Rising Electricity Costs and the Impact on American Households

Electricity costs are climbing in some areas faster than the rate of inflation, and there is plenty of blame to go around. Expensive generation sources, rising fuel costs, expanding data centers and the burgeoning impact of artificial intelligence are commonly blamed as drivers, leaving utilities, regulators and legislators playing catch up as costs hit poorer residential consumers hardest.

At a time where seemingly everything is more expensive, electricity is no exception. The U.S. Bureau of Labor Statistics released a Consumer Price Index Summary in mid-August finding that all items indexed increased by 2.7% over the past 12 months before adjustments. Electricity rose by 5.5% for the yearly period ending in July, and natural gas jumped 13.8% over the same period.

The average kilowatt of electricity used by retail residential customers rose to 17.47 cents from 16.41 cents between May 2024 and May 2025, according to the Department of Energy’s Energy Information Administration (EIA).

Rising energy prices affect lower-income households the most, with the average energy burden for these households hanging at around 8% of income, about three times the rate for other households, according to the National Energy Assistance Directors Association (NEADA).

Professional services firm WSP found that more than four in five Americans (83%) are concerned about how much electricity costs. Even in this environment, however, 65% would be willing to pay more for electricity if that money went to building a more reliable power grid.

The same survey found that an overwhelming 92% of Americans would support new transmission lines in their communities if it helped meet growing demand, even if the projects did not benefit them directly.

Tilting at Windmills

President Trump has blamed adoption of renewable energy for higher electricity bills, which he campaigned on bringing down.

“Any State that has built and relied on WINDMILLS and SOLAR for power are seeing RECORD BREAKING INCREASES IN ELECTRICITY AND ENERGY COSTS,” Trump wrote in an August 20, 2025 post to Truth Social.

Trump also halted federal wind energy permits, ordered departments to subject clean energy projects to further scrutiny, ended tax credits supporting renewable energy, and raised tariffs on steel and aluminum, which could impact construction costs for solar and wind farms.

An accelerated switch to a lower-carbon power grid that integrates more renewable energy can make electricity more affordable and lower the overall cost of living, according to analysis by the Paris-based International Energy Agency (IEA).

“In many cases, clean energy technologies are already more cost competitive over their lifespans than those reliant on conventional fuels like coal, natural gas and oil. Solar PV and wind are the cheapest options for new generation,” according to a May 2024 report from the IEA.

However, turning this potential into reality depends on upfront investment, particularly in developing countries, according to the same report.

“The data makes it clear that the quicker you move on clean energy transitions, the more cost effective it is for governments, businesses and households,” said IEA Executive Director Fatih Birol. “If policy makers and industry leaders put off action and spending today, we will all end up paying more tomorrow.”

Data Center Impacts

In hotspots in the South, the Washington, D.C. area, New England, South Dakota and Idaho, companies and their investors are pouring cash into data centers, which they are hoping will power artificial intelligence applications even as they cause local electricity bills to spike.

Data centers currently require 4% of total U.S. electricity, but thanks to the impact of AI, currently small but projected to snowball, this amount could increase just over 9%, according to a May 2024 report from the Electric Power Research Institute.

An August 2025 report from EPRI and Epoch AI describes AI’s growing power needs as “exponential,” adding that just training AI models currently eats up some 100-150 MW per training run, and could consume 1-2 GW each by 2028.

Total U.S. AI power capacity is estimated at 5 GW today and could reach more than 50 GW by 2030, according to EPRI. AI is the primary driver of data center electricity hunger, with hyperscaler companies planning $370 billion in capital expenditures this year.

Forecasts are that AI could consume 5% of U.S. power generation capacity by 2030, according to EPRI.

“Although materializing more slowly, power demands from electrification of the economy ultimately could be much larger; however, electrification creates different electric system development needs,” according to the EPRI white paper.

Tom Wilson, principal technical executive at EPRI, said data center power demands grew slowly from 2007 to 2017 even as they processed much more data. This was mostly because of small enterprise data centers moving to larger, more efficient cloud facilities that could cool themselves more effectively and use IT hardware more efficiently.

“Since 2017, efficiency has continued to improve, but societal demand for data center services (e.g., video streaming and conferencing, navigation, commerce) has increased faster, leading to a doubling of data center power demand between 2017 and 2023,” Wilson said. “That demand is projected to more than double again by 2030 despite efficiency gains. Power costs are typically less than 10-20% of data center costs. As power becomes a more significant cost or a constraint on deployment, there will be increased incentive to innovate.”

How Utilities are Coping

To use a localized example of how demand is growing, Memphis Light, Gas & Water saw 150 MW in demand to accommodate the training runs conducted by xAI’s Grok training clusters, which were carried out in a former Electrolux factory in Memphis. This represented 5% of the utility’s peak power demand, according to MLGW.

Utilities are looking for ways to manage this demand in different ways. American Electric Power (AEP) unit Indiana Michigan Power (I&M), for example, signed a contract with Google for a demand response plan meant to balance Google’s operational needs and I&M’s reliability and affordability requirements. In the plan, which is subject to the approval of regulators, Google would curtail its electricity use during periods of high demand to reduce peak load.

