Are AI data centers driving up your electric bill? A USAFacts deep dive.
Grab your favorite beverage and comfiest chair, because we’re going to learn some things.
It’s 3 am. What is the average American curious about? According to Google search data, an increasing number are spending their late nights thinking about the price of electricity.
Searches about electricity prices have exploded, with interest levels reaching about eight times higher in March of this year than average historical interest. As of June, interest remains at similarly elevated levels. (The increase started before the 2026 Iran War.)
Google searches for "electricity prices" hit a record high in March 2026, topping the pre-AI peak by more than 3x.
Google searches for "electricity prices," where peak interest in the search term equals 100
More than 1 in 10 searches about electricity prices also mention artificial intelligence (AI), which is driving a data center construction boom. Moreover, searches about the effects of AI on electricity prices have at times been most common – as a percent of total searches – between midnight and 3 am.
You may have seen a new data center pop up near your town. (According to Pew Research, about 38% of Americans now live within five miles of at least one data center.)
Get weekly insights
Subscribe for data-driven insights. No spin, just the facts.
And your electricity bills might have been noticeably higher than in the past. It is natural to wonder if the data centers and the bills are related and to wonder about what the future might hold.
In this report, we’re going to explore many questions you might have on this topic – and some questions you might not have thought to ask. We’ll look at three things:
- What’s going on with electricity prices?
- What, if anything, does AI have to do with it?
- How much growth is there going to be in AI demand? And how will this affect electricity prices in the future?
We’re not going to pretend to have all the answers. In fact, we will often point out areas where we think great caution is warranted in interpreting the data. One thing we learned in this research is how little consensus there is among experts on this issue; there is not even an agreed-upon number for just how much total electricity is used by data centers in the United States right now. Our goal is simple: to give you some more understanding of the available facts and data around this issue.
Am I paying an enormous amount more for electricity because of AI?
For most Americans, probably not – at least not yet. But the answer depends on where you live and could change in the next several years. There are four important facts that you need to know.
- Average electricity cost increases nationwide have followed general inflation patterns. While average electricity prices in the residential sector rose 27% from 2021 to 2025, most of this increase can be explained by inflation. The increase was a noticeably smaller 7% after correcting for inflation. This national average, however, obscures important state-level variations.
- Big state-level rate increases are more common than you might think. While prices have risen considerably in some states in recent years, these increases are not historically all that unusual. It is fairly common for prices to rise substantially in individual states in a short period of time. Historically, there has been roughly a 1 in 20 chance that, in the next three years, a state’s residential electricity prices will rise by more than 20% on an inflation-adjusted basis. In other words, large rate increases over a short period of time can arise from many forces in the energy market and do not, by themselves, imply a new cause such as AI data centers.
- Where data centers are dense, prices haven’t risen faster. There isn’t a clear relationship between the amount of data center capacity in a state and changes in residential electricity prices in that state over the past few years. Quite simply, some of the states that had the largest increases in electricity prices in recent years are also those with little data center presence. (That said, electricity markets often span multiple states, so data centers can affect electricity prices beyond the state in which they are located. We discuss these regional spillovers in more detail in the deep dive that follows.)
- Other factors besides AI may explain rising electricity prices. A comprehensive government-funded analysis of rate increases from 2019 to 2025 found that replacing aging infrastructure, fluctuations in natural gas prices, recovery from natural disasters, and state-level energy policies were the primary drivers in rising residential electricity prices.
Together, these facts suggest that AI data centers are unlikely to have caused large increases in residential electricity prices at the national level and for most states, at least through 2025. That said, there is some evidence from both academic and industry research of increases in residential prices caused by data centers in some localities. In addition, this is a fast-changing area, with the evidence continuing to evolve as new data emerge.
It is certainly possible that data centers could cause large increases in residential electricity prices in the future. One indication comes from the latest capacity auction run by PJM Interconnection, a regional grid operator for 13 states and Washington, DC. The auction, which secures commitments from power plants to be available from mid-2028 through mid-2029, cleared at $16.4 billion, matching the record set in late 2025. According to the grid’s independent market monitor, data centers accounted for roughly $6.3 billion (or 38%) of the total. The extent to which residential customers will bear those costs is not yet known.
