AI and Data Centres Projected to Double Global Electricity Use by 2030, Says IEA

A new report by the International Energy Agency (IEA) warns that the electricity consumption of global data centres is expected to double within five years, largely due to the rising use of artificial intelligence (AI). By 2030, data centres could require 945 terawatt-hours (TWh) of electricity annually—three times more than the current usage of the entire United Kingdom. The sharp increase is concentrated around major technology and population hubs, posing challenges for power grids, energy providers, and environmental sustainability. The report highlights that training large language models like OpenAI’s GPT-4 consumed 42 gigawatt-hours (GWh) over 14 weeks, comparable to the daily energy needs of tens of thousands of households. Even basic AI tasks, such as generating short video clips, are highly energy-intensive. In the U.S., data centres are projected to consume more electricity than traditional heavy industries combined by 2030. Despite these concerns, the IEA emphasizes AI’s potential to enhance energy efficiency and speed up the adoption of cleaner power sources. However, geopolitical tensions and high tariffs on key materials—particularly those related to renewable technologies—may hinder energy infrastructure development, especially in the U.S., while potentially giving China an edge in clean energy and AI deployment.