“As we add new large loads to our system, it is critical that we partner with our customers to effectively manage the generation and transmission resources necessary to serve them. Google’s ability to leverage load flexibility will be a highly valuable tool to meet their future energy needs,” said Steve Baker, president and chief operating officer of I&M.

In a first-of-its-kind deal, Meta Platforms agreed with Constellation Energy to keep a nuclear power plant open to help serve its power needs. The 20-year power purchase agreement would enable Constellation to relicense and retain operations at its Clinton nuclear plant for another two decades, while delivering 1121 MW of nuclear power to Meta, parent company of Facebook and Instagram.

Could AI Hit the Wall?

Improvements in hardware efficiency can cut AI’s power demand, and researchers expect these improvements to continue to net savings in electricity consumption, but the overarching expansion of AI activity means overall demand is likely to continue — perhaps even resulting in the industry hitting a predicted “wall” imposed by power generation limits.

“The 4x/year growth is very rapid and must slow eventually, almost certainly by the 2030s, absent a drastically transformed economy. The key uncertainty is when, not if, this growth will slow down,” according to the EPRI report on the forecast data center power demand from AI training runs.

It is unclear as of yet whether the current capacity of the power grid can meet the demand from AI. Generation capacity and lack of transmission lines could impact this growth. The AI industry’s future growth will depend on the power grid and how much stakeholders are willing to invest in it.

“If AI demand hits supply constraints for power, policies to unlock energy growth may be needed to enable continued growth, which may disrupt traditional planning processes and have environmental consequences. Power growth also poses challenges for tech companies that have already committed to using clean energy,” EPRI’s report states in its conclusion.

Wilson said power plants and power lines take years to build, which represents a challenge to meeting this demand.

“While there is not a “wall” at a certain size, the challenges of large infrastructure projects are clear. One opportunity that the paper discusses is spreading training geographically. This strategy has been demonstrated but likely increases training time and cost.  Distributed training could be valuable if there is plenty of power in a region, but it cannot be quickly delivered to a single data center site,” Wilson said.

Wilson said one of the key strategies that AI companies, utilities and other stakeholders are advancing is flexible data center operation, transitioning some data centers from being passive loads — unaware of what is happening on the grid —  to ones that are aware of grid stresses and are able to respond when needed.

“Historically, data centers have been viewed as running all the time, which means that when a data center requests a connection, the grid has to plan to serve increased loads in all hours. Flexibility operation can be achieved through varying computational demands at some sites, by using onsite or nearby generation or storage assets to reduce the data center’s impact on the grid, or both,” Wilson said.

Flexible data center operation could speed connection, decrease or delay needed grid buildout, and can better utilize existing grid assets, lowering grid costs and improving reliability.

Bottom of the Heap

Worsening the picture for poor families is the pressure the recent federal budget is putting on the Low Income Home Energy Assistance Plan (LIHEAP), which is the primary way for low-income people to get help paying for their electricity bills.

LIHEAP is currently in crisis, Mark Wolfe, Executive Director of NEADA, which represents those who administer LIHEAP at the state level, testified in June 2025 to the Senate Subcommittee on Labor, Health and Human Services and Education and Related Agencies.

In April, the Trump Administration eliminated the Division of Energy Assistance—the office within the U.S. Department of Health and Human Services (HHS) that oversees LIHEAP—and fired the entire staff. HHS has not released a plan for administering the program without DEA staff.

“This gutting of federal support could not have come at a worse time for the households served by LIHEAP. One-out-of-six families are currently behind on their home energy bills and the total amount these families owe their utilities is approximately $21 billion, the highest level since 2021 and up by about 30 percent since the end of 2023,” Wolfe testified, adding more than 37% of families making less than $50,000 a year reported being unable to pay an energy bill at least once in the past year.

The only way for LIHEAP to keep up without cutting supports is to get additional funds, Wolfe said.

“In order to keep up with rising energy costs, rising temperatures, and the increase in extreme weather events, LIHEAP needs $6 billion in FY26 funding plus $1 billion for the program’s contingency fund, for a total of $7 billion,” he said, urging a rejection of Trump’s recommendation to cut LIHEAP funds.

The budget proposal contains inaccuracies and a lack of understanding in how LIHEAP works and how electricity bills are paid, he said. The budget cites a 15-year-old Government Accountability Office report identifying potential cases of fraud, but the GAO since confirmed actions were taken to reduce fraud.

The administration also called LIHEAP unnecessary due to some states and utilities having policies against cutting off power to low-income families, he said.

“This statement demonstrates a gross misunderstanding of who LIHEAP helps and how it helps them,” he said. “Seasonal shut-off moratoriums that many states have put in place to protect families during periods of extreme temperatures only delay bill payment temporarily. After the moratorium ends, the bill is due. Families that could not afford to pay their bill during the shut-off moratorium will likely not be able to pay off the amount owed when the moratorium ends.”

Without LIHEAP’s ability to help pay bills, even more families could face the prospect of having their electricity shut off.