The rapid expansion of artificial intelligence is placing unprecedented pressure on electricity grids worldwide, raising the risk of delays to data centre projects and creating new challenges for energy systems, the International Energy Agency (IEA) has warned.

In a new report, the agency said recent advances in AI, driven by falling computation costs, increased data availability and major technical breakthroughs, have triggered a sharp rise in electricity demand, particularly from data centres.

“Electricity grids are already under strain in many places,” the IEA said, cautioning that unless these pressures are addressed, “around 20 per cent of planned data centre projects could be at risk of delays.”

The agency explained that long and complex grid connection queues, alongside the lengthy process of building new transmission lines and installing critical equipment such as transformers and cables, are major bottlenecks. These challenges could force trade-offs with other energy sector priorities, including electrification, industrial growth and affordability, if power supply fails to keep pace with demand.

According to the report, data centres, where AI models are trained and deployed, accounted for about 1.5 per cent of global electricity consumption in 2024, equivalent to roughly 415 terawatt hours (TWh). That figure is expected to more than double to around 945 TWh by 2030, slightly exceeding Japan’s current total electricity consumption.

The United States, China and Europe are projected to drive most of the growth. In the US alone, the IEA said data centres will account for nearly half of electricity demand growth between now and 2030. By the end of the decade, the country is expected to consume more electricity for data centres than for the combined production of aluminium, steel, cement, chemicals and other energy-intensive goods.

The surge is being driven largely by AI-focused accelerated servers, which require vast amounts of power. While data centres still represent a relatively small share of global electricity demand, the agency warned that AI-focused facilities can consume as much power as aluminium smelters and other heavy industrial plants, while being far more geographically concentrated.

In the US, nearly half of data centre capacity is located in just five regional clusters, increasing local grid risks. The IEA reiterated that without mitigation, about one-fifth of planned projects could face delays.

To address these challenges, the agency called for careful planning, including siting new data centres in areas with strong grid capacity and operating servers, onsite power generation and storage more flexibly.

At the same time, the report highlighted AI’s potential role in strengthening energy systems. The IEA said AI is already being used by energy companies to optimise supply chains, electricity generation and transmission, and energy consumption, helping to reduce costs, improve reliability, extend asset lifetimes and cut emissions.

The agency noted that advanced digital tools could unlock up to 175 gigawatts of additional transmission capacity without building new power lines, reduce renewable energy curtailment and improve fault detection across grids.

However, the report also raised concerns about energy security and critical materials, noting that AI both intensifies and helps mitigate risks in cyber and physical domains. It highlighted, for example, that China currently accounts for about 99 per cent of global refined gallium supply, a material critical to advanced electronics.

The IEA called for closer cooperation between the technology and energy sectors, saying, “Delivering the energy for AI, and seizing the benefits of AI for energy, will require even deeper dialogue and collaboration.”

It added that while risks remain, the agency will continue to provide data and analysis to help policymakers and industries navigate the rapidly evolving relationship between AI and global energy systems.