As data centers go up, North Carolina weighs how to handle energy demand
In small communities across North Carolina, data centers are already sparking conflict over land use, water use, and quality of life. Now, the debate over the facilities’ voracious need for electricity — and whether it can be met with clean sources — is heating up in the state capital of Raleigh. For months, North…
In small communities across North Carolina, data centers are already sparking conflict over land use, water use, and quality of life. Now, the debate over the facilities’ voracious need for electricity — and whether it can be met with clean sources — is heating up in the state capital of Raleigh.
For months, North Carolina’s predominant utility, Duke Energy, has forecast ballooning demand from large customers like data centers: immense buildings that house the computing devices powering AI and other software that’s become part of everyday life.
Early last year, Duke projected these “large loads” would need an additional 3.9 gigawatts of capacity, equal to about four nuclear power plants and enough to serve millions of households. By May of this year, the company’s prediction had swelled to almost 6 gigawatts.
The eye-popping estimates helped lead regulators to approve Duke’s current plan to build a massive new fleet of gas plants, alongside some clean energy investments, despite a state law requiring the utility to decarbonize. The projections are certain to factor into the next iteration of Duke’s long-term blueprint, a draft of which is due in the coming weeks.
The forecasts have “thrown everything out of whack,” said Nick Jimenez, senior attorney with the Southern Environmental Law Center.
That’s why his organization asked the state’s Utilities Commission to host a technical conference on large loads. Electricity-demand projections undergird virtually every Duke case before the panel. But at a technical conference, commissioners could grapple exclusively with the issues vexing energy experts across the country: How can data center demand be predicted with the most accuracy? Will the tech giants pay their fair share of grid upgrades and other costs? What will power the new facilities, and will it be carbon-free?
In June, the Utilities Commission granted the law center’s request and then some by opening an entire proceeding to debate these questions. Stakeholders had the summer to submit written comments, with responses due from Duke early this month. In-person presentations are scheduled for Oct. 14.
It’s not clear if the process will culminate in a discrete order from the commission, or simply inform the myriad other Duke cases before it. But Jimenez praised regulators for being proactive. “You need a proceeding to get your arms around some of these issues,” he said. “I think that’s really smart and forward-looking.”
The data center boom
In the race against other states to attract economic development, Duke and North Carolina officials keep confidential exactly which entities hope to draw power from the electric grid. And skeptics question whether all of the new facilities behind predictions of unprecedented demand growth will pan out.
But there’s little doubt that data centers are on the rise, propelled by the AI explosion. Researchers say they could account for 44% of U.S. load growth by 2028, and there’s ample evidence that North Carolina is following the national trend.
In June, Amazon Web Services announced a $10 billion, 800-acre computing campus in Richmond County, east of Charlotte, billed as the largest single capital project in North Carolina history. To the west of Charlotte, the development of a “data center corridor” is underway: Apple says its Catawba County site is included in its $500 billion U.S. expansion plans, and Microsoft envisions four new data centers nearby. Google is considering growing its facility in neighboring Caldwell County.
Not all communities are welcoming data centers with open arms. The town council of tiny Tarboro, an hour east of Raleigh, just voted to reject a $6.4 billion facility. In Apex, southwest of the city, opposition is mounting to a proposed “digital campus” that would displace 190 acres of farmland.
Still, early this month, Gov. Josh Stein, a Democrat and former attorney general, issued an executive order creating an “AI Accelerator” and a council designed to make the state “a national leader in AI literacy, governance, and deployment to the benefit of our residents, communities, and economy.”
Stein did note the technology’s downsides, including “the uncertainty around AI systems and their associated energy and water needs.” But his edict also reflects the seeming common wisdom of the moment: AI and its requisite facilities are multiplying and expanding, bringing economic opportunities that can outweigh their challenges.
“We can come to the table”
In the open docket before regulators, experts say that with the right policies in place, clean energy, efficiency, and related strategies can meet the moment. “We can come to the table,” said John Burns, general counsel for Carolinas Clean Energy Business Association, a trade group representing developers, manufacturers, and others in the clean energy industry.
In their comments, Burns and others particularly promoted “load flexibility,” a form of demand response in which data centers curtail their electricity use when the grid is strained by lots of energy consumption.
Load flexibility is feasible because data centers don’t run at maximum capacity 24/7, said Tyler Norris, former special adviser at the U.S. Department of Energy and a doctoral fellow at Duke University, which has no connection to the utility.
“You never actually run the chips and the servers to 100% of their rated nameplate power,” he said. “You wouldn’t want to, because they overheat and they don’t perform as well when they’re running that hard.”
Norris is the lead author of a February paper showing that Duke’s two utilities in the Carolinas could accommodate 4.1 gigawatts of load if data centers shave just 0.5% off their peak usage annually. In a simple example, the facilities could operate at half their maximum capacity for 88 hours over the course of a year.
A load-flexibility arrangement between Duke Energy and data centers could, in theory, avert the construction of several gigawatts of new gas plant capacity and expensive and time-consuming transmission upgrades.
Last month, Google announced demand-response agreements with the utilities Indiana Michigan Power and the Tennessee Valley Authority. In formal comments to the North Carolina Utilities Commission, Norris called the tech giant’s move the “first documented case where AI data center flexibility is explicitly integrated into U.S. utility planning.”
