Maryland is turning the AI boom into a ratepayer fight, arguing households should not cover grid upgrades built for data centers elsewhere.
The next phase of artificial intelligence is not just about chips, models and cloud contracts. It is about transmission lines, substations and the electric bills landing in ordinary homes. Maryland has now put that issue in front of federal energy regulators, challenging roughly $2 billion in power grid upgrade costs that state consumer advocates say are being shifted onto residents for data center demand largely outside the state.
According to Maryland's Office of People's Counsel, its complaint at the Federal Energy Regulatory Commission argues that PJM Interconnection's rules unlawfully assign Maryland customers responsibility for $2 billion of $22 billion in regional transmission projects advanced over the last three years. The office says those costs will show up in rates for decades and add about $1.6 billion to Maryland customer bills over the next ten years, including roughly $823 million for residential customers, or about $345 for an average household.
That is the kind of number that changes how people hear the phrase AI infrastructure. For much of the last two years, the story has been framed as a race among Microsoft, Amazon, Google, Meta, OpenAI and others to secure chips, land and power. Maryland's complaint asks a more basic question: if a data center in Virginia, Ohio or Pennsylvania creates the need for new regional transmission, why should a Maryland family help finance it?
PJM runs the regional electric grid across all or parts of 13 states and Washington, D.C., making it one of the most important pieces of machinery behind America's data center buildout. Its cost allocation method for high voltage projects spreads part of the expense across the whole PJM system based on demand, then assigns the rest through a power flow analysis. That sounds technical, but the result is simple enough for customers to understand. Costs can move across state lines even when the economic benefit does not.
Maryland officials argue that this formula no longer fits the moment. Data center growth is not evenly distributed. Northern Virginia remains the dominant hub, and new campuses are also pushing into Ohio, Pennsylvania and other PJM territories where land, interconnection access and tax incentives can make projects attractive. Maryland's projected load growth is more modest, yet its customers are being asked to help pay for transmission planned around a much larger regional surge.
The complaint also points to a political promise that has quickly become a test of policy. In March, major hyperscalers signed the White House's Ratepayer Protection Pledge, a voluntary commitment that data center developers would pay for new power delivery infrastructure needed to serve their facilities so those expenses would not be passed to ordinary households. Maryland's position is that PJM's rules can defeat that promise before it has any practical meaning.
This is where the AI economy starts to look less like a software cycle and more like a regulated utility fight. Transmission projects are capital intensive, slow to build and typically recovered through rates. If the forecasted data center load arrives, utilities and grid operators will say the investment was necessary. If some of those projects are delayed, downsized or canceled, existing customers can still be left paying for infrastructure justified by demand that never fully materialized.
A National Pattern Is Taking Shape
Maryland is not an isolated case. Michigan Attorney General Dana Nessel has challenged DTE Energy's handling of power arrangements tied to a large data center development involving Oracle and OpenAI, arguing that secret reviews and fast approvals do not give customers enough protection. The point is not whether one project should be built. It is whether utilities can make unusually large infrastructure commitments before the public understands who carries the risk.
Utah has seen a different version of the same fight. A proposed massive data center project in Box Elder County backed by Kevin O'Leary has drawn anger over energy use, water, air quality and local control. Governor Spencer Cox responded with tighter state oversight, including phased approvals and fresh scrutiny of permits. That dispute is less about regional transmission math and more about whether rural communities are being asked to absorb the physical footprint of the AI boom while outside developers capture most of the upside.
Georgia has moved the argument directly into state politics. Lawmakers have introduced bills to limit tax breaks, require more disclosure and prevent data center related power costs from being passed to residential customers. Georgia Power's plans to add nearly 10 gigawatts of capacity, largely to serve data center demand, have made electricity affordability a campaign issue. Once customers connect higher bills with server farms, the politics shift fast.
For hyperscalers, this creates a strategic problem that cannot be solved with bigger capital budgets alone. The industry needs enormous amounts of reliable electricity, but every new megawatt now brings questions about who pays, who gets priority and who carries the long-term obligation if projections prove too optimistic. A pledge is useful as a signal, but rate design, interconnection rules and enforceable contracts are what determine whether households are protected.
For regulators, the challenge is to avoid turning AI growth into a hidden subsidy. Data centers can bring tax revenue, construction jobs and digital infrastructure, but they also concentrate demand in ways the old utility playbook was not designed to handle. If the costs are socialized while the benefits are narrow, local opposition will harden. If developers pay directly for the grid they need, the market will get a clearer signal about which projects are truly economic.
Maryland's FERC complaint is therefore bigger than one state's bill. It is an early test of whether America's AI buildout can be financed without quietly loading the cost onto people who never asked for a data center and may never see one nearby. The next thing to watch is whether federal regulators force PJM to assign more costs to the zones and customers creating the demand. That decision could shape not just Maryland power bills, but the price of AI infrastructure across the country.
Also read: AI agents are turning websites into security tests. • xAI's Anthropic deal shows AI alliances are getting harder to read • China is complicating Silicon Valley's favorite AI regulation warning
