As artificial intelligence (AI) transforms industries and economies, a new report highlights its growing dependence on the energy sector.
While AI promises unprecedented productivity and innovation, its energy demands could outpace expectations – unless governments, industry and civil society act fast.
Since ChatGPT’s public launch in late 2022, AI has surged to the top of the global agenda. But behind the scenes, AI’s rise relies heavily on large data centres – the physical backbone of machine learning.
These facilities are energy-intensive, consuming vast amounts of electricity to train and deploy complex models.
According to the report, “without energy – specifically electricity – there is no AI.” A single AI-dedicated data centre today uses as much electricity as 100,000 households.
Next-generation centres will need 20 times that. While data centres currently represent 1.5% of global electricity demand, that figure is expected to double by 2030 – the equivalent of Japan’s current consumption.
In the US, data centres could soon account for nearly half of total electricity demand growth this decade.
By 2030, “the US economy is set to consume more electricity for processing data than for manufacturing all energy-intensive goods – such as aluminium, steel, cement and chemicals – combined.”
The report warns that without careful planning, the clustering of data centres near cities could overwhelm local grids.
Coordinated action is essential to avoid power bottlenecks and ensure secure, affordable and sustainable energy delivery.
Beyond electricity use, the supply chains for critical minerals like copper and gallium also face strain. These are essential for wiring and advanced chips.
At the same time, AI’s dual role in enhancing and endangering energy sector cybersecurity is under increasing scrutiny.
A new Observatory on Energy and AI will soon launch to track electricity use and emerging applications. As AI evolves, balancing its promise with sustainable energy policy will be vital.
Copyright © 2025 Energy Live News LtdELN