Study: AI, data centers using massive amounts of U.S. electricity
Data centers are fueling the growth of artificial intelligence (AI) operations and the 24/7 digital economy, but a study from the Electric Power Research Institute (EPRI) found that they could also devour electricity generation to make it happen – as much as 9 percent of U.S. generation by 2030.
That would be more than double what they currently consume, representing a major draw on the grid in general and regional supply in particular. EPRI made its estimates based on public information about existing data centers, estimates of industry growth and private electricity demand forecasts from industry experts. In the end, it laid out four scenarios, where U.S. data center electricity consumption between 2023 to 2030 would range between 4.6 percent to 9.1 percent.
“The U.S. electricity sector is working hard to meet the growing demands of data centers, transportation electrification, crypto-mining, and industrial onshoring, while balancing decarbonization efforts,” David Porter, EPRI vice president of electrification and sustainable energy strategy, said. “The data center boom requires closer collaboration between large data center owners and developers, utilities, government, and other stakeholders to ensure that we can power the needs of AI while maintaining reliable, affordable power to all customers.”
To put it in context: AI queries need approximately 10 times the electricity of traditional internet searches to perform their task. That figure jumps even higher when generating music, photos, videos and other original products. Today, EPRI stated that 5.3 billion people use the internet, so if AI is rapidly adopted, power demands will increase exponentially in a short period. The concentration of computing facilities isn’t helping, with single facilities now requesting power consumption that can run from the equivalent of 80,000 homes, to as many as 800,000 – nearly the population of San Francisco.
Annual growth rates of AI use could range from between 3.7 percent and 15 percent, by EPRI’s estimates. In that lowest case, the public would need to be more limited in AI adoption, and data centers would need to make significant gains in efficiency. Currently, 80 percent of the national data center load is also concentrated in just 15 states, with Texas and Virginia leading the pack.
In order to counter this change of pace, EPRI recommended that data centers immediately push for improved efficiency and flexibility, data center developers and utilities begin to coordinate more closely about their needs and constraints, and modeling tools be improved to anticipate and accommodate data center growth.
