Understanding data centre electricity use in the AI era

Background Information

The IEA’s Digital Demand-Driven Electricity Networks (3DEN) Initiative, the IEA 4E TCP Efficient, Demand Flexible Networked Appliances (EDNA) and the Super-Efficient Equipment and Appliance deployment (SEAD) initiative are co-organising a webinar focused on the electricity use of data centres and the role of energy consumption data in enabling effective policy and planning. The session will begin with an overview of the latest IEA analysis on the links between artificial intelligence (AI) and electricity demand, followed by two technical presentations examining data availability and methodologies for assessing data centre energy use. These will include new research from EDNA on critical assessments of current energy modelling approaches and the development of a next-generation energy forecasting tool.

Webinar Focus

This webinar will explore three interrelated topics shaping the future of data centre energy use and digitalisation:

AI and electricity demand: scene setting from the IEA Energy and AI report
AI systems are rapidly being integrated into daily life and business operations, raising critical questions for energy systems. As AI adoption grows, so too does the electricity required to power data centres which are the physical backbone of AI. This session will present key findings from the IEA Energy and AI report that quantifies AI-related electricity demand and its potential growth, and identifies implications for energy security, emissions, and policy planning. The presentation will highlight why accurate and timely data is essential for understanding and responding to this emerging challenge.

Critical review of data centre energy estimates
There is a wide range of published estimates for current and future data centre electricity use, in some cases differing by a factor of 40. EDNA’s Data Centre Energy Critical Review presents a systematic analysis of over 100 studies and projections, identifying key drivers behind discrepancies in reported figures. It seeks to determine a plausible current range for global data centre electricity use, assess how past studies approached the task, and highlight best practices and pitfalls in data collection and modelling. This presentation will offer much-needed clarity for policymakers, researchers and system planners navigating a complex evidence base.

Total Energy Modelling for Data Centres
Complementing the critical review, EDNA will also present its Total Energy Model (TEM) 4.0, a novel bottom-up hybrid model designed to forecast energy consumption across different types of data centres and workloads, including AI. Rather than relying on conventional methods based on traffic or sales extrapolation, the TEM uses processor-based capacity and efficiency data, combined with region-specific deployment profiles. It segments demand across five global regions and distinguishes between traditional, cloud, hyperscale and AI-driven workloads, offering new insights into how digital infrastructure growth and policy interventions could shape electricity demand.

This session is designed for policymakers, energy analysts, digital infrastructure planners, and technology providers seeking to better understand and respond to the growing energy footprint of data centres in the AI era.

The IEA gratefully acknowledges the support of the Italian Ministry of Environment and Energy Security as part of their contributions to the IEA’s Digital Demand-Driven Electricity Networks (3DEN) Initiative on power system modernisation and the Clean Energy Transitions Programme.