How data centers are complicating transmission expansion

The project is the Mid-Atlantic Resiliency Link: about 107 miles, roughly $960 million, developed by NextEra, running from southwestern Pennsylvania to northern Virginia through Maryland and West Virginia. PJM, the regional grid operator, approved it in 2022 as a reliability project addressing a shortage of west-to-east capacity, driven largely by growing data center load in northern Virginia, under the standard model that spreads costs across the region while assigning a bigger share to the zone expected to benefit most. Note the date: ChatGPT had not been released. NextEra then spent years in development and only began filing for state siting approvals in 2025 and this spring. The intuition Kann puts on the table is the obvious one. Demand exploded in exactly the place the line was meant to serve, so everyone should now want it faster and bigger. The reverse happened.

The reason is that regional cost sharing is a reciprocity bargain. Virginia ratepayers help fund a line in New Jersey on the understanding that New Jersey will return the favor, and it evens out because growth is roughly even. Concentrated growth destroys the evening-out. So consumer advocates now argue first that data centers should pay, and failing that, that Virginia ratepayers should carry the bulk. Maryland took the allocation to federal regulators at FERC in 2024 arguing Virginia should pay more, and was rejected. It returned in May 2026 with a broader ask: assign data-center-driven transmission costs to the specific zone, in this case Dominion’s, and let the utility work out internally how to make data centers pay, a version that would apply to future lines too. NextEra faces pushback at all four state commissions, and the March 2026 ratepayer protection pledge the hyperscalers signed at the White House now turns up in filings as leverage.

The deeper problem is that the mechanism everyone points to does not fit transmission. The hyperscaler playbook works when the cost is attributable: negotiate a tariff with the local utility covering the incremental generation and any specific upgrade your load requires, then announce it with community investment and a claim of no detrimental rate impact. A multi-state line serving many customers has no comparable instrument. Kann calls it a tragedy of the commons and cannot see how joint payment would be structured; Allsup says the systems simply do not exist yet and calls it building the plane while flying it. Timing cuts both ways. Construction is targeted for 2029 and operation years later, well past the horizon data center developers plan to, and nobody knows the retroactivity rule. If FERC decides how data centers must cover grid upgrade costs, does that reach a line approved in 2022? Allsup says plainly that she does not know and neither does NextEra, and that the question of what date divides projects presumed to serve data centers from those judged under the old framework keeps coming up in her reporting.

It would be a mistake to read this as data centers alone stopping the line. Asked to separate generalized anti-data-center sentiment from the specific northern Virginia cost question, Allsup answers that it is all of the above. The ordinary siting objections are all present, property values, agriculture, views, pace of life, and opposition runs from individual landowners through state legislators to a state consumer advocate at FERC. She also notes that every data center already running in northern Virginia sits on infrastructure the whole region paid for and nobody objected at the time; what changed is that people are watching. Kann’s counterfactual has the same shape: without the AI wave NextEra would still have needed four state approvals, and states might have made a version of this argument anyway. What the boom added was magnitude, which makes it harder to argue the line does not disproportionately serve data centers, and a generalized backlash for which this project has become a convenient vessel. The conclusion is not that data centers are hard to serve. It is that transmission fails for reasons people mostly do not name: everyone reaches first for NIMBYism, and here NIMBYism is not really the story.