Electricity use by data centers in the U.S. could nearly triple over the coming three years according to a new study by the Department of Energy, the latest indicator of a coming boom in power demand as tech companies add computing power to drive artificial intelligence.
The report released Friday from Lawrence Berkeley National Laboratory estimates that data center energy use tripled over the past decade and is expected to either double or triple again by 2028.
Data centers consumed about 4.4 percent of all the electricity produced in the country in 2023, the study found, and they are expected to consume approximately 6.7 to 12 percent of total electricity by 2028.
The DOE report is the latest projection showing a sharp increase in power demand by data centers. A September report by consulting giant McKinsey & Company found data center energy use could grow to consume 11 to 12 percent of U.S. power by 2030. In May, the Electric Power Research Institute projected data centers could grow to consume 9 percent of U.S. power by 2030. In all three reports, the authors identified the boom in AI as the main contributor to growing power demand.
EPRI Principal Technical Executive Tom Wilson said all such projections include a lot of uncertainty depending on new developments by tech companies and how much consumers ultimately use AI products.
"While projections are inherently uncertain, it's likely electricity needs will continue to increase, depending on regional needs," Wilson told Newsweek via email. "That's why it's key for data centers and power providers to communicate and collaborate to ensure reliable, affordable energy for all."
The spike in electricity demand presents a challenge to tech companies and power providers who are trying to meet climate goals to reduce dependence on fossil fuels and grow clean-energy sources.
Despite ambitious climate goals and heavy investment in renewable energy, tech companies including Google and Microsoft have seen their greenhouse gas emissions rise with the growth of data centers.
Last week, Newsweek reported on a study that showed how increasing data center energy use could cause more toxic air pollution. The study found that air pollution linked to data centers could contribute to as many as 1,300 premature deaths and billions of dollars in health costs in the U.S. each year by 2030 if tech companies don't use cleaner power.
The lead author of that study, University of California, Riverside, associate professor of electrical and computer engineering Shaolei Ren, said the DOE projects faster growth of data center demand than his study had estimated.
"If the DOE's projection holds true, the public health impact of U.S. data centers could even exceed our earlier estimate of $20 billion per year in 2030," Ren told Newsweek via email. "We must act swiftly to address this looming threat."
The DOE sought to highlight the opportunities for clean energy to power data centers.
"We can meet this growth with clean energy," Energy Secretary Jennifer Granholm said in a statement accompanying the Lawrence Berkeley study.
That's a point Granholm made in an interview with Newsweek in September. She said she was confident that new federal incentives for clean energy will speed growth of renewable and nuclear energy, especially when coupled with the purchasing power that tech companies have.
"We now have these amazing tools in the Inflation Reduction Act which make it irresistible to develop clean resources for the grid," Granholm told Newsweek. "This year we will add 60 gigawatts of clean power onto the grid, so we will be able to meet that demand."
A Newsweek Horizons event during Climate Week NYC in September took a deeper dive into the issue, as executives from utility and tech companies described how they plan to power the growth of AI.
"Our entire portfolio of AI has accelerated and increased the need for carbon-free energy, so that requires us to look outside the box," Microsoft Vice President of Energy Bobby Hollis said during the panel discussion. Among the energy options Microsoft is pursuing, the tech giant has partnered with electric power provider Constellation to reopen a nuclear reactor at the Three Mile Island generating station in Pennsylvania.
Another panelist at the event, Duke Energy Managing Director of ESG & Sustainability Heather Quinley, said that data centers will account for about 25 percent of new projects in its service areas in the Southeast and Midwest.
"We're seeing significant load growth from data centers and advanced manufacturing," Quinley said. While Duke has set a goal to decarbonize its power generation fleet, it has also been criticized by climate activists for plans to build additional gas-fired generation in North Carolina and a decision to delay the retirement of a coal-burning power plant in Indiana.
Major AI developers such as Google, Meta and Microsoft have all made major investments in renewable and nuclear energy. However, some are also looking to additional fossil fuel power sources. Meta, parent company of Facebook, announced this month that its largest new AI data center planned for construction in Louisiana will be powered with natural gas.
When Elon Musk's xAI opened its AI data center in Memphis, Tennessee, in July, the facility was powered by a fleet of mobile gas-burning generators, raising concerns about air pollution and greenhouse gas emissions.
"It seems really sketchy," LaTricea Adams, founder and president of the environmental justice organization Young, Gifted & Green, told Newsweek at the time. "It does appear that because this is Elon Musk, and this is a billionaire, that there was not transparency around what permitting took place."
English (US) ·