IEA Projects Global Data-Centre Electricity Demand to Reach 945 TWh by 2030
AI-focused data centre electricity demand soared 17% in 2025; the IEA says global grid investment must rise 50% to support a near-doubling by 2030.
Global data-centre electricity consumption is set to more than double to 945 TWh by 2030, the International Energy Agency projected, with artificial intelligence identified as the most important driver of that increase. Asian Power reported on Monday (2026-09-14) that AI workloads can draw 40kW per rack or more, compared with the 5 to 15kW around which many existing facilities were designed, a difference that points to the infrastructure mismatch forming across the sector.5
Electricity demand from data centres jumped 17% in 2025, with AI-focused facilities growing even faster, well outpacing the 3% rise in total global power consumption that year, the IEA said — a divergence that signals a structural shift in where and how densely electricity loads sit, not simply a cyclical increase in volume.4
The IEA's 2026 update, "Key Questions on Energy and AI," put per-rack intensity in household terms: a single server rack in an advanced data centre could carry a peak power demand equivalent to 65 households by 2027.2 Hong Kong's data-centre pipeline illustrates what that implies at city scale. CBRE estimated active capacity at 687MW in the first quarter of 2026, with a further 653MW in development, all of it built around power-density assumptions that advanced AI hardware has already exceeded.5
Hong Kong's decarbonisation timeline adds another constraint. Nuclear and renewables together accounted for 28% of the city's sent-out electricity mix in 2025, largely through imported nuclear power, and the government has targeted a 60 to 70% zero-carbon share of generation by 2035, with 7.5 to 10% from local renewables.5 Absorbing AI-driven load growth while shifting the fuel mix on that timetable is a harder task than either objective alone.
But the grid-investment shortfall runs wider than any single market. Spending on electricity generation has surged nearly 70% since 2015, yet outlays on transmission and distribution networks grew at less than half that rate over the same period, the IEA reported.1 To absorb demand growth through 2030, the IEA estimates annual grid investment would need to rise by about 50% from a baseline of $400 billion.3
Global electricity demand is growing at the fastest pace in 15 years. The IEA projects an annual average growth rate of 3.6% between 2026 and 2030, driven by industry, electric vehicles, air conditioning and data centres.3 The 945 TWh data-centre figure sits inside that broader demand acceleration, which means the grid investment case does not rest solely on AI adoption. Power utilities seeking to plan capacity will need to size for the combined load, not treat data-centre demand as a separate planning problem.
Capital flows confirm the shift. Global data-centre investment reached USD 580 billion in 2025, overtaking the USD 540 billion deployed on oil supply that year, the IEA noted.1 The crossover reflects how quickly digital infrastructure has come to compete with hydrocarbons for large-scale industrial capital, though the scale of that spending says nothing directly about whether the money is reaching the right points in the network.
The IEA also stated that data-centre power demand would rise by a further 75% in 2026 alone.4 If that pace holds anywhere close to forecast, the 945 TWh projection for 2030 may prove conservative. Utilities and grid planners have consistently underestimated AI load growth; if generation and transmission capacity is not expanding fast enough to meet the loads that data-centre developers are committing to now, the shortfall will become visible at peak demand, when it is hardest to manage.4,3