Artificial Intelligences Energy Paradox 2025

Foreword In today’s economy, artificial intelligence (AI) systems offer both challenges and opportunities. As integral components of digital infrastructure, the data centres that enable AI support a variety of applications, from cloud computing to complex data processing. AI’s rapid expansion, however, is accompanied by growing electricity demand, with the largest facilities in the world using the same amount of power as small cities to ensure uninterrupted operation. Data centres come in varying sizes however, ranging from large, hyperscale facilities with more than 1 gigawatt (GW) of power capacity, to smaller, micro edge deployments that may draw less than 10 kilowatts (kW) of power.1 One estimate now expects data-centre-related electricity consumption to grow from approximately 1% of global electricity demand to over 2% by 2026, potentially reaching 3% by 2030 if forecasted growth continues.2 Such projections have raised concerns about supporting this demand while also meeting net-zero commitments. Simultaneously, AI can be a powerful tool to positively support wider energy system transformation. For example, it is already being used to improve energy efficiency across industries, accelerate renewable energy integration and make power grids more resilient. This is the AI energy paradox – balancing these challenges against AI-enabled opportunities. However, current estimates of AI’s energy impact vary, and the magnitude of electricity demand growth remains unclear. Other issues include a lack of standardized taxonomies and definitions. The extent to which electricity demand growth will be offset by efficiency gains – from advancements in technologies (e.g. chips, algorithms etc.), data centre design and changing regional dynamics – is also uncertain. While a near-term rise in AI’s electricity consumption is expected, the future magnitude of this growth may decline due to the achievement of efficiency gains. To achieve this, it’s pivotal to understand innovative mitigation strategies and solutions that can effectively facilitate this balance. Over the past year, the World Economic Forum’s AI Governance Alliance has united industry and government with civil society and academia, establishing a global multistakeholder effort to ensure AI serves the greater good while maintaining responsibility, inclusivity and accountability. Players from across the AI value chain are convened to cultivate meaningful dialogue on emerging AI issues. With Accenture as a knowledge partner, the alliance’s AI Energy Impact Community (composed of over 40 global members) has facilitated cross- industry discourse towards consensus and surfaced applied use cases on AI’s energy impact. This paper highlights cross-industry insights from a diverse stakeholder group to outline mitigation strategies: –Identifying electricity use reduction strategies for AI systems –Touching upon AI’s potential for the wider energy transition –Outlining key partnerships, frameworks and policies to support sustainable AI adoption The increase in AI adoption, alongside other market factors is contributing to increased electricity use. Annual global electricity demand growth is now forecasted to reach nearly 3.5% in the coming years.3,4 This challenge is amplified by global competition for AI projects across regions. This will require stakeholders across the value chain to navigate market pressures for computing power, while balancing sustainability targets, grid constraints and community impacts.Cathy Li Head, AI, Data and Metaverse; Deputy Head, Centre for the Fourth Industrial Revolution; Member, Executive Committee, World Economic ForumJames Mazurek Managing Director, US Utilities Strategy Lead, Accenture Artificial Intelligence’s Energy Paradox: Balancing Challenges and Opportunities January 2025Roberto Bocca Head, Centre for Energy and Materials; Member, Executive Committee, World Economic ForumJeremy Jurgens Managing Director, World Economic Forum Artificial Intelligence’s Energy Paradox: Balancing Challenges and Opportunities 4