There is tremendous uncertainty about the future of AI, how much electricity data centers might consume, and what regulations might be passed affecting residential electricity prices.
We recommend great caution in reading any projection that claims certainty on the future of AI and electricity prices.
You can stop reading now (unless we’ve piqued your curiosity and you want to know more).
…
…
Still here? Okay, wonderful.
Let’s dive in.
There is tremendous uncertainty about the future of AI, how much electricity data centers might consume, and what regulations might be passed affecting residential electricity prices.
In theory, how could AI data centers influence your electric bills?
While there is fierce debate on the effects that AI will have on the world — and whether these effects will be good or bad — there is one fact that is undeniable: the data centers that power AI require a lot of electricity.
New data centers built with capacity from 100 to 1,000 megawatts of power could consume as much electricity as roughly 80,000 to 800,000 American homes over a year. For a state like Virginia, which now has data centers that can draw 9,430 megawatts of power, it is the electricity equivalent of adding at least 7.5 million homes. To give you perspective on that number, Virginia had about 3.4 million households according to recent data.
It would seem inevitable that adding so much electricity demand would increase prices. Basic economics tells us that a large surge in demand should lead to higher prices. There is only so much electricity to go around; if data centers demand a lot of it, one might it is natural to expect that other consumers would have to pay a higher price for the limited amount of it, at least until supply can catch up.
In addition, if we add a lot of electricity demand, the utility will need to procure and deliver it while still maintaining reliability. This could require building new generators or infrastructure, the costs of which may be passed on to consumers.
But some unique factors about the electricity market make the story more complex. The electricity market has what economists call high fixed costs and low marginal costs. Building the plants and other infrastructure that powers the electric grid is expensive. However, once that infrastructure is in place, supplying additional electricity can be relatively cheap. This is particularly true if the electricity is wind or solar.
To recoup the costs of building all the infrastructure, power companies split the costs among their customers. Because many grid costs are fixed, adding a large new customer can spread those costs over more electricity sales, potentially reducing the amount each customer must pay.
The electricity market also differs from many other markets because different classes of customers can be charged different prices.
In some states, regulators set electricity prices. Regulators may attempt to ensure that data centers bear more of the costs they impose on the electric grid, rather than passing those costs on to residential customers. An increasing number of utilities are creating a separate class for data centers, with potentially different rates than other customers.
In addition, some data center developers have paid for grid upgrades themselves. Recently, some builders of AI, power operators, and state governors have signed a pledge to foot the bill for modern data centers.
How much have residential electricity prices actually gone up in recent years?
You may have seen a chart showing a large increase in electricity prices paid in the residential sector (including households like yours). As mentioned in the wading pool earlier, the chart is misleading, as it doesn't account for inflation. Once you adjust for inflation — particularly rampant in the economy between 2021 and 2023 —the price increases look much smaller.
Electricity prices rose 27% from 2021 to 2025; the inflation-adjusted increase was 6.8%
Average residential electricity price (cents per Kilowatthour), 2001–2025
As we start breaking down the data by state, we’ll focus on changes between the end of 2022, when ChatGPT, the first widely used consumer AI product, was released, and the end of 2025. If AI data centers were driving electricity prices higher, we’d expect to see evidence of it during this period of rapid AI adoption.
Where have electricity prices risen?
National changes in electricity prices may have been relatively modest over the past few years once they’re corrected for inflation. However, price changes vary by state.. Electricity prices have actually fallen in some states, regardless of data center presence.
In other states, however, prices have risen sizably.
Changes in residential electricity prices varied widely across locations, ranging from large declines in Hawaii to large increases in Washington, DC.
Average residential price of electricity by state (cents per kWh)
The largest increases in residential electricity prices over the three-year period were in Washington, DC, New Jersey, and Maine.
Washington, DC, New Jersey, and Maine had the largest residential electricity price increases from 2022 to 2025.
Top 10 locations with the largest percent increases in residential electricity prices (cents per kWh)
Steep electricity price swings are nothing new.
Some of the state-level changes in residential electricity prices in recent years seem quite large. Washington, DC’s residential electricity prices rose 45% from December 2022 to December 2025. (They rose even more, 55%, from September 2022 to September 2025. Has such a thing happened before?
