Energy and AI is a flagship report published by the International Energy Agency (IEA) on April 10, 2025, under a CC BY 4.0 license. It examines the bidirectional relationship between artificial intelligence and energy: AI's projected electricity demand through 2035, and AI's potential to optimize energy systems.
Publisher: International Energy Agency Published: April 10, 2025 License: CC BY 4.0
Summary of argument
The report frames its subject as bidirectional: AI drives new electricity demand, and AI may also transform energy operations through optimization. It opens from the premise that "there is no AI without energy — specifically electricity for data centres," while treating the demand side and the optimization side as parallel concerns rather than dismissing the latter. The report describes the argument that AI can improve energy efficiency — the position that AI will help solve the energy problem it creates — as a hypothesis worth taking seriously alongside the demand-side concern, not one to be set aside.
Key claims and structure
The analysis is organized around five impact areas: energy security (implications of concentrated AI infrastructure demand), the emissions profile (climate outcomes from energy consumption and efficiency gains), grid integration (infrastructure challenges for surging demand), innovation potential (AI for grid management, efficiency, and materials science), and affordability (cost implications across sectors).
The report runs to seven chapters covering demand analysis, supply solutions, and sectoral applications across power, transport, heating, industry, and buildings.
Provenance and positioning
The report is described as the first authoritative global treatment of the AI-energy nexus from a neutral multilateral body. Its perspective is distinct from several other sources on AI's energy footprint: US and US-industry framing (Abundant Intelligence (Altman), Bain & Company — 6th Annual Global Technology Report (2025), (Source: Raw Sources/SemiAnalysis - Stargate Data Center Layer Cake.md)); US-specific grid framing (NERC 2025 Long-Term Reliability Assessment (LTRA)); and academic or compute-research framing (Epoch AI — How Much Power Will Frontier AI Training Demand in 2030?, RAND — AI's Power Requirements Under Exponential Growth (2025)).
Relationships
- supports: AI Environmental Impact, Data Center Siting / AI Power Politics, Compute Governance.
- related: Epoch AI — How Much Power Will Frontier AI Training Demand in 2030?, Epoch AI — Can AI Scaling Continue Through 2030?, RAND — AI's Power Requirements Under Exponential Growth (2025), NERC 2025 Long-Term Reliability Assessment (LTRA), Bain & Company — 6th Annual Global Technology Report (2025), GAO-25-107172 — Generative AI's Environmental and Human Effects.
Sources
- Primary: IEA - Energy and AI Report 2025