The AI Power Grab That Could Upend America's Electricity Industry

Researchers at Ohio State University College of Food, Agricultural and Environmental Sciences project that the data centers of six major technology companies could use more than twice as much electricity worldwide by 2030 compared to 2024. According to Farm and Dairy, the study focused on Amazon, Microsoft, Google, Meta, Oracle, and Apple because of their significant data center operations and AI investments. The researchers project that the combined annual electricity consumption worldwide for these six companies could increase from about 118 terawatt-hours in 2024 to between 239 and 295 terawatt-hours by 2030. Nearly two-thirds of this projected electricity use would be concentrated in fewer than 10 states and countries, with Ohio, Oregon, Virginia, Iowa, and Ireland each projected to exceed 15 terawatt-hours annually under middle-range growth assumptions. Farm and Dairy reported that 15 terawatt-hours equals roughly one-tenth of the electricity sold to all Ohio customers in 2024, based on U.S. Energy Information Administration figures. Ani L. Katchova, a professor at Ohio State, noted that whether this growth brings broad local benefits depends on how infrastructure costs are allocated among households, farms, businesses, and data centers. If other electricity customers bear too much of the cost, households, farms, and businesses could face higher costs or greater price volatility. Co-author Yongyang Cai emphasized the need for anticipatory planning so utilities and system planners incorporate large data center loads into long-term demand forecasts. Separately, OilPrice.com contributor Leonard S. Hyman compared the financial scale of big technology companies to electric utilities, pointing out that Microsoft alone has a market capitalization of $3.8 trillion, while the entire U.S. investor-owned utility industry has a market capitalization of $1.2 trillion. Due to long utility interconnection timelines and turbine backorders lasting five to seven years for large power plants from manufacturers like GE Vernova, Mitsubishi Power, and Siemens, some data center developers are turning to truck-mounted aeroderivative combustion turbines, such as the GE LM2500, to secure rapid power. OilPrice.com noted that these simple cycle turbines have higher heat rates and lower thermal efficiency than large combined-cycle plants favoured by utilities, and they can be up and running in months. Meanwhile, Nick Baileys, director of energy infrastructure solutions at Delta Electronics (Americas), discussed in an EE Times interview at the OCP Global Summit that data center operators are evaluating microgrids, on-site energy resources, and energy storage systems to overcome grid constraints and manage rapid fluctuations in computing demand.