Rising energy prices put AI and data centers in the crosshairs (techcrunch.com)

A new Sunrun-commissioned survey finds 80% of consumers worry that the AI-driven build-out of data centers will raise their electricity bills — a concern grounded in recent trends. U.S. electricity demand was flat for a decade but has shifted in the past five years toward commercial (+2.6% annually) and industrial (+2.1%) users, while residential growth is only +0.7%. Data centers already use about 4% of U.S. electricity (more than double 2018 levels) and Lawrence Berkeley Lab projects that share could climb to 6.7–12% by 2028. Renewables (solar, wind, grid batteries) have met much of the new demand — solar projects can hook up in ~18 months — but political risks (possible Republican rollbacks of parts of the Inflation Reduction Act) and supply-chain constraints for gas turbines complicate capacity planning. For the AI/ML community this matters operationally and economically. Rising demand and constrained thermal generation (new gas plants take ~4 years and turbine backlogs push deliveries up to seven) raise the prospect of higher wholesale prices and strained grid capacity, which could increase cloud costs, slow new deployments, and trigger tighter siting or procurement rules. Technical responses likely to accelerate include aggressive model/compute efficiency, carbon- and price-aware scheduling, on-site renewables and storage, and long‑term power purchase agreements. The mix of consumer backlash and policy uncertainty could reshape where and how large-scale training and inference infrastructure is built and priced.