You're paying for AI on your electric bills

If you built a remote cabin in the woods and needed power, you’d pay the local utility company to get connected. Same thing if you’re building a solar farm or a wind project—you’d pay your own way. 

But when a billion-dollar tech company builds an AI data center, something very different happens: Utilities build out massive new infrastructure, then quietly pass the bill to you.

“From most perspectives, this is morally wrong and bad policy,” said Mike Jacobs, Senior Energy Analyst from the Union of Concerned Scientists.

Jacobs recently published a report finding 130 cases where utilities spent $4.3 billion on new grid equipment to serve data centers—while shifting those costs onto everyday customers' monthly electric bills. And he told EnergySage $4.3 billion is just the tip of the iceberg.

We sat down with Jacobs and other energy experts to find out how AI is already hitting your wallet, why tech companies are being spared the costs, and what this means for your energy bills.

(To be clear: We aren’t bashing AI as a technology. We’re exposing how much power AI data centers are using and why that's going to cost you money.)

Before we get to the costs, it’s important to understand how much energy AI consumes.

Every time you ask a question to ChatGPT, Google Gemini, Copilot, etc., you're using enough electricity to power one lightbulb for about 20 minutes. That might not sound like much—until you consider that millions of Americans are asking questions every minute of every day.

Those questions don't just float around "the cloud." They're processed in physical AI data centers, and these buildings are becoming some of the largest energy users on Earth. Jacobs says one data center in Louisiana uses three times more electricity than the entire city of New Orleans.

While traditional data centers have a power demand around 300 megawatts, new AI data centers demand upwards of 1,000 megawatts—"roughly what some small states have for utility-owned generation in the state—Vermont-sized states," said Jacobs.

Jacobs’ report focuses on PJM Interconnection, the grid operator for Illinois, Maryland, New Jersey, Ohio, Pennsylvania, West Virginia, and Virginia. The latter is now nicknamed "Data Center Alley."

"We were hearing Virginia was having increased energy needs, and then gradually over the last two years, the transmission discussions were about meeting those needs. And then it just blew up like a bomb," said Jacobs.

Carnegie Mellon researchers estimate that data center electricity demand will grow 350% by 2030. But utilities usually plan for only 1-2% annual growth, so the math isn’t mathing.

Plus, our grid is already under strain. A 2024 report found that homes within 50 miles of a data center were experiencing electricity issues, and experts globally are warning about reliability risks and power shortages.

"The ecosystem of how these things are operating is really stressing our resource supply,” said Lizzie Irwin, Policy Communication Specialist at the Center for Humane Technology. “The way the tech is being developed and deployed, it's happening at a rate that is really fast and sort of unfair and maybe unreasonable to ask people to just simply adapt.”

When energy demand explodes, but the system wasn't built for it, utilities must build more power plants, substations, and transmission lines. All of that costs money.

But the AI tech companies aren't paying for it. You are. And you probably didn't even know it.

Barklie Estes, President of Nova Solar in Virginia, lives in the heart of Data Center Alley. He says that even though data centers have been a hot topic for a long time, there's been zero communication from utilities about how these costs are being passed to the community.

"The big boys don't even try to justify. They just send you the bill," said Estes.

We reached out to Dominion Energy, the major utility in Virginia, for comment, but didn't hear back. However, Dominion officials spoke with NPR about one of their upcoming utility projects. While the officials said it's being built to meet growing demand in the community, they also admitted the data center "is largely driving the need for this project."

According to NPR, the project would be a data center and substation about the size of eight football fields with 120-foot steel poles. The cost to connect that data center to the grid? $40 million. And that's just one project.

Jacobs reviewed utility documents across seven Mid-Atlantic states and found 130 data center projects in 2024 costing $4.3 billion just for grid connections. That number doesn't include the cost for power plants needed to actually power the data centers, which Jacob says will add another $20 billion in the PJM region alone.

That's billions of dollars placed on regular people—families, small businesses, renters—to build infrastructure for some of the richest companies on the planet.

