When Maryland’s Office of People’s Counsel published its breakdown of rising electricity rates earlier this year, the culprit it identified was not extreme weather, not aging power plants, and not anything Maryland households had done. It was data centers, most of them located in other states, whose explosive growth is pushing up the wholesale capacity prices that every utility customer in the region must help cover.
Across the 13-state territory managed by PJM Interconnection, the regional grid operator coordinating power delivery from New Jersey to Illinois, the cost of reserving enough electricity to meet surging demand is climbing fast. Under PJM’s wholesale capacity market rules, those costs don’t stay in the states where the data centers sit. They spread to every utility in the footprint, reaching residential customers through supply charges, capacity riders, and base-rate adjustments. A retiree in Hagerstown, Maryland, who has never used a cloud computing service is helping fund grid expansions driven by AI server farms in Loudoun County, Virginia, the densest data center market on Earth.
As of mid-2026, this dynamic represents one of the most consequential shifts in American electricity cost allocation in at least two decades, according to consumer advocates and state regulators grappling with the fallout.
A two-decade flatline, then a spike
For most of the 21st century, U.S. electricity consumption barely moved. Between 2005 and 2019, total demand grew at roughly 0.1 percent per year, a figure consistent with historical load data published in the EIA’s Monthly Energy Review. Efficient lighting, better appliances, and tighter building codes held consumption in check, and many utilities deferred building new power plants altogether.
That world ended abruptly. U.S. Energy Information Administration data shows annual demand growth accelerating sharply starting around 2020, with the agency attributing much of the increase to data-center-heavy regions, particularly PJM’s territory. The EIA has since projected the strongest stretch of electricity demand growth since at least 2000, naming data centers as a primary driver.
The International Energy Agency’s Electricity 2026 report reinforces that outlook, forecasting sustained U.S. consumption growth through the end of the decade, with large-scale computing loads singled out as a defining factor.
In practical terms, the country has shifted from a planning environment where utilities could defer investment to one where they are racing to build generation and transmission fast enough to keep the lights on while feeding clusters of facilities that each draw as much power as a small city.
How data center costs reach your kitchen table
The mechanics are straightforward, even if the bills they produce are not.
PJM runs a capacity market designed to guarantee that enough power plants exist to meet demand three years into the future. Every year, the grid operator holds a Base Residual Auction: generators bid to supply capacity, and the market clears at a single price reflecting how tight supply is relative to projected load. When data centers file massive new interconnection requests, PJM’s demand forecast rises. More generation must be procured. The auction clearing price climbs. Utilities across the entire PJM footprint then pass their share of that higher price to customers, typically through line items on monthly bills labeled as capacity or supply charges.
PJM’s most recent Base Residual Auction results showed clearing prices that were multiples of levels seen just a few years earlier, a spike that grid analysts and consumer advocates have tied directly to the surge in projected data center load. Maryland’s Office of People’s Counsel has explicitly identified data centers, most located outside the state, as a force pushing up the capacity prices Maryland households must cover.
This cost-sharing structure is not new. PJM’s capacity market has worked this way for years. What changed is the sheer scale of demand being added by a single industry in a compressed window, which amplifies the price signal far beyond what incremental residential or commercial growth would produce.
The states feeling it first
Maryland is the most publicly documented case, but it is not alone. Utility commissions in Virginia, Ohio, Indiana, and other PJM states are reviewing rate filings that cite rising wholesale capacity and transmission costs. In Virginia, where Dominion Energy serves the epicenter of U.S. data center construction, regulators have faced pointed questions about how infrastructure investments driven by hyperscale tenants should be split between those commercial customers and the broader residential base.
The challenge for regulators is that rate cases rarely isolate a single cost driver. A typical filing bundles fuel expenses, storm recovery, grid modernization, and new generation into one request. Extracting the precise share attributable to data centers requires combing through testimony and workpapers that are dense even by regulatory standards. No single public database yet compiles approved or pending residential rate increases linked specifically to data center demand across all PJM states.
Consumer advocates say that opacity works against households. If the cost driver is invisible on the bill, it becomes harder for ratepayers to push back through the regulatory process or demand that technology companies shoulder a larger share of the infrastructure tab.
What tech companies are and aren’t paying
Amazon Web Services, Microsoft, Google, and Meta have each announced data center capital spending plans measured in tens of billions of dollars, and alongside them, large renewable energy procurement deals. These companies regularly highlight their commitments to carbon-free electricity and direct power purchase agreements with wind and solar developers.
What they do not publicly disclose is how much of the regional grid-upgrade spending their facilities require will be covered through direct contracts versus recovered from general ratepayers. A data center operator may sign a power purchase agreement for solar output, but the transmission lines, substations, and peaking plants needed to integrate that facility into the broader grid are often socialized across all customers.
Data center operators also point out, with some justification, that their facilities generate local tax revenue, construction jobs, and long-term employment in host communities. In Loudoun County, data center property tax revenue has helped keep residential property tax rates among the lowest in the Washington, D.C., metro area. But those local fiscal benefits do not flow to ratepayers in other states who are absorbing higher capacity costs through PJM’s regional market.
Without transparent reporting from both utilities and technology companies, the precise split between corporate and residential cost responsibility remains impossible to independently verify. This is the central gap in the public record, and it is one that state regulators, federal agencies, and consumer groups are only beginning to press on.
Why household bills will keep climbing alongside data center load
Several forces suggest the cost pressure on households will intensify before it eases.
PJM’s interconnection queue, the pipeline of projects seeking to connect to the grid, contains data center requests totaling tens of gigawatts, a volume that dwarfs anything in the operator’s history. New generation takes years to permit and build, which means the supply side will lag demand growth for the foreseeable future, keeping capacity auction prices elevated.
At the same time, policymakers are starting to respond. Some state legislators have proposed requiring data center operators to fund a larger share of grid upgrades directly, rather than allowing those costs to flow through to all ratepayers. The Federal Energy Regulatory Commission has been examining transmission cost allocation through ongoing rulemaking proceedings, including questions about how expenses should be assigned when a small number of very large customers drive the need for new infrastructure. As of June 2026, FERC has not published a final rule, and the relevant docket numbers and hearing dates have not been consolidated into a single public notice.
There is also the question of generation mix. Several companies have explored restarting retired or idled nuclear plants to serve data center load, including discussions around the Three Mile Island Unit 1 facility in Pennsylvania. If those restarts proceed, the costs and benefits of keeping aging nuclear capacity online will add another layer to the rate-setting debate.
For now, the bills are arriving. The EIA’s demand data, PJM’s auction results, and Maryland’s consumer-advocacy filings together trace a clear chain: AI-driven data center growth is pushing up regional electricity costs, and residential customers are absorbing a real share of the increase. The exact dollar amounts vary by utility, by state, and by season, and no single public source yet aggregates the per-household impact across the PJM footprint. But the direction is consistent, and every credible forecast points the same way.