Yep. Looking at residential electricity prices in every month since 2000 and comparing them with three years earlier reveals 10 instances in which a state’s prices grew by more than 50%. There were 758 instances in which they were more than 20% higher.
One way to think about this is that, historically, there has been roughly a 1 in 20 chance that a state’s electricity prices will rise 20% (on an inflation-adjusted basis) over the next three years.
In other words, well before AI and the accompanying electricity-guzzling data centers, there have been plenty of times that state electricity prices rose a substantial amount in a short time.
Since 2000, there have been 10 instances of state residential electricity prices rising more than 50% in three years
Largest residential electricity price percent increases over three-year periods, by location
Where are data centers located?
Data centers have spread to every corner of the United States — every state in the lower 48, plus Washington, DC, now has at least one. But they are far from evenly distributed. Roughly 41% of existing data center capacity is in just two states – Texas and Virginia.
Roughly 41% of data center capacity is in Texas and Virginia
Megawatts of existing data center capacity by state, 2025
Why these two states? Texas has abundant energy sources, vast land availability, and a policy mix that includes fast-tracked infrastructure permitting and targeted financial incentives to qualifying data centers.
Historically, it has also operated its own electrical grid with little federal regulation, allowing new projects to connect to the grid faster than in other states, though the state recently enacted a temporary freeze and mandatory audit on new data center connections.
Virginia has strong network infrastructure, proximity to federal clients that demand near-instant data delivery, and policies that are favorable to the tech industry.
About 75% of existing data center capacity is in 10 states:
- Texas
- Virginia
- Ohio
- Oregon
- Georgia
- Arizona
- Iowa
- California
- Illinois
- Washington
How much electricity do data centers use?
The government does not comprehensively track electricity consumption from data centers.
A 2024 report commissioned by the Department of Energy estimated that data centers consumed roughly 176 terawatt-hours in 2023, representing about 4.4% of US total electricity consumption. But there has not been a government estimate since then.
A Congressional Research Service report notes that there can be great variation in data center energy consumption estimates depending on the methodology. There has been a bi-partisan push to collect more data on the matter.
Has there been a relationship between data center capacity and residential price changes?
One simple, though far from perfect, test for whether data centers are driving up residential electricity prices is to see there were large price increases in states with more data centers. But the data from 2022 through 2025 does not show a statistically significant relationship between data center concentration and changes in average residential electricity prices
States with some of the largest residential electricity price increases don’t have many data centers compared to other states. Meanwhile, residential electricity prices in Texas and Virginia didn’t rise as much as in the average state.
States with more data centers did not see larger electricity price increases
Percent change in average residential electricity prices by megawatts of existing data center capacity, Dec 2022–Dec 2025
That said, we have to be cautious in inferring too much from this analysis. Some states share electricity because of the way the grid is organized. Electricity demand from data centers in one state might cause higher prices in a state that shares electricity.
Can other factors explain recent price increases?
A recent analysis by the Lawrence Berkeley National Laboratory and funded by the Department of Energy examined state-level changes in retail electricity prices from 2019 to 2025. In the words of a Congressional report summing up the analysis, “They found that the main driver behind increases in electricity prices during this period was utility investments in grid infrastructure, mostly in response to aging infrastructure and resilience needs. Other key drivers included natural gas prices, recovery from natural disasters (e.g., storms, wildfires), and state energy and environmental policies. In other words, this analysis did not identify data centers as major influencers of electricity prices in most areas of the country between 2019 and 2025.”
Notably, the original report found that states that had the highest load growth over this time period – e.g., the highest increase in electricity consumption – actually saw the all-sector average of electricity prices go down.
That said, it did show that the PJM capacity auction was driving higher capacity prices, which in turn contributed to higher retail electricity prices in some Northeastern states, particularly Pennsylvania and New Jersey.
What about the future? Will AI drive up the electricity prices?
How much electricity will data centers consume in the future?
There is one official government report that predicted future data center load. It came out in December 2024 and predicted that US data centers would consume 6.7% to 12.0% of total electricity in 2028.
