NPR.
Recently, Senator Bernie Sanders and Representative Alexandria Ocasio-Cortez introduced a bill to pause new AI data center construction.
Among their complaints?
Electricity costs.
People's energy bills around the country are skyrocketing in order to pay for these AI data centers for them.
Over the last five years, electricity prices have more than tripled in areas close to data centers.
That's according to Bloomberg.
These price rises have contributed to local opposition, blocking billions of dollars worth of data centers.
And this week Maine's House passed initial approval on pausing most data center construction in the state.
Both the Senate and governor support the bill, so it's expected to pass.
J.J.
Assyria is a utilities consultant and he says that data centers are a scapegoat for longstanding issues in the electricity market.
I think data centers are a factor, but to say that they're the ones solely responsible for higher electricity prices, I think, is perhaps misleading.
So who's right?
This is The Indicator from Planet Money.
I'm Darian Woods.
And I'm Waylon Wong.
Today on the show, data centers and your power bill.
We excavate years of underinvestment in the grid and ask whether the current wave of AI data centers could possibly even help.
Jay Jayasiriya is a consultant in Houston, Texas for Sendero Consulting.
That's where Jay works with electricity companies that power data centers.
And Jay has a provocative thesis.
He believes that AI data centers are unfairly getting the blame for more than a decade of underinvestment in the US grid.
What's the state of America's electricity grids?
They are functioning.
You see a lot of the grid upgrading in certain ways.
There's new lines and transmission lines going up.
There's the upgrade to be able to transport more power across the grid.
But Jay thinks that America's power grid has been underinvested in for 15 years, way before the AI boom.
You've got power lines that can be vulnerable to storms and could spark wildfires, big renewable projects that can't connect to the grid as it is.
Jay thinks that companies building AI data centers are, by contrast, actually trying to find new ways to get power.
A lot of the data centers that are going up are responsibly finding ways to acquire power through these long-term deals with providers or spending the money, investing the money to build a co-located generator that will serve the data center needs.
Right, they're putting natural gas plants on the data center site sometimes.
Yeah, I mean a natural gas plant, a combined cycle plant, you know, even looking at smaller nuclear options.
Jay says that having the data center providing its own power on site could have spinoff benefits for the rest of us.
That's if the data center doesn't need all the power it's generating.
Whatever excess they have, they can provide into the grid.
Another way data centers could actually help the grid is paying for fixed costs maintaining power lines, clearing overgrown trees from those lines, building substations to convert electricity into the right voltages.
These would be costs for the utilities, no matter how much power an individual house or business bought each day.
And so when there's more electricity being generated and sold, those fixed costs as a share of your final power bill are lower.
It's just like how a car company can bring down the price of cars by selling more of them.
Economies of scale.
In fact, earlier this year, the big power company in California PGE, said that data center growth had helped it cut electricity rates by 13 since 2024.
And they cited the ways that data centers paid for more fixed costs as contributing to this.
But it's worth a reality check here.
Philip Krein is a professor of electrical and computer engineering at the University of Illinois, Urbana-Champaign.
We asked Philip Krein what he thought of Jay Jayasuriya's idea that data centers are the scapegoat for years of grid underinvestment.
I don't accept that.
I think that what's going on right now is data centers are walking in and say hey, we want 20 more capacity.
We want it today. that's just not a normal situation.
Yeah so Philip, it's the speed at which they want it that's causing strains and, in many places, higher electricity bills.
Data centers and computing basically are sucking up 100% of potential future capacity.
And that is getting reflected back into the current rate base.
Phillips says it doesn't matter whether or not the grid had been properly investing over the last 15 years.
Either way, the grid would struggle to meet this rapid increase in demand.
Yeah, not everywhere is like California where power bills are going down.
Take Northern Virginia.
Its electricity grid is already strained to its existing capacity.
So a new data center needs new grid investments, new high-voltage transmission lines, new substations, for example.
That means higher bills.
That said, Philip believes it's possible that the US can increase its electricity generation and transmission by a lot.
After all, it's already done this in the past.
Between the late 1920s and the late 1970s, the U.S. power grid grew by about a factor of 12.
So the notion that we need to grow again by 20 or 30 or 50 percent is a little bit of a been there, done that kind of a thing.
On the other hand, Phillips says the National Grid kind of stopped growing in the 1980s.
That's because appliances and factories got more energy efficient, so there wasn't much push for expanding capacity.
Now we have growth in electric vehicles, air conditioning, home heating converting to electric and, of course, AI data centers.
Whether we can get back into that mode is the interesting question.
But we've done this before.
We know how to do it.
Phillips says growing the electricity sector is a matter of getting the right financial incentives, making sure it's economic for generators to build new capacity and send it to customers.
One example, electricity at the moment is regional.
That means, say, New York in the evening can't buy cheap solar power from the Texas afternoon.
What could solve this would be building huge transmission lines that could move electricity from state to state.
The United States needs very significant growth in the cross-country transmission infrastructure.
I don't think anybody who knows the situation disagrees with that.
But You know, a transmission line from Chicago to New York or California to Chicago goes through so many jurisdictions and so many layers of approval that it's a very unwieldy kind of a project.
Yeah, I'll be honest.
All that compliance and permitting sounds impossible.
But Philip says it's not.
We need to go back and see what we've done before and some of what we need to do again.
Data centers in general are contributing to rising electricity bills.
California is an example of one state that's an exception to this.
But both Philip and Jay agree there's an opportunity to use this spike in power demand to build a better grid.
This episode was produced by Cooper Katz McKim with engineering by Kwesi Lee.
It was fact-checked by Sierra Juarez.
Kate Kincannon edits the show and The Indicator is a production of NPR.
