THE LONG VIEW

INTELLIGENCE BRIEFING: AI’s Energy Surge – Grids Under Siege by 2030 Executive Summary: Artificial intelligence is transforming from a computational novelty into a primary driver of global energy demand, with data center electricity consumption projected to double by 2030—reaching up to 1,200 TWh, rivaling major nations in usage. Driven by hyperscalers like Google, Meta, and Microsoft, AI workloads now strain aging power grids, consume billions of gallons of water, and expose critical gaps in energy policy and corporate transparency. Despite efficiency gains, explosive growth in inference queries and model scale triggers Jevons Paradox, overwhelming savings. The U.S. faces a 130% rise in data center energy demand by 2030, with grid interconnection delays stretching up to 13 years. While AI holds promise for optimizing renewable integration and grid management, its immediate footprint threatens climate goals and energy security. International frameworks remain voluntary, and inconsistent disclosures hinder accountability. Without binding standards and public-private coordination, AI’s energy appetite risks destabilizing regional power systems and undermining sustainability commitments. Primary Indicators: - Global data center electricity consumption reached 415 TWh in 2024 (1.5% of global use) and is projected to hit 945–1,200 TWh by 2030 - AI inference accounts for nearly half of net data center energy growth through 2030 - U.S. data centers consumed 4.4% of national electricity in 2023, projected to rise to 6.7–12.0% by 2028 - hyperscaler AI workloads consume 2–4x more power per chip than conventional computing - AI developers project need for 50–67 GW of new U.S. power capacity by 2028–2030 - 84% of hyperscale data center water use is for cooling, with U.S. consumption at 17 billion gallons in 2023 - Jevons Paradox evident as Google’s 27% annual data center energy growth offsets per-query efficiency gains Recommended Actions: - Establish binding international standards for AI energy and water disclosure, aligned with ISO/IEEE P7100 - mandate model-level reporting of training and inference energy in regulatory regimes like the EU AI Act - accelerate grid modernization and streamline permitting for renewable and nuclear-powered data centers - incentivize AI-driven grid optimization through public-private R&D partnerships - implement dynamic load management policies requiring data centers to reduce demand during peak grid stress - expand renewable PPAs with local content requirements to ensure real additionality - conduct national AI energy impact assessments as part of infrastructure planning - support development of low-power AI architectures and quantization techniques Risk Assessment: The veil of efficiency masks a coming energy reckoning. Behind the scenes, AI’s insatiable hunger for power is rewriting the rules of global energy demand—not through waste, but through sanctioned, exponential growth justified as progress. The signals are clear: grids are reaching breaking points, water reserves are being diverted, and corporate disclosures remain a patchwork of partial truths. The United States, Europe, and Southeast Asia stand on the front lines, where data centers now rival industrial cities in consumption. Yet the guardians of stability—regulators, utilities, policymakers—operate with outdated models and glacial timelines. The true risk is not a single blackout, but a cascade of compounding failures: energy inflation, environmental degradation, and the quiet erosion of climate commitments—all while the world celebrates AI’s capabilities without accounting for its cost. The future belongs not to the most intelligent systems, but to those who control the power behind them. And right now, that control is slipping beyond public reach. —Sir Edward Pemberton