US Hyperscalers To Consume 22% More Grid Power By End of 2025
Article Date: 17 October 2025
Article URL: https://hardware.slashdot.org/story/25/10/17/2051238/us-hyperscalers-to-consume-22-more-grid-power-by-end-of-2025
Article Image: 
Summary
Research from 451 Research (part of S&P Global) finds that US hyperscale datacentres will draw around 22% more grid power by the end of 2025 versus a year earlier, driven mainly by demand to build and train new machine-learning models on GPU-heavy servers. The report estimates utility power to hyperscale datacentres will increase by about 11.3 GW to 61.8 GW this year, then rise to 75.8 GW in 2026 and to 134.4 GW by 2030 (figures exclude enterprise-owned facilities and focus on hyperscalers, leased sites and crypto-mining).
The research highlights Virginia and Texas as the largest current concentrations of datacentre power demand — Virginia’s load is forecast to reach about 12.1 GW in 2025, and Texas about 9.7 GW — while operators also scout locations such as Idaho, Louisiana, Oklahoma and smaller West Texas towns for cheap or “stranded” power and alternative generation options.
Key Points
- US hyperscale datacentre grid demand is forecast to be 22% higher at the end of 2025 than a year earlier (451 Research).
- Projected hyperscale demand climbs from ~61.8 GW in 2025 to 75.8 GW in 2026, 108 GW in 2028 and 134.4 GW by 2030 (excluding enterprise-owned sites).
- Growth is largely fuelled by training and developing new machine-learning models requiring GPU-dense servers and heavy cooling infrastructure.
- Virginia and Texas are currently the biggest hotspots for datacentre power demand; other states like Idaho, Louisiana and Oklahoma are being considered for cheaper or stranded power.
- Figures cover hyperscalers (Amazon, Apple, Google, Meta, Microsoft), leased facilities and crypto-mining sites, not enterprise-owned datacentres.
Context and Relevance
The rapid expansion of AI workloads is changing how and where operators build datacentres, with significant implications for local grids, planning and energy policy. Large, concentrated power draws can force utilities and regulators to invest in new generation and transmission capacity, potentially affecting consumer bills and regional energy strategies. For planners, investors and policymakers, the report signals where infrastructure bottlenecks and investment needs are most likely to arise over the next five years.
Why should I read this
Short version: if you care about energy prices, local grid reliability or where Big Tech plants its next massive facility, this is worth two minutes. The AI boom isn’t just about clever models — it’s reshaping demand for electricity, driving new builds and making some states the next battleground for power and policy.
Author’s note
Punchy take: this is a big deal. The numbers point to rapid, concentrated demand growth that will force choices — build more generation, upgrade networks, or shift where datacentres go. All of which matters to taxpayers, regulators and anyone paying an electricity bill.