Data centers consume roughly 1.5% of global electricity, a figure heavily concentrated geographically. Artificial intelligence (AI) is rapidly adopted, raising concerns about its energy demand's environmental impact, strain on local supplies, and potential as a technological bottleneck.

AI energy consumption includes training and running models, with inference likely dominating. Estimates encompass server power, cooling, and infrastructure, excluding user device power or cryptocurrency mining. The International Energy Agency (IEA) estimates data centers used 485 terawatt-hours (TWh) globally in 2025, about 1.5% of world electricity. AI-focused data centers accounted for a third of this, approximately 0.5% of global electricity. Projections indicate AI's share will grow significantly by 2030.

Estimates for data center electricity demand vary, with sources like S&P Global reporting higher figures, partly due to including cryptocurrency mining, which the IEA excludes. Geographic concentration is a key challenge; for instance, data centers account for over 20% of electricity consumption in Ireland. Individual AI queries, such as a text-based query to Google's Gemini model, consume an estimated 0.24 watt-hours (Wh) of electricity.