And don't think you're off the hook if you live outside the Mid-Atlantic. "There's additional money in a separate conversation buried outside of the PJM databases," said Jacobs. "This is likely happening on more of a nationwide level, but it's just not as transparent."

Carnegie Mellon's research backs this up: Electric bills are on track to rise an average of 8% nationwide by 2030. But in places like Virginia, bills could rise as much as 25%—about $70 more every month for the average household.

"People are telling me 'this is early days' with generative AI. So the wave is still coming," said Jacobs.

So, how exactly are tech companies getting these billion-dollar grid connections while we pay for it?

AI data centers use so much energy that utility companies can classify their work as a transmission project instead of a simple connection.

"If the utility can make the argument that this is a transmission extension, it goes into the rates for everybody to pay. And they just haven't really asked the data center companies anything else," said Jacobs.

When Amazon, Google, Meta, or others want massive new connections, utilities build them and make all of us ratepayers split the bill. When Jacobs confronted utilities about his findings, he said "they all fully concede that I've described it correctly. They just say there's nothing wrong with that."

Several governors from the PJM states met with the Trump Administration in January to discuss forcing data centers to pay for new power plants and protecting ratepayers by capping utility prices. While Jacobs supports the notion, he's skeptical these changes will happen anytime soon.

"There's a whole lot of 'How are you going to do this?' still to be worked out," he said.

Another problem Jacobs has with the new proposal? The Trump Administration called PJM's plan to use wind and solar energy to power data centers a "serious market failure," instead suggesting they use "reliable baseload generation resources (coal, natural gas, nuclear)."

"These days, it's whack-a-mole with clean energy," said Jacobs.

Meanwhile, many of these massive data centers are rolling in dozens of diesel generators on flatbed trucks just to stay running.

"Nobody in their right mind does that—nobody in their right mind except in an emergency or a forward operating military base," said Jacobs. "These are customers in our neighborhoods who are acting like it's a war zone or a disaster zone. That's not sustainable."

Jacobs has a better solution: "You've got to do this with renewable energy. The best thing here would be energy parks that have both supply and demand with lots of wind farms inside that park, or directly attached lots of solar and batteries to handle both the variations of the generation and that spiking and dropping behavior of the data centers."

But clean energy isn't just being ignored—it's being actively blocked. Virginia counties are approving massive data centers while rejecting solar projects that could help power them.

"We had a couple of examples in 2024 where Hanover County approved a data center, and they simultaneously rejected the application for a utility-scale solar field. Culpeper County did the same thing," said Estes.

"This has been the saddest story I've ever seen in my career," said Jacobs. "If this were just about money, that would be one thing. The truth is, we're talking about first blocking clean energy. But now we actually have an all-out shortage of power."

Is AI worth the cost?

Tech companies say that AI will revolutionize healthcare, education, and scientific research—and maybe it will. But Irwin says AI is currently being deployed too quickly and recklessly.

"I think there's this real gold rush mentality among big tech companies and the developers that's leading them to acting this way, and it's not really considering the effects on the environment and the communities and the way that it's impacting costs," said Irwin.

Aside from the increased electricity demand, MIT News reports that AI is straining water supplies and disrupting local ecosystems. Additionally, data centers currently account for 2.5-3.7% of global greenhouse gas emissions, according to Columbia Climate School.

"I candidly don't think that the promises of AI justify this bad behavior,” Irwin said.

Even though this is a national problem, it's fought locally.

If this news is making you angry, grab your neighbors and get to work: City council, county boards, utility commissions. Call your state utility commission, email your representatives, and demand transparency.

In the meantime, protect yourself from rising electric rates and grid instability.

"Get yourself enough solar panels and batteries so you feel good about your own electricity supply, because right now it's being corrupted by these data center requirements," said Jacobs.

If you can't install solar and batteries, look into community solar or, if you live in a deregulated state, sign up for a clean retail energy plan. If those aren’t available, push your officials to allow more clean energy in your community.

You may think local action is a waste of time, but here's proof it's not: In the last year, 25 data center plans were canceled as a direct result of opposition from local people. Concern from locals has made its way up to senators from both parties, who are now speaking out against AI data centers. That means change is possible.