Maryland Ratepayers Face Multi-Billion Dollar Grid Bill for Out-of-State AI Data Centers

The infrastructure cost of AI compute is landing on people who never signed up for it.

Maryland ratepayers are facing multi-billion dollar charges for regional transmission grid upgrades, with a significant portion of that demand attributed to AI data centers operating in neighboring states. The state has filed a complaint with federal energy regulators arguing the cost allocation breaks existing ratepayer protection commitments.

The mechanics here matter. Regional transmission organizations like PJM allocate grid upgrade costs across member utilities based on load-flow models. When a new large load — a hyperscale data center — connects in Virginia or Pennsylvania, the resulting transmission upgrades can be socialized across the broader region. Maryland customers pay a share even if no data center sits within the state.

This is not a new problem. Cost socialization has been contested in PJM proceedings for years. What changed is the scale. AI inference and training facilities draw sustained, high-density power loads that stress transmission capacity in ways that episodic industrial loads do not. The volume of new large-load interconnection requests has accelerated sharply since 2022.

For engineers and founders building on cloud infrastructure, the downstream signal is straightforward: physical compute capacity is increasingly constrained by transmission buildout, not just generation. Data center expansion timelines in the Mid-Atlantic are already slipping due to interconnection queues. That pressure propagates upstream to GPU availability and colocation pricing.

The regulatory fight in Maryland is unlikely to resolve quickly. FERC proceedings move slowly, and the cost allocation formulas embedded in regional tariffs require broad stakeholder consensus to change. In the near term, ratepayers absorb the cost.

The pattern will repeat in other states as AI load growth continues to outpace planned grid capacity. Builders who depend on East Coast infrastructure should treat interconnection constraints as a first-order variable in deployment planning, not a background condition.

Source