require so much electricity compared with ordinary computing equipment?

AI servers need electricity to run their processors, memory and other computing equipment. AI workloads can involve many calculations across large sets of data, especially when training models or generating responses. This general explanation comes from established computing knowledge; the article does not detail the technical reasons for AI servers’ energy use or compare them directly with ordinary equipment.

The article does show how quickly AI computing is expanding. Annual global shipments of AI servers have climbed eighteen-fold since 2020. The IEA says AI accounts for three-quarters of the projected growth in datacentre electricity use, and forecasts an extra 350 terawatt-hours a year for AI datacentres within four years.

More servers and AI factories mean greater demand for electricity, though the article gives no per-server power figures. It says datacentre capacity has reached 17 gigawatts after climbing tenfold in 24 months. By 2030, total datacentre electricity consumption is expected to be twice today’s level, with the power mix varying across regions.

07

What determines how much carbon is emitted when electricity is used?

Electricity use does not have one fixed carbon impact. As a general principle from established energy knowledge, emissions depend on the amount of electricity consumed and the emissions intensity of the sources generating it. Coal and gas generation release carbon dioxide, while wind and solar generation do not release carbon dioxide while producing electricity. The article does not set out a calculation method.

Its regional forecasts illustrate why the power mix matters. The IEA expects coal to provide about half of China’s AI datacentre electricity by 2030, and gas to provide about 60% of US datacentre electricity. In the EU, renewables are expected to supply about 70% of datacentre power, followed by nuclear, gas and very little coal.

The IEA says emissions associated with datacentres are on track to rise 70% by 2030 as electricity demand increases and fossil generation grows. Developers’ renewable-energy purchases are part of the picture, but the article’s forecasts also show rising fossil-fuel use. Actual emissions therefore depend on both the scale of electricity demand and the sources supplying it.

This brief was written by AI from the original reporting and checked by other models. Names, figures and quotes come from the source; read it for full context.

Read more in the JupiteX app

Pulse is free. New stories every 4 hours, each one broken into the questions that explain it.

Or read more news on the web