AI Consumed 0.5% of Global Electricity in 2025, IEA Report Finds
Key Takeaways
- ▸AI consumed 0.5% of global electricity in 2025; all data centres consumed 1.5% of global supply
- ▸Data centre demand is severely concentrated geographically, with 5% of US electricity and 25%+ in states like Virginia
- ▸IEA projects data centre electricity will roughly double to 3% of global supply by 2030, driven primarily by AI
Summary
According to the International Energy Agency's latest Energy and AI report, artificial intelligence systems accounted for approximately 0.5% of global electricity consumption in 2025, with all data centres collectively consuming 1.5% of worldwide electricity. AI-focused data centres used roughly half of that total electricity, while non-AI data centres—supporting streaming, banking, email, and other digital services—consumed more than twice as much. However, the demand is highly concentrated geographically: data centres account for 5% of US electricity consumption and over 20% in Ireland, with some states like Virginia seeing data centres consume more than 25% of local electricity supply.
Looking ahead, the IEA projects data centre electricity demand will grow to 3% of global electricity by 2030 under its base scenario, with AI-focused facilities becoming the primary growth driver and eventually consuming roughly equal shares with non-AI data centres. The agency published four different demand scenarios to account for variables including rapid AI scaling, accelerated efficiency improvements, and lower-than-expected demand growth, reflecting significant uncertainty in medium-term projections.
- Non-AI data centres still use 2x more electricity than AI-focused ones, but AI is the growth engine
Editorial Opinion
This IEA data punctures inflated claims about AI's energy impact while highlighting a genuine emerging challenge: geographic concentration. Half a percent of global electricity is not a global crisis, yet it becomes one locally—Virginia's grid bears an outsized burden for the world's AI consumption. The real policy opportunity isn't global coordination but targeted investment in renewable energy and grid resilience in AI clusters. As AI demand accelerates, the question isn't whether we can afford it globally, but whether specific regions can afford it locally.


