Summary

  • Duke Energy’s two North Carolina utilities and several large customers have filed a settlement that would require covered new loads to pre-fund dedicated facilities, post security for transmission upgrades, pay against at least 75% of contracted demand and accept long contract and exit terms.
  • The proposal is not yet a Commission order. It defers a methodology for assigning portions of shared network upgrades to particular customers, and neither the filed tariff nor Duke’s summary requires clean-energy procurement.
  • The commercial test is whether the final tariff makes project-specific costs, shared-system costs, reliability obligations and exit exposure observable before utilities commit capital—not whether a headline promise says data centres will pay their way.

When a data centre needs a new substation, the customer-specific part of the bill is easy to point to. The more difficult question is what happens when that project also relies on a transmission upgrade that serves several loads—or strengthens the grid for customers who have not yet signed a contract. North Carolina’s new large-load settlement puts a detailed price on the first boundary and leaves the second one open for more work.

On October 6, Duke Energy Carolinas and Duke Energy Progress filed an Agreement and Stipulation of Settlement on Large Load Tariff in two North Carolina Utilities Commission proceedings. North Carolina Public Staff, Amazon Data Services, Andale, Carolina Industrial Group for Fair Utility Rates, federal agencies, Google and Microsoft are among the signatories. Duke described the agreement the next day as a way to protect existing customers from costs associated with large loads. But this is a negotiated filing, not a Commission decision. The Commission may accept, reject or modify it; the parties say the settlement is binding only if accepted in its entirety. (Filed settlement; Duke’s announcement)

Who would be covered—and when

The proposed tariff is aimed at a defined class, not every data centre or every large employer. It covers new letter agreements and electric service agreements executed on or after June 1, 2026, at a single premise or multiple premises within one mile, if the load either contracts for at least 50 megawatts and is expected to run at an 80% or higher rolling annual load factor, or contracts for at least 150 MW at any load factor. Incremental service that independently crosses the threshold can also be covered. The agreement contains exceptions for some customers already receiving service or holding earlier agreements.

The actual filed language, rather than the shorthand “50 MW or more,” determines whether a particular project falls inside it.

That threshold matters commercially. A customer below the applicability test is not automatically subject to the proposed terms, while a covered project has to model them before choosing where and how quickly to build. The settlement would use Duke’s High Load Factor schedule or a future large-load rate schedule; it does not publish the rates for a specific campus or identify any customer contract.

The connection has an advance price; the network does not yet have a final allocation rule

For facilities built only to serve one customer—such as a dedicated substation—the tariff requires the customer to advance the estimated cost as a contribution in aid of construction within 60 days of a letter agreement. The customer must provide additional money if actual cost exceeds the estimate. Refunds depend on whether an electric service agreement is completed and what costs Duke has already incurred; some unspent funds after early termination can be applied to damages.

Transmission network upgrades receive a different treatment. The customer posts a letter of credit equal to the estimated cost of upgrades needed to serve its load, also with a mechanism to top up if actual costs are higher. After the customer has reached at least 75% of its full contract demand for three consecutive billing months, the letter of credit can be reduced ratably over the remaining term. If service ends early, the utility may draw against the security for the remaining net book value of relevant upgrades under the tariff’s terms.

That is meaningful protection against a project that never energizes or leaves before the investment is recovered. It is not the same as a final rule assigning every shared upgrade to the load that caused it. The settlement directs Duke to file a report within 12 months after Commission approval, following stakeholder discussions, on possible methods for directly assigning portions of a network upgrade to a customer that triggers or uses it. The discussions are to include demand charges, exit fees, customer contributions and how costs should be divided when multiple large loads use the same system elements.

The parties remain free to argue for different methods later, and the Commission retains the final decision. A method would apply to agreements entered after that method is approved.

This timing creates the central regulatory gap. The new connection itself can be assigned and funded before work begins; the allocation of broader network improvements may depend on a study, another proceeding and a later Commission decision. For developers and utilities, that distinction affects the project’s cost forecast and the risk that capacity built for one anchor tenant has value to the wider system. The filing establishes a process for resolving that question, not the resolution itself.

