Americans fear AI data centers will increase their electricity bills. Congress has developed a plan to protect taxpayers, but there is no easy way to know what AI is costing them.
Transcription
ELISSA NADWORNY, HOST:
The construction of huge new AI data centers has become a political lightning rod. Communities near proposed projects are raising questions about their impact, particularly on electricity costs. A big question is who will end up paying for the expensive new infrastructure needed to power these massive facilities. NPR’s Scott Neuman has been looking into this question, and he joins me now. Hi, Scott.
SCOTT NEUMAN, BYLINE: Hello, Elissa.
NADWORNY: So, Scott, most of us know by now that these large-scale AI centers require a lot of energy and power grid upgrades. Who pays for all this?
NEUMAN: Well, the simple answer is: everyone who buys electricity. This of course means the people who build and use these data centers, but also local businesses, landlords and tenants, even if they don’t think about using AI. While reporting last week, I spoke with Ari Peskoe. He is the director of the Harvard Electricity Law Initiative. He says ultimately everyone bears the burden of the billions of dollars spent on new improvements to the power grid.
ARI PESKOE: Everyone pays these costs. They are distributed throughout the region for every business and every resident with an electric meter.
NEUMAN: Now, in many cases, the pace and scale of demand is coming at a truly dizzying speed, which makes it all unpredictable. Demand for new transmission lines, substations and other infrastructure is growing faster than utilities can in many cases meet. And that creates a huge planning challenge. But you know, Peskoe brings up another interesting point. He says the speed with which this is all happening can actually drive up the cost of these improvements. Components, like transformers, are rare, and everyone wants them, and wants them now. So there is a shortage, and that makes everything more expensive.
NADWORNY: OK, so we all pay for AI upgrades, but how are those costs distributed?
NEUMAN: That’s a really difficult question to answer. Even if there is an AI data center in your area, that is the reason for these improvements, everyone will be able to benefit from it. But Peskoe says no one has really figured out how to distribute the costs fairly.
PESKOE: There is no agreement on what data center infrastructure is versus public infrastructure. This in itself is a controversial point.
NEUMAN: So in 2024, Maryland’s Office of People’s Counsel, an independent state consumer agency, projected that new data center costs would cost residents up to $216 a year. But the rapid construction of these data centers creates a lot of uncertainty.
NADWORNY: Do individual customers benefit from all this new network infrastructure?
NEUMAN: Yes, so the fact that everyone can benefit from these network improvements means that everyone benefits equally. Lucy Qiu is a professor at the University of Maryland who studies all of this. She says in many places the improvements are long overdue. They will be needed for AI data centers, but they will also be needed for things like meeting electric vehicle charging capacity.
LUCY QIU: On the positive side, you can accelerate the upgrade of network infrastructure.
NEUMAN: Qiu says studies show that people who live near data centers where the network has been made more robust to accommodate them, residential ratepayers experience fewer power outages during extreme weather events.
NADWORNY: Okay, so a positive side too, thank you. It was NPR’s Scott Neuman.
NEUMAN: Thank you.
(SOUNDBITE OF MUSIC)
Copyright © 2026 NPR. All rights reserved. Visit the terms of use and permissions pages on our website at www.npr.org for more information.
The accuracy and availability of NPR transcripts may vary. The text of the transcript may be revised to correct errors or match updates to the audio. Audio on npr.org may be edited after its original broadcast or publication. The authoritative record of NPR’s programming is the audio recording.
Gn bussni
