Data centre electricity use surged last year despite power consumption per AI task declining rapidly, according to new figures from the International Energy Agency, a result of the hundreds of billions that tech firms invested in expanding their physical footprints.

The IEA’s report, Key Questions on Energy and AI, found that the combined capital expenditure of the world’s largest technology companies, including Amazon, Google, Meta, and Microsoft, surpassed $400 billion in 2025, largely due to data centre investment, pushing global electricity demand for these facilities up by 17%.

According to the IEA, electricity demand from AI‑driven data centres in particular skyrocketed 50% last year, far outpacing the 3% growth in global electricity demand, a symptom of big tech’s push to expand AI infrastructure as both enterprise and consumer appetite grow.

With AI adoption still gathering pace, and energy-intensive uses like AI agents on the rise, the IEA said electricity consumption by AI-focused data centres is set to triple by 2030 due to an influx of new users, new applications, and businesses pushing to scale the technology as fast as possible.

The report claims that AI will continue to be resource hungry despite energy efficiency “improving at a rate unprecedented in energy history”, with simple text‑based AI queries now using less electricity than running a TV for the same period.

As these efficiencies scale, the IEA said that AI could even help firms in the most energy-intensive industries reduce their energy costs by up to 10%, with the right mix of policies and infrastructure helping to keep costs stable elsewhere.

Significant challenges remain, however, as data centres increasingly come up against physical bottlenecks, limiting the rate at which they can expand in the near term, an issue becoming acute as supply chains for essential components like gas turbines and transformers, not to mention chips and IT hardware, have tightened over the past year.

The flood of new data centre projects is also straining global planning and regulatory systems, with grid connections and other necessary approvals slowing to a snail’s pace – according to the UK Government, the queue for connection to the National Grid network climbed 460% over the first half of 2025, with projects now facing waits of up to fifteen years. 

These delays are driving more developers, particularly in the US, to explore on‑site gas power, but rapid swings in demand mean meeting their power needs can stretch the capabilities of on-site gas plants, added to which many of these facilities remain in their early stages due to technical and financial hurdles.

Chasing reliability, tech firms are increasingly turning to on-site battery storage, as well as nuclear and advanced geothermal solutions.

Recommended reading

• Experts Call for Urgent Action to Help UK Meet Net Zero Targets
• Will AI Energy Demand Undermind Climate Gains?
• Comment | Building Adaptive Energy Systems with AI

The IEA said that the pipeline of conditional offtake agreements between data centre operators and small modular reactor nuclear projects has grown from 25 gigawatts at the end of 2024 to 45 gigawatts today, indicating that the momentum behind AI could accelerate the commercialisation of new energy technologies. 

“We see that 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 Fatih Birol, IEA executive director.

To help countries that seize on this opportunity to modernise their energy systems, and to tackle bottlenecks and other concerns associated with AI’s rapid growth, collaboration between policymakers and the energy and tech sectors remains crucial.”