For the first time in over a decade, Americans used significantly more electricity than the year before. Total U.S. power consumption rose 2% in 2025, according to the International Energy Agency’s Global Energy Review 2026, snapping a long stretch of nearly flat demand. The single biggest reason: data centers, which accounted for roughly half of the entire national increase in electricity consumption.
That lopsided growth is now forcing a reckoning over who picks up the tab. Grid operators need new transmission lines and power plants to serve facilities that can each consume as much electricity as a small city. Residential customers, state regulators, and federal officials are all asking the same question: should ordinary ratepayers subsidize infrastructure built to power the artificial intelligence boom?
A demand spike with a clear return address
The IEA’s global trends analysis puts the figure plainly: data centers were responsible for around 50% of total U.S. electricity demand growth in 2025. No other single category of end use, not electric vehicles, not manufacturing, not residential heating, came close to matching that share. The underlying data is available through the agency’s full report package for independent verification.
The concentration is especially stark in Virginia. The U.S. Energy Information Administration reported that commercial electricity sales have soared in the state, driven by clusters of hyperscale data centers in its northern counties. PJM Interconnection, the regional grid operator covering 13 states and the District of Columbia, expects Dominion Energy’s Virginia zone to post the largest absolute jump in summer peak demand anywhere in its territory, according to the same EIA analysis.
The timing is not a coincidence. The explosion of generative AI services over the past two years has sent computing demand surging. Training and running large language models requires vast arrays of specialized processors that draw far more power per rack than traditional cloud workloads. Tech companies have responded by racing to build new facilities, and the grid is feeling the consequences.
The cost fight moves to federal court
On May 7, 2026, the Maryland Office of People’s Counsel filed a formal complaint against PJM Interconnection at the Federal Energy Regulatory Commission, listed under docket EL26-63-000. The filing challenges how PJM allocates the cost of transmission upgrades, arguing that residential customers should not bear expenses created by large commercial loads they had no part in requesting.
“Marylanders did not ask for this wave of data center construction, and they should not be forced to finance the grid expansions it requires,” Maryland People’s Counsel David Lapp said in a statement accompanying the filing. The complaint does not yet include specific dollar figures, a list of contested transmission projects, or a proposed timeline for resolution. But its existence marks a concrete escalation. For months, the debate over data center cost allocation played out at industry conferences and in utility planning meetings. Now it sits before a federal regulator with the authority to issue binding orders.
Other states inside PJM’s footprint that host fast-growing data center clusters have not filed similar complaints as of early June 2026. Whether they follow Maryland’s lead, or whether FERC uses the docket as a springboard for broader rulemaking, will shape the financial landscape for millions of ratepayers across the mid-Atlantic and Midwest.
What the data shows, and where the gaps remain
The IEA and EIA figures represent the strongest available evidence because they rest on metered electricity consumption and official utility filings, not models or surveys. When the IEA says demand grew 2%, that number reflects actual power flowing through the grid.
But important gaps persist. No primary dataset currently breaks out the exact gigawatt-hour contributions from manufacturing reshoring, EV charging, and heat pump adoption alongside data centers for 2025. The IEA attributes roughly half the growth to data centers; the composition of the other half is described only in broad terms. Until the EIA or another agency publishes a full sectoral decomposition, the relative weight of competing drivers remains an estimate.
The Department of Energy, through Lawrence Berkeley National Laboratory, maintains a dedicated data center energy tracking portal with facility-level measurement data and historical trends. LBNL’s published series has not yet been updated with final 2025 figures, so it cannot independently confirm the IEA’s growth share. It does, however, document the long-run trajectory: computing intensity has risen steadily even as individual facilities have become more efficient.
That efficiency question matters. Data center operators continue deploying advanced cooling systems and more power-efficient server hardware. Some of the newest facilities achieve lower power-usage effectiveness scores than their predecessors. Whether those gains can keep pace with the sheer volume of AI workloads is an open question no existing dataset answers conclusively.
Public opinion is split
A Pew Research Center survey published in March 2026 found that Americans hold mixed views on data centers’ local effects, weighing job creation and economic investment against environmental concerns and strain on local resources. The survey captures a national mood, but it lacks state-level breakdowns for Virginia or other high-growth zones, making it hard to know whether residents living next to these facilities feel differently from the country at large.
That ambivalence could shape the political calculus for state legislators and utility commissioners who must decide how aggressively to protect residential ratepayers. Public support for data centers is not guaranteed, especially if electricity bills start climbing in ways voters can trace directly to new facilities.
Rate pressure is already building in data center corridors
Several developments between now and mid-2027 will determine whether the current tension over data center power demand hardens into a lasting policy fight or eases as the grid adapts.
New data releases from the EIA and DOE will begin filling in the 2025 sectoral picture. A more granular breakdown of commercial, industrial, and residential consumption in states with heavy data center buildout will test whether these facilities are uniquely responsible for recent spikes in peak loads and total sales, or whether other fast-growing categories deserve more attention.
The Maryland complaint at FERC will signal how far federal regulators are willing to go. If the commission opens a broader inquiry, other state consumer advocates could file their own challenges, potentially reshaping how transmission expansion is financed across PJM and in other regions.
State-level decisions in Virginia and neighboring jurisdictions will reveal whether policymakers choose to ring-fence residential customers from some portion of data center-related infrastructure costs. Options on the table include special rate classes for large loads, upfront contribution requirements for new connections, and targeted incentives for on-site generation or efficiency improvements.
And the pace of construction itself bears watching. If higher financing costs, community opposition, or new siting rules slow the pipeline of proposed projects, pressure on regional grids could ease. If demand for AI and cloud services keeps accelerating, the tensions now surfacing in PJM may spread to regions that have, until now, treated data centers as a straightforward economic development win.
For now, the most defensible conclusion is narrow but significant: data centers played an outsized role in pushing U.S. electricity demand higher in 2025, and that growth is colliding with long-standing assumptions about who pays to keep the grid running. The trend is established. The fights over fairness are just getting started.
More from Morning Overview
*This article was researched with the help of AI, with human editors creating the final content.
