greenai.guide

Home › Articles › AI and data centre electricity demand

Article

AI and data centre electricity demand

Updated · 7 min read

In short

Data centre energy consumption was about 415 TWh in 2024, around 1.5% of the world's electricity, and the International Energy Agency expects it to roughly double to around 950 TWh by 2030, with AI data centres the fastest-growing part. AI power demand will not break the grid worldwide, but it is already straining grids where data centres cluster, such as Ireland and Virginia, and it is shaping new gas, nuclear and renewable investment.

Key points

  • The IEA estimates data centres used about 415 TWh in 2024 and 485 TWh in 2025, and projects around 950 TWh by 2030, about 3% of global electricity demand.
  • Electricity use by AI-focused data centres grew 50% in 2025, according to the IEA.
  • Data centres used 23% of Ireland's metered electricity in 2025, up from 5% in 2015.
  • US data centres used 4.4% of the country's electricity in 2023; Lawrence Berkeley National Laboratory projects 6.7–12% by 2028.
  • The IEA's range for 2035 is 700–1,700 TWh, and its high-efficiency case is 20% below its base case.
  • In the United States, data centres account for nearly half of electricity demand growth to 2030, the IEA estimates.

Data centres are the buildings full of servers that run cloud services, websites and, increasingly, AI. This page explains how much electricity they use, how fast AI data centres' electricity demand is growing, where it bites, and what grid operators and governments are doing about it.

How much electricity do data centres use?

About 1.5% of the world's electricity, and rising fast. The International Energy Agency (IEA) estimates that data centre energy consumption was around 415 terawatt-hours (TWh) in 2024, about 1.5% of global electricity. Its 2026 update put the figure at 485 TWh in 2025, up 17% in a year.

A terawatt-hour is a billion kilowatt-hours. The IEA says data-centre use has grown by around 12% a year since 2017, much faster than electricity demand overall.

How much of that is AI?

Nobody knows exactly, but AI is the fastest-growing part. Operators don't report AI separately, so there is no precise global figure for AI's share. What the IEA does report is that electricity use by AI-focused data centres surged 50% in 2025, and that AI is the most important driver of growth to 2030. It projects that electricity use by accelerated servers, the chips mainly used for AI, will grow by 30% a year in its base case, against about 9% for conventional servers.

How much will AI power demand grow?

The IEA expects data-centre electricity to roughly double by 2030, with a wide range after that.

Projection Figure Source
World, 2030 Around 945–950 TWh, about 3% of global electricity IEA, 2025 and 2026
World, 2035 700–1,700 TWh depending on scenario; base case about 1,200 TWh IEA, 2025
United States, 2028 325–580 TWh, 6.7–12% of US electricity Lawrence Berkeley National Laboratory, 2024
United States, 2030 4.6–9.1% of US electricity generation Electric Power Research Institute, 2024

The IEA notes that 945 TWh is slightly more than Japan's total electricity consumption today. The spread for 2035 reflects real uncertainty about how fast AI is adopted, how efficient it becomes, and whether chips, power and grid connections can be supplied fast enough.

Where is data centre demand concentrated?

In a few countries and, within them, a few regions. The IEA says the United States and China together account for nearly 80% of global growth in data-centre electricity to 2030, with the US by far the largest share. National averages hide much bigger local effects.

Place Data centres' share of electricity Source
Ireland, 2025 23% of metered electricity, up from 5% in 2015 Central Statistics Office, Ireland
Virginia, US, 2023 Nearly a quarter of state electricity demand Electric Power Research Institute, 2024
United States, 2023 4.4% Lawrence Berkeley National Laboratory
World, 2024 About 1.5% IEA

Ireland's data centres used 7,663 gigawatt-hours in 2025, up from 6,973 in 2024, according to the Central Statistics Office. In the US, the IEA estimates that data centres account for nearly half of all electricity demand growth between now and 2030.

Will AI break the grid?

Not globally, but it is straining local grids where data centres cluster. The IEA expects data centres to account for around one-tenth of global electricity demand growth to 2030, less than industrial motors, air conditioning in homes and offices, or electric vehicles. The problem is local and about timing: the IEA says a typical AI-focused data centre uses as much electricity as 100,000 households, and it wants that power within a few years, while transmission lines and power stations take much longer to build.

