For years, the U.S. power grid has been a bastion of predictability and stability. U.S. electricity demand remained flat throughout the 2010s, as gains and losses in efficiency in energy-intensive industries like manufacturing masked the coming digital age.
But the grids of the past may not be able to meet the technological needs of the 2020s. Retail electricity prices have soared in recent years, rising faster than inflation over the same period, in part because of rising electricity costs due to an AI-driven infrastructure boom. Electricity costs have been one of the factors behind AI’s recent sharp decline in public opinion polls, and a new study suggests the pain for residential utilities related to this decade’s technology needs may be just beginning.
A study published by the magazine last week showed that data centers accounted for 1.9% of total U.S. electricity use between 2018 and 2023 to 4.4% Environmental Research Letters.
According to the study, which simulated several different energy usage scenarios based on existing electricity demand forecasts, national average wholesale electricity costs could rise between 6% and 29% by the end of the century. The increase in utility prices is mostly related to the expansion of data centers, with cryptocurrency mining also included in higher-cost models.
In some areas, price increases may be greater. For example, in Virginia, one of the epicenters of the country’s data center boom, power generation costs can soar as much as 57%.
urgent energy needs
Grid power flowing to data centers surged 22% last year and could account for 17% of total U.S. electricity use by the end of the decade, according to research from S&P Global.
To meet this demand, the study’s models predict utilities will rely heavily on natural gas — a fuel source whose price fluctuations add some uncertainty to future consumer costs.
Jeremiah Johnson, associate professor of civil and environmental engineering at North Carolina State University and lead author of the study, also found that data centers may turn in part to underutilized coal-fired power plants to meet their energy needs. Research shows that data center expansion could actually increase carbon dioxide emissions from electricity generation by up to 28% by 2030, reversing some of the power industry’s efforts to phase out coal over the past two decades.
Renewable energy will also play an important role in meeting this demand, although the ability of wind and solar to compensate has become heavily dependent on policy.
The study simulated scenarios with and without federal clean energy incentives that were comparable to those under the Inflation Reduction Act, the subsidy that Congress largely repealed earlier this year. Without these incentives, about 70% of the additional power needed for new data centers would be provided by natural gas, with coal, wind and solar providing the remainder. If these incentives were restored, natural gas’s share would fall to about 41%, wind would account for 29% of incremental load, and solar would account for 15% of incremental load.
The energy mix is equally important for costs and emissions. The study found that in areas where renewable energy development is slow or restricted, such as Virginia, where traditional fossil fuel power plants stay online longer, consumers may have to import power from neighboring countries, driving up wholesale costs for everyone on the grid.
“The challenge here is the scale of this demand we’re talking about is very large. The scale dwarfs some of the other changes we’ve experienced in the power industry in recent years,” Johnson told wealth.
“It’s going to take all hands on deck to get the generation needed to meet such a huge demand.”
Not in my backyard
With electricity prices expected to soar, the financial anxieties of American households have begun to register in public opinion.
Utilities are asking states to approve a record $31 billion in rate hikes nationwide in 2025. While electricity prices were rising long before the current data center boom (due in part to investments in modernizing grid infrastructure and improving weather resilience), AI and related infrastructure development have become obvious scapegoats.
Seven in 10 Americans oppose the idea of building artificial intelligence data centers near their homes, according to a Gallup poll released last week. The main concern is how construction will affect local resources, including electricity use. Fifteen percent of respondents specifically cited concerns about rising utility and energy costs.
This result is part of a worsening of views on AI that has also been shown in YouGov and other recent polls. this economist The study found that more than half of Americans say artificial intelligence is developing too fast and that the technology is unlikely to deliver significant widespread economic benefits.
This resistance has manifested itself in the form of protests and blockades of data centers in a growing number of communities across the country. Last year alone, opposition delayed or halted planned construction of 48 data center projects worth more than $156 billion, according to research firm Data Center Watch.
“There’s a lot of local resistance to data center siting, and what we’re finding is that being close to these large centers increases local electricity bills, which I think will make the siting process more contentious and more important,” Johnson said. “I think a very important aspect of site selection is understanding who pays the incremental costs associated with generating electricity and who bears the benefits.”
This story originally appeared on Fortune.com