AI data centers are changing the price formation problem for Mid-Atlantic electricity markets. The shift is most visible in PJM Interconnection, where concentrated load growth from large computing facilities has collided with slower-moving generation, transmission, and capacity planning cycles.
The evidence does not support a simple claim that computing load alone explains every price move. Fuel costs, generator retirements, transmission limits, weather, capacity auction rules, and local demand patterns all affect wholesale prices. The data center effect is still material because it changes forward load forecasts, raises the value of firm supply, and concentrates demand in a few grid zones rather than spreading it evenly across the region.
For related coverage on the same regional pressure point, Abacus News has analyzed how large computing loads are affecting Mid-Atlantic power planning. Abacus News and a related site, Natewin, are both part of the same network.
Why AI data centers Are Moving PJM Prices
AI data centers In The Price Signal
PJM operates wholesale markets across 13 Mid-Atlantic and Midwest states and the District of Columbia. The market has to match supply and demand in real time, while also procuring enough future capacity to meet reliability needs. Large new loads matter because electricity cannot be delivered at scale without generation availability, transmission capacity, and distribution upgrades in the right locations.
The U.S. Energy Information Administration projected that, under a high electricity demand growth case, PJM’s wholesale price for 2027 would be $2.60 per megawatt-hour higher than the agency’s February 2026 forecast, about a 4% increase. The EIA linked that change largely to higher data center power demand in its analysis of data center growth and fossil generation risk EIA data center demand analysis.
That number should be read carefully. A $2.60/MWh increase is not the same as a retail bill increase, and it is a forecast scenario rather than a measured invoice. Still, it shows that data center load has become large enough to alter federal price modeling for one of the country’s most important power markets.
The Local Concentration Problem
The effect of AI data centers is not evenly distributed. Northern Virginia has long been a major data center hub, and the research record provided for this article points to stronger price pressure there than in the national average. A Dallas Fed working paper revised in October 2026 found that by 2025, data centers had raised wholesale electricity prices by about 6.9% nationally and by 15.0% in northern Virginia.
Local concentration changes the technical problem. A megawatt of demand added near available transmission and generation has a different cost profile from a megawatt added in a constrained zone. If a cluster of large users requests firm service in the same area, planners may need new substations, high-voltage lines, reactive support, or local reliability upgrades before that load can be served without raising reliability risk.
Energy, Capacity, And Transmission Costs
Wholesale Energy Is Only One Layer
Wholesale energy prices reflect the cost of serving electricity demand at a given time and location. The research notes show that between the first quarter of 2025 and the first quarter of 2026, PJM-region wholesale power prices rose from about $77.78/MWh to $136.53/MWh, a nearly 76% year-over-year increase. The notes attribute that move primarily to growing energy demand from data centers.
That figure captures a sharp market change, but it should not be treated as a clean laboratory measurement. Weather, fuel prices, outages, and generator availability can affect quarterly comparisons. The stronger analytical point is that large computing loads have shifted the demand baseline used by system planners and market participants. Once that baseline moves, both spot prices and forward expectations can change.
Maryland’s Office of People’s Counsel reported that from January through September 2025, PJM energy prices grew almost 50% from the same period a year earlier, with the fastest growth in data center hot spots such as Baltimore. The office also stated that between 2024 and 2025, PJM advanced roughly $12 billion in new transmission infrastructure, mainly driven by service requests for data centers in Northern Virginia, with costs expected to appear on Maryland customers’ supply bills Maryland data center cost review.
Capacity Prices Reflect Future Reliability Risk
Capacity markets are designed to pay resources for being available in future delivery years. When forecasted demand rises faster than dependable supply, capacity prices can increase even before all the new load is connected. That is why data center forecasts can affect bills ahead of physical energization.
The research notes state that PJM’s 2026/27 capacity auction price rose from around $28.92 per megawatt-day in 2024/25 to $329.17 per megawatt-day. They also cite PJM’s independent market monitor as saying data center demand has been the primary reason for recent and expected capacity market conditions, adding about $23 billion to capacity market costs across the last three auctions.
