North Carolina Regulators Denied Duke Energy a $584 Million Gas Plant. The 3-1 Vote Came Down to a Question About Data Centers Duke Couldn't Answer.

Utility commissions approve gas plants. That is, overwhelmingly, what the job consists of — a utility files for a certificate of public convenience and necessity, staff reviews the load forecast and the cost, and the commission signs off, usually with little drama. On September 18th, North Carolina's Utilities Commission did the opposite. It voted 3-1 to deny Duke Energy Progress a certificate for a 255-megawatt gas-fired combustion turbine at its Smith Energy Complex near Hamlet, in Richmond County — the sixth turbine Duke has proposed for that site, at an estimated cost of $584 million. The complex sits inside the same industrial park as Amazon's $10 billion AWS data center campus, now under construction a short distance away. Duke never managed to tell the commission how much of the new turbine's output, or its cost, belonged to that campus versus everyone else on the system. That gap is what sank the application.

A rare no, and a sharper word than usual

Commissioners Bill Brawley, Tommy Tucker, and Donald van der Vaart made up the majority that denied the certificate; Floyd McKissick dissented, arguing the turbine would provide "critical firm capacity" during extreme weather and other periods of elevated demand — a reasonable argument on its own terms, and the one that usually wins in cases like this. It didn't here. Tucker's concurring opinion zeroed in on the number itself, calling the $584 million price tag "staggering" and "very expensive," language that staff testimony had used first and the commission adopted as its own. A single adjective doesn't normally decide a rate case. In this one, it's the plainest signal of how far outside Duke's usual approval pattern this request fell.

The 70 percent nobody had checked yet

The commission's actual objection was narrower and more procedural than the word "staggering" suggests, and more damaging to Duke's case for it. Roughly 70 percent of the load growth Duke is modeling systemwide comes from data centers — but that figure is still sitting inside Duke's ongoing 2025 Carbon Plan and Integrated Resource Plan proceeding, a separate docket the commission hasn't finished vetting. Commissioners found Duke was asking them to approve a $584 million, standalone generation asset built on a load forecast the same commission hadn't yet signed off on in the case where that forecast actually belongs. Layered on top of that: Duke's application didn't specify how much of the new turbine's 255 megawatts would actually serve data center customers as opposed to the broader system, and didn't propose how the cost would be allocated if the answer turned out to be "most of it." Asking regulators to approve a plant before its own load forecast clears review, without saying who the plant is really for, is a hard ask in any docket. It was a harder one next to a campus the size of Amazon's.

A pledge Duke signed and then couldn't show its work on

The commission's order leaned on something that didn't exist the last time Duke built a turbine at Smith: the Ratepayer Protection Pledge, which the White House rolled out on March 4th and which Duke, along with utilities covering roughly 80 percent of the power delivered to U.S. homes and businesses, signed onto. The pledge commits signatories to have large loads — data centers chief among them — build, bring, or buy their own additional power, pay the full cost of the transmission and distribution infrastructure built to serve them, and accept rate structures that charge them for capacity reserved on their behalf even if they end up using less of it. It's a voluntary commitment, not a statute, which is exactly why the commission's order matters: this is the first time a state regulator has told a signatory, in a live rate case, that signing the pledge isn't the same as demonstrating compliance with it. Duke's application didn't explain how recovering $584 million through general rates would square with a pledge premised on large loads covering their own costs. The commission's denial order made that explanation a condition of any refiling, not a nice-to-have.

Next door to the campus driving the math

The Smith Energy Complex has been running since 2001 and already carries a pair of combined-cycle gas units; the proposed turbine would have been its sixth combustion unit on site. It sits in Rockingham's Energy Way Industrial Park, adjacent to the nearly 800-acre, 20-building AWS campus Amazon broke ground on as the largest single investment in North Carolina history at the time it was announced — $10 billion, roughly 80 miles east of Charlotte, with 500 jobs promised over five years. None of that proximity is incidental to why this case drew attention beyond the usual rate-case audience. But proximity is also exactly the thing Duke's application failed to turn into an actual allocation. A gas plant built next to a data center campus is a plausible story about where the power is going. It is not, on the evidentiary record the commission had in front of it, a demonstrated one — and the commission's majority wasn't willing to approve $584 million in ratepayer-backed construction on the strength of a plausible story.

Denied without prejudice is not denied for good

Duke's setback here is real but bounded. The commission denied the certificate without prejudice, which means Duke can refile rather than start the whole proceeding over from a worse position, and the order spells out exactly what a refiled application has to fix: tie the turbine's need to the final outcome of the Carbon Plan and IRP case rather than Duke's own interim forecast, explain how cost recovery complies with the Ratepayer Protection Pledge, state plainly whether Duke intends to seek recovery of construction costs while the plant is still being built, and show how the turbine stacks up against other resources that could serve the same need. Duke called the ruling disappointing and said it's assessing next steps, stopping short of committing to either a quick refiling or an appeal. Every item on that list is answerable. None of them is quick, and the Carbon Plan case Duke would need to lean on for the first item isn't finished either.

The diligence angle

This is the same gap we keep finding on the other side of the country in different forms — Texas's new interconnection deposit rule exists because regulators stopped trusting a developer's claimed megawatt figure until collateral backs it, and the federal Ratepayer Protection Act tried, and mostly failed, to force states to adopt cost-allocation standards by statute rather than by voluntary pledge. North Carolina's commission just did informally, inside one rate case, something closer to what that federal bill couldn't accomplish by law: it treated a utility's claim about who a new plant serves as something that has to be proven on the record, not assumed from a map. A gas turbine built next to a data center campus implies a relationship between the two. It doesn't establish one, and the commission's majority wouldn't let Duke's ratepayers underwrite $584 million on the strength of the implication alone.

GridPermit works Colorado, not Duke's Carolinas territory, and nothing in Xcel Energy's system looks exactly like a sixth turbine bolted onto a two-decade-old gas complex next to a hyperscale campus. But the evidentiary standard the commission just applied is one every state's regulators are drifting toward, ours included: a generating asset sited near a specific load is not the same rung on the evidence ladder as a signed, allocated interconnection agreement naming that load directly, and increasingly, state commissions are the ones enforcing the difference rather than taking a utility's word for it. In a Kill Test, "there's a power plant going in next door" has never counted as power confirmed for a specific site. North Carolina's commission just applied the same logic to the utility's side of the transaction, not just the developer's — and that cuts both ways on every site we evaluate.