When NextEra Energy announced on May 18 that it would acquire Dominion Energy in an all-stock deal valued at roughly $67 billion, the headline number obscured what the deal was actually buying. NextEra is not paying a premium for poles, wires, or even Dominion's offshore wind project. It is paying for the one input the artificial intelligence industry can no longer source on open markets: dispatchable electricity, physically close to data center cores, with the regulated right to deliver it. For the 3.6 million electric customers Dominion serves in Virginia, North Carolina, and South Carolina, the question is simpler. "Ratepayers are all an afterthought," said Ari Peskoe, director of the Electricity Law Initiative at Harvard Law School.
The combined company, with a $249 billion market capitalization and $420 billion enterprise value, would become the world's largest regulated electric utility. Dominion brings to the table roughly 51 gigawatts of contracted data center capacity as of March 2026, tied to customers including Amazon, Microsoft, Alphabet, Meta, Equinix, CoreWeave, and CyrusOne. That pipeline has more than tripled from 16.5 gigawatts in July 2023. One gigawatt can power roughly 750,000 homes; Dominion's 51-gigawatt data center pipeline is a grid-scale number, not a conventional corporate sales backlog.
To soften regulatory scrutiny, NextEra preemptively offered $2.25 billion in bill credits to Dominion's customers in Virginia and the Carolinas, spread over two years post-close. The deal is structured as tax-free, with a fixed exchange ratio of 0.8138 NextEra shares per Dominion share plus a $360 million cash payment at closing. Closing is targeted in 12 to 18 months pending approvals from the Federal Energy Regulatory Commission, the Nuclear Regulatory Commission, and state-level commissions in Virginia, North Carolina, and South Carolina.
That is the surface story. The deeper one is about what money can and cannot buy in 2026.
Power Prices in Virginia, Maryland, Pennsylvania Rise 75.5% in One Year
For most of 2024 and early 2025, the dominant scarcity story in AI was about GPU allocation. That story has not disappeared, but it has been joined — and in many cases overtaken — by a harder-to-compress constraint: power delivery. Big Tech is on pace to spend more than $650 billion on AI infrastructure in 2026, yet industry analysis projects 30 to 50 percent of planned 2026 data center capacity will slip to 2028. The marginal constraint is no longer silicon alone. It is also steel, copper, and grain-oriented electrical steel.
The pivot point is the high-voltage transformer. Before 2020, lead times on power transformers ran 24 to 30 months. According to Wood Mackenzie's Q2 2025 industry survey, large power transformers now average 128 weeks and generator step-up units 144 weeks; some orders extend to four years. Demand for generator step-up transformers has grown 274 percent since 2019; substation power transformers are up 116 percent. Even with roughly $1.8 billion in announced North American manufacturing expansions, analysts expect transformer supply to remain tight into the end of the decade.
Grid interconnection is the second choke point. More than 2,000 gigawatts of US generation and storage capacity were in interconnection queues at the end of 2025, according to Lawrence Berkeley National Laboratory — a figure that surpasses the country's existing generation capacity. New grid connections take three to seven years. None of this is fixable with capital alone. A hyperscaler with $50 billion in free cash flow cannot order a 765-kV substation the way it orders Hopper or Blackwell GPUs. There are only a handful of suppliers globally — Hitachi Energy, Siemens Energy, GE Vernova, Hyundai Electric — and most large-power-transformer slots through 2028 are already spoken for.
The price signal is unambiguous. Monitoring Analytics, the independent monitor for PJM Interconnection — the grid covering Virginia, Maryland, Pennsylvania, and 10 other states — reported on May 15 that wholesale electricity costs in Q1 2026 surged 75.5 percent year-on-year, from $77.78 to $136.53 per megawatt-hour. The monitor's report identified data center load growth as "the primary reason for recent and expected capacity market conditions." PJM capacity auction prices have moved in the same direction even more violently: from $28.92/MW-day for the 2024/2025 delivery year to $269.92 for 2025/2026, then to the FERC-approved cap of $329.17 for 2026/2027 and $333.44 for 2027/2028. The December 2025 auction for the 2027/2028 delivery year fell 6,625 megawatts short of reliability requirements for the first time in PJM's history.
"The price impacts on customers have been very large and are not reversible," Monitoring Analytics stated in the May 15 report.
Nuclear Deals Signal Hyperscalers Seek Power Sources Outside the Shared Grid
Faced with a grid that cannot grow fast enough and gas plants slowed by their own equipment queues, hyperscalers have made an unusual move: they have effectively become anchor financiers for new nuclear generation. The pattern that formed over the past 18 months is consistent.