Long contracts make demand assumptions part of the financing

The tariff also makes a large customer’s commitment extend beyond its initial build schedule. An agreement below 100 MW would have a minimum initial term equal to the greater of 10 years or five years plus the ramp period. At 100 MW or above, the minimum would be the greater of 15 years or 10 years plus ramp. Once contract demand is in effect, the billing demand cannot fall below 75% of the contracted amount for demand charges.

Exit becomes more expensive as the requested load grows. A customer seeking to terminate or reduce demand must give 24 months’ notice below 250 MW, 36 months between 250 and 499 MW, and 48 months at 500 MW or more. The early-termination formula includes all minimum-demand and extra-facility charges during the notice period, the net present value of remaining extra-facility charges, and 25% of remaining minimum-demand charges after the notice period, with credits for amounts paid during notice. A one-time reduction of up to 20%, capped at 200 MW, is allowed without liquidated damages after the sixth anniversary, under the stated conditions.

Exhibit 2 makes the scale visible but should not be mistaken for a forecast. It models a 500 MW Duke Energy Carolinas customer, a 15-year term, four-year ramp and four-year notice period, and assumes $50,000 a month in extra-facility charges. The formula produces a maximum illustrative damages/security basis of about $587.4 million. That is a worked example under selected assumptions, not a fee attached to a named project, an expected customer bill or proof that a customer would default.

The long term and minimum bill can make utility investment more financeable by reducing the chance that a customer reserves capacity and then walks away. For a data-centre operator, the same terms raise the cost of a slower-than-expected occupancy ramp, a change in compute demand, a customer loss or a shift toward on-site supply. The project’s power contract is therefore a capital commitment, not just a utility connection application.

Reliability is another condition of access

During the load-ramp period and the first year of service, each agreement must include one of two curtailment options or a Commission-approved alternative that accelerates capacity resources on the utility system. The options allow interruption of 95% of contract load for up to 50 hours per year, or 50% for up to 100 hours per year. Curtailment is capped at six hours per day and is not compensated. A customer can seek a Commission waiver for good cause; the agreement also describes treatment for federal customers.

Those provisions make reliability a contractual term rather than an assumption that all requested megawatts will be firm from day one. An operator has to decide whether its workloads can be shifted, backed by on-site generation or held through interruptions. A utility can use that flexibility while integrating load, but the settlement does not establish how often curtailment will occur or how much capacity it will ultimately accelerate. Those are operational results to observe after approval.

Consumer protection is disputed, not proven by the settlement’s title

Duke says the agreement memorializes and enhances protections it had already implemented, including upfront customer payments, deposits and guarantees, and a separate High Load Factor rate. The Southern Environmental Law Center and allied groups take a different view: they argue that the proposed terms are too lenient on exit costs and do not create a customer class that fully tracks the cost large loads impose on the system. They have also called for a clean-transition tariff that would link large-load growth to new clean energy. Those are stakeholder positions, not findings by the Commission. (SELC’s response)

The text supports both a narrower conclusion and an important caveat. There are concrete protections for dedicated facilities, credit support for transmission upgrades, minimum demand bills, long terms and termination damages. At the same time, a future cost-of-service analysis and a later assignment methodology are still contemplated; signatories keep their ability to argue their positions. The tariff’s reliability section does not require a customer to procure clean energy. A proposed customer contribution is not the same as a complete accounting of generation, transmission, local distribution and emissions effects.

The Commission’s decision, expected by Duke in mid-November, is the next public test. If the settlement is approved without material change, the follow-on work should be judged by whether the promised report produces an auditable method for assigning shared upgrades, whether customer security declines only as utilities recover costs, and whether the approved agreements disclose assumptions about ramp, contract demand and exit liability. The Commission can still reject, modify or condition the package, so the forecast date is not an approval commitment.

For data-centre investors, a site is not power-ready merely because a utility has published a customer-protection framework. The diligence file needs the applicable threshold, the utility’s load-ramp assumptions, dedicated-facility estimates, network-upgrade security, minimum-bill exposure, interruptibility, exit formula and a map of costs that have not yet been assigned. For regulators and ratepayers, the decisive question is whether those contracts are later reconciled against the cost of serving the load and the value of upgrades used by other customers.

North Carolina has proposed a stronger contract boundary around the first customer-specific assets. It has not yet closed the broader allocation question. The distinction is why “data centres pay for the grid” is too broad a conclusion: the settlement makes some payments explicit, leaves other costs for a later record, and awaits the Commission’s judgment.

Sources