For customers, the distinction matters. Energy market costs reflect electricity consumed. Capacity costs reflect the system’s obligation to maintain enough dependable supply for peak and stressed conditions. A data center that runs with high utilization can affect both categories: it consumes large amounts of energy and raises the peak-demand planning target if its load is treated as firm.
Market Design Is Under Pressure
Interconnection Rules Face A Large-Load Test
PJM’s existing market design was not built around sudden, multi-gigawatt growth from a small number of very large electricity users. By May 6, 2026, PJM was evaluating changes to how electricity is bought and sold because of risks that data center demand could outstrip supply and challenge reliability across its footprint, according to the research notes.
Federal regulators also moved toward closer scrutiny. On June 18, 2026, the Federal Energy Regulatory Commission issued a draft show-cause order directing grid operators outside Texas to justify or update processes for connecting very large energy users such as data centers. The stated concern was that connection practices could raise costs or reliability risk if they did not reflect the scale and speed of new load requests.
These reforms are not only legal or administrative issues. They determine who pays for upgrades, how much certainty large customers must provide before reserving capacity, and whether other consumers are protected from costs linked to speculative or delayed projects. A stricter queue process could reduce phantom demand in forecasts, but it may also slow legitimate projects that have financing and construction plans in place.
Operational Flexibility Remains Limited
AI data centers can, in theory, shift some workloads across time or geography. In practice, the amount of flexible load depends on service-level requirements, latency tolerance, contractual obligations, cooling systems, backup power design, and the type of computation being performed. Training jobs may be more movable than latency-sensitive inference, but the research notes do not provide enough evidence to quantify how much PJM load could be curtailed or shifted during stress events.
That uncertainty makes planning harder. If grid operators assume data centers are fully firm loads, they may procure more capacity and transmission. If they assume too much flexibility and the load does not respond during tight conditions, reliability risk rises. The policy challenge is to create tariffs and interconnection terms that reflect actual operating behavior rather than marketing claims.
Who Bears The Cost Shift

Retail Customers See Effects Indirectly
Wholesale price increases do not pass through to households and businesses in a uniform way. Retail rates depend on state regulation, utility procurement, fixed contracts, capacity charges, transmission allocations, and rate design. Still, wholesale energy, capacity, and transmission costs can flow into supply bills over time.
Maryland is a useful case because the Office of People’s Counsel has explicitly linked data center-driven transmission upgrades to customer bills. The office’s position, as reflected in the research notes, is that requests for data center service in Northern Virginia can create costs that are allocated beyond the immediate location of the new load. That is a central fairness issue for regional transmission organizations: benefits and costs do not always sit in the same jurisdiction.
Generators And Utilities Face Planning Risk
Generators may see stronger price signals for new supply, but development timelines are long. Transmission projects can also take years, particularly when siting, permitting, and regional cost allocation are contested. Utilities must serve new load requests while maintaining reliability for existing customers, yet load forecasts tied to data center projects can change if financing, chip supply, or customer demand shifts.
This creates a timing mismatch. Computing demand can appear in forecasts before the grid assets needed to serve it are complete. If demand materializes faster than supply, prices rise and reliability margins narrow. If demand is overestimated, customers may be left paying for assets that were sized for load that did not arrive as expected.
AI data centers And Mid-Atlantic Price Risk
The supported evidence points to a measured but significant conclusion: large computing loads are now a price-setting factor in the Mid-Atlantic, not a side issue. EIA modeling shows higher PJM wholesale price projections under faster demand growth. Maryland’s consumer advocate data shows sharp 2025 energy price increases in PJM and large transmission costs tied to data center service requests. Other research cited in the provided notes indicates stronger impacts in northern Virginia than in the national average.
The next technical question is not whether data centers use large amounts of power. They do. The harder question is how much of that demand should be treated as firm, how upgrade costs should be assigned, and what market rules can separate committed projects from speculative load requests.
AI data centers will remain a planning challenge for PJM if load growth stays concentrated in constrained zones. Better forecasting, clearer interconnection deposits, flexible-load tariffs, and transparent cost allocation can reduce risk. None of those measures removes the need for generation and transmission investment. They can, however, make the price signal more accurate and reduce the chance that households and smaller businesses absorb costs created by poorly timed or poorly screened large-load additions.