Microsoft signed a 20-year power purchase agreement with Constellation Energy to support the restart of Three Mile Island Unit 1 — now renamed the Crane Clean Energy Center — which will add 835 megawatts of carbon-free output. Constellation has targeted 2028 for the restart, pending Nuclear Regulatory Commission approval. Amazon invested in X-energy, a small modular reactor developer, as part of an agreement to bring more than five gigawatts of new projects online across the US by 2039, and in June 2025 expanded its nuclear offtake agreement with Talen Energy to 1,920 megawatts through 2042. Google signed a 25-year power purchase agreement with NextEra Energy in October 2025 to restart Iowa's Duane Arnold Energy Center — a 615-megawatt nuclear plant shuttered in 2020 — with restart targeted for the first quarter of 2029. Meta has procured existing nuclear output from Constellation and Vistra and issued a request for proposals seeking one to four gigawatts of new nuclear capacity. Oracle has announced plans for a gigawatt-scale data center powered by three small modular reactors.
Trackers put the collective Big Tech commitment at roughly 9.8 gigawatts of nuclear capacity earmarked for AI infrastructure, though many of those projects remain at the request-for-proposals, planning, or pre-commercial stage. Nuclear is attractive because it offers what AI training clusters require but the rest of the grid cannot reliably guarantee: 24/7 carbon-free baseload power at gigawatt scale, behind a meter that hyperscalers can effectively control. The trade-off is timeline: small modular reactors remain at least five years from US commercial operation.
That is the gap NextEra-Dominion is built to fill. The merged entity arrives with NextEra's nationwide development platform — including its Duane Arnold restart in Iowa already contracted to Google — bolted onto Dominion's Virginia regulated franchise. The combined construction backlog of 130 gigawatts exceeds the two companies' current generation. That is the asset hyperscalers cannot replicate at any price: a vertically integrated developer that can site, permit, build, and interconnect generation inside a regulated rate base, in the geographies where compute already sits.
NextEra Track Record on Ratepayer Impact Fuels Regulator Concern
What NextEra is buying is not assets. It is queue position — physical, regulatory, and political. That concentration concerns the people who will pay for it.
"This deal would hand control of Virginia's electric grid to a company with a deeply troubling track record," said Brennan Gilmore, executive director of Clean Virginia, a ratepayer advocacy group. Gilmore's statement pointed specifically to NextEra's record in Florida, where rate increases and lobbying conduct have drawn scrutiny. Virginia State Sen. Danica Roem, whose district includes Prince William County, was blunter: "How on earth will it benefit the ratepayers?" Sen. Richard Blumenthal of Connecticut, a member of the Senate Judiciary Subcommittee on Antitrust, Competition Policy and Consumer Rights, said he has "serious doubts about the anti-competitive effects" and intends to scrutinize the deal.
The companies' track record gives those doubts historical grounding. NextEra's attempted acquisition of a Texas utility was blocked by state regulators who found it "would subject Oncor ratepayers to new and potentially substantial risks." Its South Carolina bid was withdrawn after state lawmakers passed legislation to obtain documents related to the company's lobbying and campaign contributions. Its attempted acquisition of Hawaiian Electric was rejected by Hawaii regulators in 2016.
Morningstar equity analyst Andrew Bischof acknowledged the deal makes financial sense for both companies. But Harvard Law's Peskoe offered the structural critique that no bill credit can address: "For the Dominion shareholders, they are selling their shares at a premium. The executives are getting massive payouts for facilitating this, assuming it all goes through, and obviously NextEra believes the transaction is going to add value to the company. Ratepayers are all an afterthought."
NextEra CEO John Ketchum has argued the opposite — that scale is the only mechanism by which the country can build the grid fast enough to serve customers affordably. "Scale matters more than ever — not for the sake of size, but because scale translates into capital and operating efficiencies," Ketchum said in the merger announcement.
Regulators in Virginia, North Carolina, South Carolina, FERC, and the NRC will need 12 to 18 months to decide whether one company should control that much of the country's AI power supply. The Constellation-Calpine merger required divestitures of gas plants in Pennsylvania and Texas before clearing regulators; NextEra-Dominion will face similar pressure, particularly given that PJM ratepayers are already absorbing roughly $13 billion in additional capacity costs attributable to data center load.
The deal's defenders will argue that only a utility at this scale can build fast enough to meet demand without pushing residential bills higher. Its critics will argue the same scale is what makes the asymmetry dangerous: the more compute demand on a single regulated franchise, the more the costs of that demand are socialized across customers who never contracted for a data center.
Either way, the message of the merger is unambiguous. The next chapter of the AI race will not be decided in Santa Clara. It will be decided in Richmond, Harrisburg, and inside every substation queue between them.
ⓒ 2026 TECHTIMES.com All rights reserved. Do not reproduce without permission.
