How much energy do data centers and artificial intelligence use?
Key Takeaways
- ▸AI consumed ~0.5% of global electricity in 2025, equivalent to about one-third of all data center electricity demand
- ▸Data centers currently use 485 TWh annually (~1.5% of world electricity), with AI facilities projected to match non-AI data centers by 2030
- ▸Energy demand is dominated by inference (running models) rather than training; concerns span environmental impact, local grid strain, and potential resource bottlenecks for AI expansion
Summary
A comprehensive analysis examining AI and data center energy consumption finds that artificial intelligence consumed approximately 0.5% of global electricity in 2025, equivalent to roughly one-third of all data center electricity use worldwide. According to the International Energy Agency (IEA), data centers collectively consumed around 485 TWh annually (about 1.5% of global electricity generation), with AI-focused facilities accounting for the remaining third. The article breaks down energy usage for both model training and inference, clarifying that estimates include facility operations (cooling, lighting) but exclude end-user devices and cryptocurrency mining.
As AI adoption accelerates, energy demand projections suggest rapid growth. The IEA's base-case scenario estimates data centers will consume 3% of global electricity by 2030, with AI and non-AI facilities reaching parity. However, these projections carry significant uncertainty, with some analysts arguing the IEA's forecasts are conservative compared to industry growth trajectories.
- Projections are uncertain; some analysts believe AI energy demand growth could exceed even the IEA's base-case scenario
Editorial Opinion
This analysis underscores a critical tension in AI's future: rapid capability expansion is coupled with growing resource demands that will pressure grids and climate goals. While current AI energy use (0.5% globally) remains manageable, the trajectory toward parity with non-AI data centers by 2030 demands urgent investment in energy efficiency, renewable capacity, and smarter infrastructure. Without proactive solutions, AI's energy footprint could become a genuine bottleneck—not just for business growth, but for broader decarbonization efforts.
