America’s largest power grid is billing households for data centers that have not been built yet
By Mary Seal
There are fewer sectors of the U.S. economy that appear so confident in the future as the data center sector. It seems every couple of weeks there is a brand-new campus announced somewhere in America using enough electrical power to light up a small town. The country’s grid operators those responsible for forecasting demand over years ahead have accepted the projections of the data center companies. Now, the costs for accepting those projections are appearing on residential electric utility bills.
American data centers consume approximately 4.4 percent of all electricity produced in the United States in 2023, and the amount could be as high as 12 percent in 2028. Much of the growth is taking place in a few key areas, which is part of the reason it hurts. Currently, Virginia has the largest number of data centers, 498, of any state in the country, followed closely by Texas, with 407. Residents living near these developments are starting to feel the impact. According to a Consumer Reports survey conducted in November 2025, seventy-eight percent (78%) of Americans reported being “somewhat” or “very” concerned about increased energy rates due to new data centers.
The data illustrating who pays for the growth is provided by PJM Interconnect, the grid operator for thirteen states and Washington D.C. PJM does not merely purchase electricity; PJM also conducts an auction to ensure sufficient generating capacity exists well into the future. Customers ultimately bear the burden of whatever PJM determines to clear during the auction. During PJM’s most recently completed auction, data centers represented forty percent ($6.3 billion) of the $16.4 billion in capacity charges awarded through the auction process. According to PJM Market Monitor, since 2011, PJM has held fourteen auctions to determine generating capacity requirements and the resulting capacity charges. Since then, data-center driven capacity charges totaled $29.4 billion, representing nearly fifty percent (46%) of the total $63.6 billion charged to consumers during the same time period.
The contentious nature of this issue stems from the fact that much of the projected load does not currently exist. PJM Market Monitor determined that forty-five percent (45%) of the $47.2 billion in capacity costs assessed to PJM’s customers from capacity costs incurred in three recent auctions were based upon forecasted data-center loads beyond current levels of consumption. The monitor stated that this level of uncertainty is unprecedented and raises significant questions regarding how much an auction clearing price represents if that clearing price is based on such uncertain forecasts. In essence, households are underwriting the data-center industry’s optimistic view of its own growth. When a data-center project is canceled or when a campus is constructed at a smaller size than initially forecasted by developers, the capacity purchased on behalf of that campus is still borne by PJM’s customers regardless.
Data-center supporters deserve to be heard as well. The numbers referenced above represent PJM Market Monitor estimates of what would have occurred had data-center loads been removed from PJM’s auction processes. PJM emphasizes that its auction process cost does not directly translate to total cost to customers. Furthermore, while headlines may suggest otherwise, the percentage of PJM’s total auction cost attributable to data-center loads is lower than suggested. In the first six months of 2026, wholesale costs averaged $114.50 per megawatt hour (MW-h), with data-center loads representing nine point seven-seven percent (9.77%) of those costs attributable to capacity market payments. Moreover, the large increase in wholesale costs throughout 2025 is attributed to multiple factors. A study released by Columbia Center on Global Energy Policy in July 2026 indicates that no single factor can explain increases in wholesale costs experienced in 2025; instead, factors including aging grid infrastructure, increasing grid hardening measures, disaster recover efforts, fluctuating fuel prices and mandated policies by government agencies contribute to higher costs.
Both points could be true. A data centre does not necessarily need to be the sole cause of increasing costs in order to be poorly run. The increasingly popular argument is that large-scale consumers will either generate their own power or pay for the strain they put on the grid. PJM’s Board has recommended holding a one-time backstop auction for 6.8 GW, and as the grid approaches emergency levels, curtailing new large loads which do not provide their own energy. Simply stated, PJM’s Board believes that current customers should not have to fund the generation capability created by new large-scale loads, since those new large-scale loads are supplying their own generation. The regulators were less aggressive than PJM. FERC approved PJM’s designation of a “large load” (a site of 50 MW or greater) along with a registration process beginning Oct. 12; however, FERC rejected PJM’s remaining tariff revisions. Therefore, while PJM has survived at least part of what identifies who constitutes a “large user,” the part that determines who pays remains undecided.
In terms of political activity, the matter is relatively clear-cut. Few citizens know anything about capacity auctions, but everyone knows what a bill is. According to a poll conducted by Global Strategy Group and the Chesapeake Climate Action Network Action Fund (the latter being an advocacy organization), approximately three-fourths of Virginians surveyed (where the bulk of the industry resides) attribute the increase in electric utility bills to the existence of the facilities themselves. Thus, politicians who previously competed for the support of data centers through providing tax incentives now find an opportunity to demonstrate that data centers also need to help bear some portion of the burden of increasing rates.
The data center community’s most logical response is to simply pay their share. There are several companies currently signing long term power contracts and the Monitor’s own recommendation for large loads is to enter long term generation contracts. This position moves the debate away from a disagreement regarding responsibility for increased costs toward questions of how the system should be designed. It is theoretically possible to allow the growth of a technology sector using the same grid used to serve residential customers which are not subsidizing such development. However, it is difficult to continue operating under previous assumptions where the cost of the data center industry’s forecasted needs is transferred to individuals who did not agree to finance them.
Image Credit: Unsplash