Seattle Ferguson stood before the Virginia State Corporation Commission during Dominion Energy's 2025 biennial rate review and did something regulators rarely encounter in their wood-paneled proceedings: she converted a utility's revenue request into the arithmetic of her own life. Twenty dollars a month, she told the commissioners, "might not seem like a lot to the Dominion CEO…but for me, it's massive." Ferguson, a Virginia Commonwealth University student, said her June electricity bill had come in $50 higher than expected after running air conditioning through a heat wave. She was testifying about what it costs to stay cool.
She was one voice in nearly 3,000 pages of public comments filed ahead of the September 2025 hearings, most saying some version of the same thing: bills are climbing on top of rising rents, groceries, and gas. Dominion had requested $1.17 billion in additional revenue, adding roughly $14.73 per month to a residential bill the first year, with another $4.84 layered on after that. For Ferguson, that's the difference between groceries and not.
What most of those commenters probably couldn't see is where the pressure on the grid is actually concentrated. Drive west from D.C. along the Dulles Toll Road into Loudoun County and the answer sits between the subdivisions and the strip malls: windowless concrete buildings the size of aircraft hangars behind security fences, rooftop cooling units running around the clock. This is Data Center Alley. More than 300 facilities operate in Loudoun County alone, making it the densest cluster of data-processing infrastructure on Earth. And according to the Virginia Joint Legislative Audit and Review Commission's November 2024 report, unconstrained demand could double Virginia's data-center electricity load within a decade.
A typical Dominion residential customer could face $14 to $37 more per month by 2040, driven by generation and transmission investments needed to serve data-center load growth.
The rate case is the place where that growth meets a family's kitchen table. When a utility needs to build billions in new generation and transmission capacity to serve an industry concentrated in one part of its territory, how do those costs get divided among everyone who shares the grid?
Forty-five cents of every tax dollar
To understand why Loudoun County can't simply stop approving data centers, look at the FY2027 adopted budget.
Real property taxes assessed on data centers: $417.7 million. Personal property taxes on the computer equipment inside them: $879.1 million. Together, about $1.3 billion, or roughly 45 percent of the county's $2.9 billion in local tax revenue. The computer-equipment line alone grew from $581.6 million in FY2023 to $879.1 million in FY2027, increases the county's own fiscal trends analysis attributes to the "strong performance and expected future performance of the data-center industry." That single personal-property-tax line accounts for about 77 percent of all personal property tax revenue the county collects, excluding state reimbursement.
Mike Turner, vice chair of the Loudoun County Board of Supervisors and the Ashburn district supervisor representing the heart of Data Center Alley, has been unusually direct about the bind.
"Loudoun has taken a maximum benefit from data centers financially."
— Mike Turner, to Le Monde
He acknowledged what he hears back from residents: they don't care that data-center revenue keeps property taxes 25 percent lower. They'd rather raise taxes than keep building.
Residents are doing their own math, weighing the tax benefit they can see against the costs they're absorbing in ways the tax bill doesn't capture. Higher electric rates. A landscape transforming around them. A water system whose capacity during peak summer demand nobody can verify with public data. They've concluded the tradeoff isn't working, and they're saying so out loud.
The county knows the revenue stream is fragile. Its FY2027 budget established a Revenue Stabilization Fund sized at 10 percent of data-center-associated tax revenue, designed to cushion the county if the industry contracts or relocates. You don't build a rainy-day fund around the possibility that your largest taxpayer might leave unless you've already accepted how precarious the arrangement is.
And the revenue picture omits a statewide counterweight. JLARC found that Virginia's sales and use tax exemption for data-center equipment cost the state $928 million in forgone revenue. The industry driving the need for grid expansion is simultaneously exempt from nearly a billion dollars in taxes that could otherwise fund public infrastructure.
Where the cost lands
JLARC projected that a typical Dominion residential customer could face $14 to $37 more per month by 2040, driven by generation and transmission investments needed to serve data-center load growth. That projection landed in November 2024. By April 2025, the first installment was already in the rate case.
The range between $14 and $37 depends on two variables: how fast data-center demand actually grows, and how costs get allocated between customer classes. The second variable is where rate design becomes a question about fairness that most ratepayers never get to weigh in on.
Dominion proposed creating a new customer class for large-scale electricity users, including data centers, with minimum charges to protect other customers and the utility if a large customer abandons a contract after infrastructure has been built. Amazon, Google, Microsoft, the Data Center Coalition, and some industrial users countered with lower floors.
| Generation costs | Transmission & distribution | |
|---|---|---|
| Dominion proposed | 60% minimum | 85% minimum |
| Industry counter | 50% minimum | 75% minimum |
The gap between those two sets of numbers determines how much of the cost of building power plants and stringing high-voltage lines to serve data centers gets recovered from data centers, and how much gets spread across every other Dominion customer. In November 2025, the SCC approved creation of the new rate class for the biggest electricity users.
A new rate class is an institutional milestone. It means the SCC recognized that data-center load is categorically different from other commercial demand. But a rate class is a container. What minimum obligations it imposes, and whether those obligations keep pace with the speed of load growth, are the conditions that determine what shows up on a residential bill in 2028 or 2035. For Seattle Ferguson, the question is concrete: does the approved rate structure mean her share of grid-expansion costs goes down, stays flat, or keeps climbing? The answer depends on the minimums set inside that container, and on whether data-center demand grows at the pace Dominion projects or faster.
When the grid flinches
The cost question assumes the grid holds. There are reasons to pay attention.
In July 2024, a high-voltage transmission line in Virginia failed, and roughly 70 data centers disconnected from the grid simultaneously. In February 2025, another line malfunctioned and about 40 data centers switched to backup power. Neither incident triggered an emergency. But PJM's senior vice president of operations noted the obvious next question: what happens when it's 3,000 megawatts, or 5,000?