Other companies and agencies have come up with their own projections, though each differs slightly in what they are predicting and when. Here are the most cited estimates:
- McKinsey & Co. predicts that data centers will use 11.7% of total US power demand by 2030.
- The Electric Power Research Institute projects data centers will use 9% to 17% of all US electricity by 2030.
- Boston Consulting Group projects that data centers could use as much as 1050 Terawatt hours of electricity by 2030, which would represent about 25% of all US electricity generated in 2023.
- Goldman Sachs predicts data centers will take up 8 % of total US power by 2030.
Why is predicting the future of AI’s electricity use so hard?
There is a famous saying attributed both to the Nobel Prize-winning physicist Niels Bohr and the witty baseball manager Yogi Berra: it’s tough to make predictions, especially about the future.
This is always true with a new technology and it is easy to find a long list of laughably bad predictions about both future technologies and the energy market.
For example, in 1999, a highly publicized study claimed that the internet would consume as much as 50% of the entire United States power grid within two decades. The study was presented to Congress and featured prominently in testimony before the U.S. House Subcommittee on National Economic Growth. To be fair, some of the reason for this faulty prediction was that it was based on a faulty projection of current internet use, exaggerating the amount of electricity the internet was using at the time by a factor of 2,000.
In addition to the usual difficulty predicting the future, there are some unique challenges to predicting the future of AI and electricity prices. There are many questions that are extraordinarily difficult to answer.
How much will AI use grow?
Artificial int elligence is growing rapidly. ChatGPT became the fastest-growing consumer application ever, reaching 100 million monthly active users in two months.
Global use of AI tools reached 53% of the global population in three years, faster than the personal computer or the internet.
Given how fast this technology is moving, it's no surprise that expert predictions about where it's headed vary wildly.
Demis Hassabis, the CEO of Google DeepMind, says that artificial intelligence will be “10 times bigger than the Industrial Revolution and maybe 10 times faster.” Dario Amodei, the CEO of Anthropic (the parent company of popular AI tool Claude), has said that AI is a “general labor substitute for humans” and might be doing the majority of entry-level white-collar work within a few years.
On the other hand, the Nobel Prize-winning economist Daron Acemoglu is skeptical and thinks AI can realistically only take over about 5% of jobs.
We might amend the Bohr or Berra quote: Prediction is incredibly difficult, especially about the future of an undoubtedly powerful but unprecedented and rapidly-changing technology.
In other words, the future of AI usage depends on a whole host of factors: technological breakthroughs or limitations; popularity and utility of the consumer products; and government policies, all of which are highly uncertain.
How will AI energy consumption change?
Just as AI adoption has grown rapidly, the cost of running AI models is changing. Some reports suggest the cost of answering a user query dropped more than 250-fold in just a two-year period, though there is some debate about the accuracy of these numbers.
Companies and venture capitalists are investing to develop new chips that lower the energy demand of AI models.
How will AI and data centers be regulated?
We are not the only ones who noticed increased interest in AI and electricity prices. When USAFacts visited Washington, DC, recently, this was clearly top of mind for both Democrats and Republicans.
In the first three months of 2026, there were more than 1,500 AI-related bills introduced in state legislatures. This already surpassed the total for the entire year of 2024.
And there have already been recent actions to regulate data centers.
Pennsylvania Governor Josh Shapiro sued PJM, which lead to a four-year cap on how high the capacity auction could go.
In 2026, New York state passed a one year moratorium on the construction of large new data centers. And Seattle passed a temporary freeze on new data centers.
Virginia has implemented a new rate class for large customers, such as data centers, effective at the start of 2027.
Ohio has implemented a tariff that forces large customers to pay 100% of buildout costs.
There is little consensus about which bills might be passed in the future or which new bills might be introduced.
Usually, at USAFacts, we try to give you answers. And there are striking — and sometimes counterintuitive — patterns in electricity prices over the past few years. We hope they have helped give you a more nuanced understanding of how AI data centers have influenced electricity prices. But when projecting the future of data centers and electricity prices, we are left with many open questions. We will be tracking these closely and sharing the data as it becomes available.
Read about AI use in the US, see how AI and tech are evolving at USAFacts, and get fresh data delivered to your inbox weekly by signing up for our newsletter.