The signs of strain are visible:

  • Connection queues. The IEA's 2026 report says developers are constrained by slow grid connections, and recommends reforming how connection queues are managed.
  • Capacity prices. PJM, the largest grid operator in the US, saw capacity prices for much of its area rise from $28.92 per megawatt-day in its 2024/25 auction to $269.92 for 2025/26. PJM attributed this to generator retirements, a higher forecast of peak demand and market rule changes.
  • Rules on new connections. Ireland's energy regulator decided in December 2025 that new data centres must provide new renewable and dispatchable electricity generation, rather than simply drawing on the grid.

How are grids and power companies responding?

With new generation of every kind, and deals signed directly with tech companies. The IEA projects that, to 2035, renewables supply the largest share of extra data-centre electricity, growing by over 450 TWh, with natural gas adding about 175 TWh and nuclear a similar amount. Its 2026 report says 15–27 GW of on-site gas generation could be powering data centres by 2030, mostly in the US.

Some of the deals so far:

  • Nuclear restart. Constellation agreed a 20-year power purchase agreement with Microsoft in 2024 to restart a reactor at the former Three Mile Island site as the Crane Clean Energy Center, about 835 MW, expected online in 2028.
  • Small modular reactors. Google agreed in 2024 to buy up to 500 MW from Kairos Power reactors, with the first by 2030. Amazon announced in October 2026 a plan with X-energy and AEP aimed at more than 5 GW of new nuclear capacity in the US by 2039.
  • Corporate demand. Google's 2026 environmental report says its total electricity consumption grew by 37% in 2025.

New gas plants lock in emissions for decades, and new nuclear and long-distance transmission take years. That is why using less electricity per unit of useful AI work matters now, not later.

Can efficiency keep up with growth?

Efficiency is improving fast, but so far use is growing faster. Google reported that the energy of its median Gemini text prompt fell 33-fold in a single year, yet data-centre electricity use worldwide still grew 17% in 2025. When AI gets cheaper per task, people and businesses use more of it. Economists call this the rebound effect, or the Jevons paradox.

The IEA's own scenarios show the limits of efficiency alone: its high-efficiency case for 2035 is 20% below its base case, a large saving but not a reversal of growth. Efficiency buys time and capacity; whether it reduces total demand depends on choices about how much AI is used and for what. See efficiency is not sufficiency and energy-efficient AI.

What are policymakers doing about data centre electricity?

Moving from encouraging growth to setting conditions on it, with very different approaches by country.

  • European Union. Larger data centres must report energy and sustainability indicators to a common EU scheme under Delegated Regulation (EU) 2024/1364, and providers of general-purpose AI models must document the known or estimated energy consumption of their models under the AI Act (Annex XI).
  • Germany. The Energy Efficiency Act requires new data centres operating from July 2026 to reach a power usage effectiveness (PUE) of 1.2 or better, with tighter limits for existing sites from 2027 and 2030, and requires 100% renewable electricity from January 2027. Check the current text, as amendments have been proposed.
  • Ireland. The regulator's 2025 connection policy requires new large users, including data centres, to bring new generation, with dispatchable capacity for the largest.
  • Singapore. The government paused new data-centre growth in 2019, reopened it in 2022 through a selective call for applications favouring efficient, sustainable proposals, and in 2024 announced at least 300 MW of extra capacity with a push for green energy.
  • United Kingdom. Data centres were designated as Critical National Infrastructure in 2024, and in 2025 the government launched AI Growth Zones designed to scale to 500 MW or more each, the first at Culham in Oxfordshire.

Many governments are trying to attract AI investment and manage its energy needs at the same time. See green AI policy, sovereign AI and sustainability and how to encourage greener AI as a policymaker.

What can organisations and communities do?

Choose where and how AI runs, and take part in local decisions. Organisations buying AI can favour providers that publish energy figures, use efficient data centres and run on clean grids; see green AI hosting and data centres and how to choose a sustainable data centre for AI. People living near proposed sites can respond to planning and grid consultations; see how to have a say on a data centre near me.