The counterintuitive part: during those incidents, the data centers didn't overwhelm the grid by drawing too much power. They vanished from it too fast. They're designed to switch to backup generators at the first sign of a voltage disturbance, protecting the servers inside. When dozens of facilities do this simultaneously, the grid operator loses a city's worth of load in seconds, creating the kind of sudden imbalance that can cascade into wider instability. Dominion has been working with tech companies on ways for facilities to ride through brief grid faults rather than switching to backup power. NERC convened a February 2026 conference with Google, QTS Data Centers, and others to address the issue.
Residents worry about whether there's enough power to go around during a July heat wave. Grid operators share that concern, and they carry an additional one: the scenario where the grid's largest customers all flinch at the same moment.
Nationally, NERC's 2026 summer assessment found that peak demand grew by about 11 gigawatts from the prior year, driven by data centers and other large customers. The assessment identified 58.5 gigawatts of new resources entering service, mostly solar and batteries, and concluded the country should have sufficient supply under normal conditions. Normal conditions means no prolonged heat dome, no drought reducing hydropower, no wildfire smoke forcing plant curtailments. PJM, the grid operator serving 67 million people across the Mid-Atlantic and Midwest, received Department of Energy emergency permission to curtail power to data centers and other large customers with onsite backup generation as a last resort.
Northern Virginia sits at the sharpest edge of this national pattern. Across the country, proposed power plants and transmission projects are stacking up in interconnection queues, waiting years for approval to connect to the grid. The load growth driving those queues clusters where data centers cluster, and nowhere do they cluster more densely than Loudoun County.
The water nobody counts
Electricity dominates the data-center debate, but water runs beneath it. JLARC found that Virginia data centers consumed about 2.1 billion gallons in 2023, an 86 percent increase since 2019. Roughly a third came from reclaimed water rather than potable supplies.
JLARC concluded that current data-center water use is sustainable at the state level but acknowledged that water is more limited in some localities, particularly those without large surface-water access. The commission recommended authorizing local governments to require water-use estimates as part of the permitting process. That recommendation points to its own gap: localities currently lack the authority to ask how much water a proposed data center will use before approving it.
No publicly available dataset reviewed for this story breaks out data-center cooling water against residential water use in the Loudoun Water service area on a peak summer day. A 2026 Caltech preprint (not yet peer-reviewed) argues that direct data-center water withdrawals can create peak water-capacity bottlenecks for host communities during the hottest days, even when annual volumes appear manageable. The gap between annual sustainability and peak-day competition is the one that hits a family running sprinklers during the same July week that data centers are cycling cooling water at maximum capacity.
The absence of the local number is itself a finding. A county that derives 45 percent of its tax revenue from an industry cannot tell its residents, with public data, how much of their shared water supply that industry uses on the hottest day of the year. I've learned to pay attention to what doesn't get measured. The absence of the record becomes the absence of the problem.
Living inside the tradeoff
Juli Briskman, the Algonkian district supervisor, has said that voting for data centers is "almost political suicide" because county voters are "done with them." When Amazon proposed a data center on the former George Washington University campus in Ashburn, Turner said he'd be "shocked" if the board approved it, calling Loudoun's infrastructure "pretty well saturated."
Saturation and dependence pull in opposite directions. A county that has built its budget around $1.3 billion in annual data-center revenue cannot easily absorb the fiscal shock of that revenue declining, even as residents demand a halt to new construction. The Revenue Stabilization Fund provides a cushion, not a solution. If the industry's growth plateaus or its equipment depreciates faster than new facilities are built, the gap between what the county needs and what it collects will open quickly.
The families living inside this don't experience it as a rate-design proceeding or a fiscal trends analysis. They experience it as a summer electric bill that keeps climbing, a water system whose peak-day capacity they can't verify, a grid whose growth has outpaced the infrastructure built to serve it, and a landscape filling with windowless buildings that nobody in the neighborhood asked for. JLARC's $14 to $37 monthly projection is a range whose lower bound assumes cost allocation shifts more burden to the industry, and whose upper bound assumes it doesn't. Where it lands depends on proceedings most ratepayers will never read, in a regulatory body most have never heard of, adjudicating formulas that determine whether the grid next door is a shared resource or a subsidy flowing in one direction. Residents like Ferguson, and the thousands who filed comments alongside her, have already done their own calculation. Whether any institution in the chain, from the SCC to the county board to the General Assembly, is structured to hear that answer and act on it before the next rate case arrives is the question Loudoun is living with right now.
-
FERC's cost-allocation order: In June 2026, federal regulators directed six regional grid operators to speed access for large data-center loads while requiring data centers to bear the full cost of necessary grid upgrades, a ruling that could reshape how Dominion and PJM divide infrastructure costs between industrial and residential customers.
-
The interconnection backlog: Berkeley Lab's queue tracker shows that more than 2,060 gigawatts of generation and storage capacity were actively seeking grid connection at the end of 2025, a bottleneck that determines how quickly new power sources can come online to meet data-center demand without straining existing residential service.
-
PJM's emergency curtailment authority: The Wall Street Journal reported that PJM received Department of Energy clearance to curtail power to large customers with onsite backup generation as a last-resort measure this summer, a tool that tests whether data centers can be treated as interruptible load during grid emergencies.
-
Peak-day water competition: A 2026 preprint by researchers at Caltech and other institutions projects that U.S. data centers could require 697 to 1,451 million gallons per day of new water capacity by 2030 if current water-use intensity persists, a figure that sharpens the question of whether Loudoun's water system can serve both its residents and its largest industry during summer peaks.