The first data centres consumed as much power as 20,000 homes, but the largest today use the equivalent amount of power as 100,000 homes and are on track to burn up 5mn household’s worth of power by 2030 – the same amount of power that Japan consumes, the International Energy Agency (IEA) said in a report on May14.

The rapid expansion of artificial intelligence is set to significantly increase global electricity demand. While AI has evolved into a multi-trillion-dollar industry since the public launch of ChatGPT in late 2022, the IEA stressed that, ”without energy – specifically electricity – there is no AI.”

The energy requirements stem from vast data centres used to train and deploy AI models. Electricity used by data centres alone already consumes as much as that of Germany or France and will be comparable to that of India by 2030, according to the International Monetary Fund (IMF).

These facilities are already highly energy-intensive, and their consumption is accelerating. They will leapfrog over the projected consumption by electric vehicles, using 1.5 times as much power than EVs by the decade’s end. Although data centres currently account for just 1.5% of global electricity demand, their share is expected to double to 3% by 2030.

The impact is projected to be especially pronounced in the United States, where data centres are expected to account for nearly half of the country’s total increase in electricity demand through the end of the decade. Power needed for US server farms is likely to more than triple, exceeding 600 TWh by 2030, according to a medium-demand scenario projection by McKinsey & Co.

“Driven by AI use in particular, the US economy is set to consume more electricity in 2030 for processing data than for manufacturing all energy-intensive goods – such as aluminium, steel, cement and chemicals – combined,” the report said. Under current energy policies, the AI-driven rise in electricity demand could add 1.7 gigatonnes in global greenhouse gas (GHG) emissions between 2025 and 2030, about as much as Italy’s energy-related emissions over a five-year period.

The IEA also noted that data centres are often clustered near cities, unlike other high-demand infrastructure such as heavy industry. This creates new pressures on urban grids. “Without greater and more proactive coordination between government, the tech industry, the energy sector and civil society, there’s a risk this could lead to local grid strains and bottlenecks,” the report finds.

The world’s data centres consumed as much as 500 TWh of electricity in 2023, according to the most recent full-year estimate by the Organization of the Petroleum Exporting Countries (OPEC). That total, which was more than double the annual levels from 2015-19, could triple to 1,500 TWh by 2030, OPEC projects.

“AI has the potential to raise the average pace of annual global economic growth according to scenarios in our recent analysis, included in the IMF’s April 2025 World Economic Outlook,” the IMF said in a recent blog. “AI, however, needs more and more electricity for the data centres that make it possible. The resulting strain on power grids has major implications for global electricity demand.”

Beyond power supply, the report highlighted growing risks around the supply chains for critical minerals used in data centre hardware. These include copper for electrical wiring and gallium for high-performance chips. The complexity and geographic spread of these supply chains pose long-term strategic challenges.

Cybersecurity is also a growing concern. While AI offers tools to improve digital defences, it is simultaneously increasing the frequency and sophistication of cyberattacks targeting energy infrastructure. “This is an issue that requires greater attention,” the IEA concluded.