Electricity demand from data centres rose by 17% in 2025, according to a new report from the International Energy Agency published on 16 April 2026. Demand from data centres focused on artificial intelligence grew even faster. Both outpaced the 3% growth in global electricity demand. The report, titled Key Questions on Energy and AI, builds on the IEA's first Energy and AI report from April 2025.
The IEA's data show global data centre electricity consumption of about 485 TWh in 2025, with AI-focused data centres accounting for about 155 TWh, roughly a third of the total.
Investment surge
The capital expenditure of five large technology companies rose to more than $400 billion in 2025 and is set to increase by a further 75% in 2026, according to the IEA. Most of that spending goes to data centres, chips and related infrastructure.
The IEA expects electricity consumption from data centres to double by 2030, and consumption by AI-focused data centres to triple. The agency notes that energy use per AI task is falling rapidly, with efficiency improving at a rate it describes as unprecedented in energy history. However, more people are using AI, and energy-intensive applications such as AI agents are growing, so total demand continues to rise.
Physical bottlenecks
The report finds that AI deployment is increasingly running into physical constraints that limit how fast data centres can expand. Supply chains for gas turbines and transformers, as well as for advanced chips and IT components, have tightened over the past year. The growing pipeline of data centre projects is straining planning and regulatory systems, holding up grid connections and other approvals.
Constrained by slow grid connections, developers are advancing many projects with on-site natural gas generation, mostly in the United States. The IEA used satellite-based tracking to assess these projects and found that many remain at early stages, which it said highlights the technical and financial hurdles involved. AI data centres have rapid and large swings in demand, which can stretch the capabilities of on-site gas plants. As a result, the IEA says, on-site battery storage is becoming a critical technology for the next generation of AI data centres.
Clean energy procurement
The technology sector accounted for around 40% of all corporate power purchase agreements for renewables signed in 2025, according to the report. The sector has also become a major source of momentum for nuclear power and advanced geothermal. The pipeline of conditional offtake agreements between data centre operators and small modular reactor projects grew from 25 GW at the end of 2024 to 45 GW at the time of the report.
"While AI is still an energy taker, it is also becoming an energy maker, driving forward innovative solutions like next-generation nuclear reactors, flexible data centres and long-duration energy storage," said IEA Executive Director Fatih Birol.
Affordability concerns
The report notes that social concerns about AI have grown, with data centres becoming a visible flashpoint for worries about electricity prices and the environment. The IEA finds that rising electricity demand does not necessarily raise prices if the right mix of policies and infrastructure investment is in place. However, data centres can create particular challenges for affordability because they are large, concentrated loads that scale up quickly, often requiring new generation and grid investment.
The IEA says policymakers have tools to manage these issues, including encouraging smart integration of data centres into grids and incentivising them to operate more flexibly. In the United States, grid operators such as PJM and ERCOT have been developing rules for large loads, including requirements for flexibility and for new loads to bring their own generation in some cases.
Emissions implications
The report's findings matter for emissions. Where data centre demand is met by new gas generation, as with many on-site projects in the United States, emissions rise. Where it is met by new renewables, nuclear or geothermal, the emissions effect is limited. The technology companies building most AI capacity have climate targets, and their sustainability reports have shown rising emissions as data centre construction and electricity use grow. Microsoft, Google and Amazon have each reported increases in emissions linked to data centre expansion in recent years.
In its 2025 report, the IEA estimated that emissions from data centres would rise from around 180 million tonnes of CO2 to 300 million tonnes by 2035 in its base case, a small share of global energy emissions but one of the fastest-growing. The agency also argued that AI applications could reduce emissions elsewhere in the energy system if widely adopted.
Regional concentration
Data centre demand is highly concentrated. The United States and China account for most global consumption, and within the United States a handful of regions, including northern Virginia, Texas, Georgia, Ohio and Arizona, host a large share of new capacity. That concentration means the effect on local grids can be much larger than the global numbers suggest. In Ireland, data centres already account for a large share of national electricity use, and the grid operator has restricted new connections in the Dublin area. Local concentration is one reason the IEA emphasises grid connection queues and planning systems as bottlenecks.
AI in the energy sector
The report finds that proven AI applications could help firms in energy-intensive industries reduce energy costs by 3 to 10 percentage points. However, the energy sector is not yet taking full advantage of AI, according to the IEA, with a lack of digital skills and data availability emerging as key barriers. Birol announced that the IEA will launch a platform for governments and industry to discuss energy and AI issues regularly.
Climate and water
The IEA report focuses on electricity, but data centres also use water for cooling, either directly through evaporative systems or indirectly through the water used by power plants. During heatwaves, cooling demand rises and water constraints can tighten. Several jurisdictions, including the EU, now require data centres to report water use, and companies have been shifting to designs that use less water in water-stressed areas.
What to watch
Key items include how quickly gas turbine and transformer supply chains expand, the progress of on-site generation projects, the development of large-load rules in US grid markets and how technology companies' emissions develop as their 2026 sustainability reports are published in the summer.