What to do next

Frequently asked questions

How much electricity do data centres use?

The IEA estimates about 415 TWh in 2024, around 1.5% of global electricity, rising to 485 TWh in 2025. It expects around 950 TWh by 2030, about 3% of global demand.

How much of data centre electricity is AI?

There is no exact public figure, because operators don't report AI separately. The IEA says AI-focused data centres are the fastest-growing part, with their electricity use up 50% in 2025, and that AI is the most important driver of growth to 2030.

Will AI break the grid?

Not globally: the IEA expects data centres to account for about a tenth of global electricity demand growth to 2030. Locally, grids are already strained where data centres cluster, which is why Ireland now requires new data centres to bring their own generation and Singapore paused new builds from 2019.

Which country uses the most data centre electricity?

The United States, followed by China. The IEA says the two account for nearly 80% of global growth in data-centre electricity to 2030. As a share of national electricity, Ireland is the outlier at 23% in 2025.

How much electricity will AI data centres use in 2030?

The IEA's central estimate for all data centres is around 945–950 TWh by 2030, roughly double 2024. For the US, Lawrence Berkeley National Laboratory projects 325–580 TWh by 2028, 6.7–12% of US electricity.

Are data centres raising electricity bills?

In some markets, demand growth is pushing up capacity costs. PJM, the largest US grid operator, saw its capacity price jump from $28.92 to $269.92 per MW-day in its 2025/26 auction, citing higher forecast peak load, plant retirements and rule changes.

Can efficiency stop data centre electricity growing?

It slows it but has not stopped it. Energy per AI prompt has fallen sharply, but use is growing faster. The IEA's high-efficiency case is only 20% below its base case in 2035.

Sources

  1. Energy and AI: executive summary, International Energy Agency (2025)
  2. Energy and AI: energy demand from AI, International Energy Agency (2025)
  3. Key questions on energy and AI, International Energy Agency (2026)
  4. DOE releases new report evaluating increase in electricity demand from data centers (2024 United States Data Center Energy Usage Report, LBNL), US Department of Energy and Lawrence Berkeley National Laboratory (2024)
  5. Data centres metered electricity consumption 2025: key findings, Central Statistics Office, Ireland (2026)
  6. Large Energy User connection policy decision paper (CRU2025236), Commission for Regulation of Utilities, Ireland (2025)
  7. Powering data centers: U.S. energy system and emissions impacts of growing loads, Electric Power Research Institute (2024)
  8. EPRI study: data centers could consume up to 9% of U.S. electricity generation by 2030, Electric Power Research Institute (2024)
  9. PJM capacity auction procures sufficient resources to meet RTO reliability requirement, PJM Interconnection (2024)
  10. EDB and IMDA launch pilot data centre call for application, Singapore Economic Development Board and IMDA (2022)
  11. Singapore to expand data centre capacity by at least one-third, pushes for green energy use, Singapore Economic Development Board (2024)
  12. Energieeffizienzgesetz (EnEfG), section 11: energy efficiency requirements for data centres, Federal Ministry of Justice, Germany (2023)
  13. Commission Delegated Regulation (EU) 2024/1364 on the first phase of the establishment of a common Union rating scheme for data centres, European Commission (2024)
  14. Data centres to be given massive boost and protections from cyber criminals and IT blackouts, UK Government (2024)
  15. Government fires starting gun on AI Growth Zones, UK Government (2025)
  16. Constellation to launch Crane Clean Energy Center, Constellation Energy (2024)
  17. New nuclear clean energy agreement with Kairos Power, Google (2024)
  18. Amazon, X-energy and AEP nuclear energy agreement in Indiana, Amazon (2026)
  19. Google 2026 environmental report, Google (2026)
  20. Measuring the environmental impact of delivering AI at Google scale, Elsworth et al., Google (2025)
  21. Regulation (EU) 2024/1689 (AI Act), Annex XI, European Union (2024)

Join us

Help cut AI's impact on the planet

Get new guides, funding calls and directory listings by email, about once a month. Or add your product, course, PhD or research to the directories.

No tracking. Unsubscribe in one click. Other ways to join · Privacy