A new report by the American Council for an Energy-Efficient Economy (ACEEE) has laid out the challenges posed by a growing industry of data centers. These facilities, sprouting across the nation, are consuming massive amounts of electricity and water, often with little oversight from regulators. The data center industry is being driven by the increasing computational demands of artificial intelligence, growing at a faster rate than clean energy production, putting regional sustainability plans and climate goals in jeopardy.
Many local leaders “just want to get data centers into the community. But the scale and the speed was not expected,” said Nora Wang Esram, one of the authors of the October report, Turning Data Centers into Grid and Regional Assets: Considerations and Recommendations for the Federal Government, State Policymakers, and Utility Regulators.
From 2010 to 2023, electric demand in the United States was nearly flat. Starting about two years ago, the nation began to see a resurgence in manufacturing, driven in part by federal legislation like the Inflation Reduction Act and the CHIPS and Science Act. AI data centers, new manufacturing facilities and the growth of electric vehicles are all expected to continue to grow electric demand, requiring the nation’s electric grid to expand 33 percent to 67 percent in the next decade, according to the report.
“We are seeing an increased interest in data centers as our collective appetitive for data, games, crypto and AI drives a seemingly endless need for computing power and data storage,” said Lori Mitchell, director of San José Clean Energy, an electric utility serving city residents. These facilities can offer benefits to cities in the form of revenue generated from property tax expansion or other sources, she added, but still require “careful consideration” concerning land use, water consumption and energy resources.
Generative AI computing requires about four times more power than the standard cloud-computing data center. And it’s not just electricity data centers are consuming. In 2018 — before the release of tools like ChatGPT — data centers in the U.S. consumed about 135 billion gallons of water, used largely for cooling, the study found.
“Those water uses may not be metered and measured in a way that the city can foresee the impacts,” Esram said. “By the time they realize, it’s a little bit too late. And in many cases, they’re reacting to it.”
The issue of water use by data centers is still a new one, said Tony Willardson, executive director of the Western States Water Council (WSWC), and is being closely watched. The WSWC, made up of water authorities and public officials from 18 states, passed a resolution in March calling for "integrating water and energy policies and engaging water and energy planners to maximize program and project effectiveness and efficiencies.”
Willardson recalled a recent meeting of commercial real estate developers related to water and electricity supplies in the area covered by the Intermountain Power Agency, an electric utility in Utah.
“Data centers’ power consumption was raised by the audience, which wasn’t satisfied with the answers from energy providers that the power would be there when needed,” Willardson said in an email. “The water requirements for cooling both thermal electric plants and data centers would not be insignificant and are likely to become increasingly important in the future.”
In San José, South Bay Water Recycling serves 19 data centers that use about 1 million gallons of recycled water every day, or 12 percent of the total recycled water for approximately 1,000 customers per day, Mitchell said, adding data centers are helping to “catalyze other developments” like new, more sustainable energy systems.
The growth of data centers and their significant electric demands, could jeopardize climate and sustainably goals established by a number of regions, as utilities look to more conventional methods of generating electricity, like natural gas-powered peaker plants — which typically operate during times of high, or peak, electrical demand.
Complicating the issue is the lack of regulatory structure around an industry that touches multiple agencies, including city and county planning and economic development offices, utilities and water authorities.
The issue, with its multiple players, has become more of a “who cares about what” question, Esram said.
“The complexity here is when cities and counties are looking at data centers, they are more thinking about potentially the noise, as well as the water use. That kind of belongs to the local jurisdiction,” she said. “When you think about the grid, that is really more at the state level. The utility, the distribution network, the retail market, is being regulated by the state-level commissioners.”
Data center operations can be adjusted to energy demands, she said, and the report points out some of AI’s most energy-intense processing can be shifted to times of low energy demand.
“Training AI” requires more computation, which means more power usage, she explained. “Will you train the AI when the [electric] grid is congested? Probably not.” The data center industry is still so new, it’s hard to say how much of an impact AI will place on a region’s electric grid, she added.
One thing’s for sure, say industry watchers: Data centers are inseparable from a modern, growing economy.
“Data centers are absolutely crucial for the operation of online services. And they’re providing, literally … some aspect of the physical infrastructure that’s powering the digital world we all continue to live in,” Melissa Farney, director of marketing at TECfusions, said during a recent webinar on the growth of data centers. The event was part of the weekly Broadband Breakfast podcast.
“Digital infrastructure is the foundation for future economies,” Bill Thomson, vice president of marketing and product development at DC BLOX, a multitenant data center, said during the podcast. “The smart cities and the smart counties of the world will see the future economic benefit of this investment.”
Editor’s note: This article has been updated to include perspective on data centers in San Jose, Calif.
