Data center underwriting is now a county-level power game

The National Association of Realtors has put a figure on what every data center site selector already knows from the field: ten U.S. counties hold 42 percent of the nation's data centers, while 92 percent of counties have none at all, and that imbalance now dictates the shape of every private real estate deal.

The NAR 2026 Data Center Impact Report, first reported by Commercial Observer, drew on county, state, and city data across 3,200 U.S. counties. Loudoun County and Prince William County in Northern Virginia account for 19 percent of all U.S. data centers on their own; Silicon Valley and two central Ohio counties, Franklin and Licking, add another 10 percent, putting 29 percent of national supply in a handful of geographies with different land values, power grids, and local politics.

An investor who treats data centers as a single asset class is implicitly betting that a project in Northern Virginia, one in Ohio, and one in California face the same risks. The NAR map suggests each county is a separate market with its own power capacity, permitting climate, and land price; the underwriting mistake would be to average those into one number.

The economic data make the local variation concrete. In counties with more than ten data centers, the median home value is $431,750, against $174,500 in counties with none; home values in those counties rose 95 percent between 2014 and 2024; and median incomes reached $89,000, compared with $64,000 in areas without data centers. Employment in those same counties grew 16 percent over that decade, against 2 percent where data centers are absent.

NAR is careful not to convert those correlations into simple cause and effect. "There is no single data center effect," Lawrence Yun, the association's chief economist, said. "Instead, the story varies significantly depending on the local market." He warned that "the number of data centers alone does not tell us what will happen to home values, jobs or utility costs," and the report's own text notes that counties with more data centers have higher incomes, younger populations, and more college-educated residents — all factors that push housing demand on their own.

Agents on the ground feel the trade-offs. In an NAR survey of 2,357 commercial and residential real estate agents, 22 percent reported negative effects from data center development while 25 percent reported positive effects; the rest, less than a majority on either side, see a mixed picture, which may explain why local opposition and support can both be loud in the same county, and that political friction belongs in the underwriting.

For a private capital investor, the 95 percent home-price appreciation in counties with more than ten data centers carries a double lesson: it points to a wealthy, educated local market with strong demand for space, but it also means land has already repriced for the data center premium, so the next site carries higher property tax assessments and a more expensive land basis than the facilities that generated that appreciation. The swing variable shifts to power acquisition — how fast a developer can get capacity, and at what cost?

The concentration shapes the exit as much as the entry. If the asset sits in one of the ten counties that dominate national supply, an owner can expect a buyer pool that understands the local grid and land market; in counties with no data centers, there is no established resale market to price against, and an investor who builds the first facility is underwriting an exit that has never happened there.

The energization calendar is now the cap rate in data center underwriting, and the NAR data make that point from the opposite side. The 92 percent of U.S. counties with no data centers are not blank sandboxes waiting for development; most lack the grid capacity to serve a hyperscale load, or the political consensus to build the infrastructure that brings one. A cheap land price in such a county does not create an entry point; it warns that the project will carry the cost of energization itself and years of permitting uncertainty.

The report's practical value for allocators is the map, which tells investors where data center supply has already concentrated and, by inference, where the power and land constraints are bound to bind first. The next winning strategy will be a portfolio of county-level positions, each underwritten against its own local power queue, land assessment, and permitting politics, rather than a national absorption forecast. Ten counties now hold 42 percent of supply; the rest will be priced one energization queue at a time.