Across a 13-state stretch of the eastern United States, surging data center construction is colliding with aging power infrastructure, prompting federal regulators and grid operators to take faster action on reliability and cost questions. Residents in states like Virginia are watching electricity bills climb while grid operators and regulators warn that AI-driven data center growth could strain generation and transmission planning if supply doesn’t keep pace. The tension between tech-driven demand growth and the physical limits of the nation’s electricity supply has reached a critical point, with consequences that extend well beyond server farms and into ordinary households.
Data Centers Are Devouring the Grid’s Spare Capacity
The numbers tell a stark story about how quickly the power balance is shifting. Data centers consumed approximately 4.4% of total U.S. electricity in 2023, according to a Department of Energy report evaluating the increase in electricity demand from these facilities. That analysis projects consumption will jump to between 6.7% and 12% by 2028, driven largely by the computational demands of AI training and inference. A separate projection estimates data centers could consume as much as 17% of U.S. electricity by 2030, a figure that would represent a roughly fourfold increase from current levels in under a decade.
The strain is especially acute in Virginia, the country’s largest data center market, where these facilities could eventually draw up to 57% of the state’s electricity. That concentration creates a fragile dependency: when multiple data centers disconnect from the grid simultaneously during a fault event, the sudden swing in load can destabilize the broader system. Grid engineers and operators have warned that this kind of sudden load swing is an operational risk they are actively trying to manage, alongside the familiar challenge of keeping the lights on during heat waves and winter storms.
PJM’s 13-State Region Faces a Supply Crunch
PJM Interconnection, the grid operator responsible for coordinating electricity across 13 states and the District of Columbia, sits at the center of this crisis. Increasing demand from the tech industry threatens to max out generation capacity in PJM’s territory, and the problem is compounded by the retirement of older coal and gas plants that have not been replaced quickly enough. The result is a tightening margin between available supply and peak demand that leaves little room for error during extreme weather or unexpected equipment failures. For utilities, that means higher costs to secure backup power and upgrade transmission lines; for customers, it means rising rates and a greater risk of outages even as power-hungry tech campuses continue to expand.
The Federal Energy Regulatory Commission has approved steps by PJM to speed up parts of its reliability process in the mid-Atlantic portion of the grid, a move tied to concerns about tightening supply margins. That decision has drawn complaints from clean energy advocates who argue the expedited process disproportionately benefits natural gas projects over renewable alternatives. The tension reflects a deeper policy conflict: regulators are caught between the immediate need to add generation capacity and the longer-term goal of decarbonizing the electricity supply. For residents already facing rate increases tied to grid upgrades, the choice of fuel source is not abstract. It determines both the cost and the environmental footprint of the power they will be paying for over the next several decades.
FERC Tackles the Co-Location Question
One of the most consequential regulatory battles is playing out around a practice known as co-location, in which large electricity consumers like data centers connect directly to or near power generation facilities. FERC formally instituted a proceeding focused on co-location arrangements in PJM’s territory, linking the inquiry to reliability concerns and questions about how costs should be distributed among ratepayers. The core worry is straightforward: if a data center locks up a significant share of a power plant’s output through a direct connection, less electricity flows through the shared grid, and remaining customers bear a larger share of transmission and maintenance costs. That dynamic could undermine the traditional social compact of regulated utilities, in which all users help pay for a robust, resilient system.
Federal regulators are weighing how to handle proposals that would connect data centers directly to power plants, but the policy stakes extend far beyond engineering logistics. The proceeding must answer who pays for grid reliability when major loads bypass the shared system, who gets priority access to generation during shortages, and how oversight keeps pace with deals struck between individual generators and corporate buyers. These are questions with direct financial consequences for households. If co-location agreements shift costs onto residential and small-business ratepayers without corresponding benefits, the arrangement amounts to a subsidy flowing from ordinary electricity customers to some of the world’s most profitable technology companies.
Grid Reliability Faces Growing Strain
The demand surge from data centers arrives at a moment when the broader grid is already under stress. The reliability of the electric systems powering the United States and Canada is trending in the wrong direction, with major winter storms exposing weaknesses in aging infrastructure, fuel supply chains, and planning assumptions. Analysts warn that climate change is intensifying both heat waves and cold snaps, increasing the likelihood that extreme weather will collide with tight capacity margins in regions like PJM. When that happens, grid operators are forced into last-minute measures such as emergency imports, voltage reductions, or controlled outages to keep the system from cascading into a wider blackout.
Layered on top of these stresses is the sheer speed of AI-driven load growth, which makes long-term planning more uncertain. Traditional industrial facilities often take years to build and ramp up, but data center developers can move faster, locking in large blocks of demand with relatively short notice. That volatility complicates the work of regulators and utilities who must decide when to approve new plants, transmission lines, and efficiency programs. If they underestimate demand, reliability suffers; if they overshoot, customers are left paying for underused infrastructure. The result is a more fragile equilibrium in which the margin for forecasting error continues to shrink.
Searching for Solutions in Technology and Policy
Policymakers and engineers are increasingly looking to innovation to square the circle of rising demand and finite infrastructure. Within the Department of Energy, programs overseen by the Advanced Research Projects Agency-Energy are directing funds toward next-generation grid technologies that could make transmission networks more flexible and efficient. These efforts include advanced power electronics, better forecasting tools, and hardware designed to integrate large volumes of variable renewable energy without sacrificing stability. While such projects will not solve PJM’s immediate capacity crunch, they point toward a future grid that can accommodate data-intensive industries with less reliance on fossil fuel peaker plants.
At the same time, federal initiatives like the DOE’s Genesis platform aim to streamline the siting of new energy projects by providing data on land use, environmental constraints, and grid interconnection points. By helping developers identify locations where new generation or transmission can be added with minimal conflict and maximum benefit, tools like this could reduce the long delays that currently plague clean energy buildout. In combination with more stringent efficiency standards for servers, cooling systems, and building design, such policies could buy the grid valuable breathing room. But absent clearer rules about how costs and risks are shared among data center operators, utilities, and ordinary customers, the scramble to power the AI boom will continue to test both the resilience of the system and the patience of the people who depend on it.
More from Morning Overview
*This article was researched with the help of AI, with human editors creating the final content.
