For two years the debate about artificial intelligence and electricity has run on company announcements, utility pipelines and back-of-envelope arithmetic. On 10 April the International Energy Agency published its special report Energy and AI, the first attempt by a major intergovernmental body to build a consistent global baseline. Its central projection is easy to remember. Electricity consumption by data centres worldwide, about 415 TWh in 2024, is set to more than double to around 945 TWh by 2030, slightly more than Japan's total electricity consumption today.
The report's value lies less in the headline than in the breakdown. It tells us where the load is, how fast it has been growing, how it is likely to be supplied and where the bottlenecks sit. It also sets out the uncertainty honestly, with a range of cases that differ by a factor of more than two by 2035.
Where the load sits today
Data centres used about 1.5 per cent of the world's electricity in 2024. The United States accounted for 45 per cent of that consumption, China for 25 per cent and Europe for 15 per cent. In other words, 85 per cent of global data centre demand is concentrated in three markets. Globally, consumption has grown by about 12 per cent a year since 2017, more than four times faster than total electricity consumption.
Emerging and developing economies other than China account for half of the world's internet users but less than 10 per cent of global data centre capacity. That gap matters for the next decade, because it means that much of the world's digital activity is hosted far from where the users are, and that countries hoping to build domestic AI capacity will first need reliable and affordable power.
Where the growth goes
AI is the most important driver of growth, with electricity demand from AI-optimised data centres projected to more than quadruple by 2030. The United States accounts for by far the largest share of the projected increase, followed by China. In the United States, data centres account for nearly half of electricity demand growth between now and 2030, and by the end of the decade the country is set to use more electricity for data centres than for producing aluminium, steel, cement, chemicals and all other energy-intensive goods combined.
The country-level effects are where the report becomes useful for policy. Fatih Birol, the IEA's executive director, said data centres are on course to account for almost half of electricity demand growth in the United States, more than half in Japan, and as much as one-fifth in Malaysia. In advanced economies as a group, data centres account for more than 20 per cent of demand growth to 2030, after decades of essentially stagnant electricity demand. In emerging and developing economies, where demand is already growing quickly, data centres account for about 5 per cent of the increase.
Globally, the picture is more modest than the headlines suggest. Data centres account for around one-tenth of world electricity demand growth to 2030, less than industrial motors, air conditioning or electric vehicles. The pressure is local and regional, not global, which is why the policy response has to be local and regional as well.
How it gets supplied
The IEA expects renewables and natural gas to take the lead. Half of the global growth in data centre demand is met by renewables, supported by storage and the wider grid, and renewable generation is projected to grow by over 450 TWh to meet data centre demand to 2035. Natural gas expands by 175 TWh, particularly in the United States. Nuclear contributes about the same amount of additional generation, particularly in China, Japan and the United States, with the first small modular reactors arriving around 2030.
That mix reflects what can be built fast enough. Renewables have short lead times and the large technology companies have been the biggest corporate buyers of clean power. Gas offers firm supply in markets with cheap fuel. Nuclear contributes mostly through existing plants, uprates and restarts before new designs arrive at scale.
The bottlenecks
The section on grids is the most sobering. The IEA estimates that, unless risks are addressed, around 20 per cent of planned data centre projects could face delays. Building transmission can take four to eight years in advanced economies, and wait times for transformers and cables have doubled in the past three years. Gas turbine deliveries face lead times of several years, potentially pushing some new plants beyond 2030.
The report's suggested responses are practical: locate new data centres where power and grid capacity are available, and operate servers or on-site generation and storage more flexibly. It acknowledges a catch. An AI-focused data centre is ten times more capital-intensive than an aluminium smelter, so curtailing it is expensive. Many facilities, however, run with a buffer of spare server capacity, and regulators could encourage operators to use that buffer, or their backup generation and storage, to support the grid. The IEA also finds that half of the data centres under development in the United States are in pre-existing large clusters, which concentrates the strain on particular grids.
The range of outcomes
The report's sensitivity cases deserve as much attention as the base case. By 2035 the base case reaches around 1,200 TWh. A Lift-Off Case with faster AI uptake and fewer bottlenecks goes higher, and a Headwinds Case with slower uptake and infrastructure constraints goes lower. A High Efficiency Case, with faster gains in hardware and model efficiency, sees demand 20 per cent below the base case in 2035. Across all cases, data centre demand in 2035 spans 700 to 1,700 TWh.
On emissions, the IEA is measured. Emissions from data centre electricity use rise from 180 million tonnes today to 300 million tonnes by 2035 in the base case, and up to 500 million tonnes in the Lift-Off Case, still below 1.5 per cent of energy sector emissions. Broad adoption of existing AI applications could cut emissions by an amount equivalent to about 5 per cent of energy-related emissions in 2035, far more than data centres emit but far less than climate goals require.
What the baseline changes
The report gives governments, utilities and investors a shared reference point. It also shifts the burden of proof. Claims that AI will consume a large share of the world's electricity now have to be squared with a 1.5 per cent share today and only around one-tenth of global demand growth to 2030. Claims that AI is a marginal issue have to be squared with the United States, Japan and Malaysia, where it is the main source of new demand. The truth, as so often in energy, depends on where you stand.