On November 9, 2026, Constellation Energy disclosed a product it had never sold before: a 20-year agreement for 270 megawatts of unit-contingent energy, capacity, and nuclear attributes from its existing nuclear fleet, set to begin in 2028. It is the first transaction under a new integrated offering called Peak+, and the customer was not named. The deal is small next to the headline numbers in the same release — a 20-year, 690-megawatt agreement with Amazon and a 20-year, 890-megawatt uprate agreement with Google — but it may be the most revealing of the three: hyperscaler contracts are becoming negotiated bundles of capacity, hourly shape, and delivery risk, not just purchases of clean volume.

A data center draws power nearly flat, around the clock, at 70 to 90 percent of contracted capacity. The grid it draws from does not. Demand peaks on hot summer afternoons and cold winter mornings, and the capacity market that pays generators to be available in those hours is priced around those peaks. A flat block of nuclear energy solves the average hour, not the peak. Peak+ is the attempt to sell both at once.

What Peak+ actually is

Constellation’s release describes Peak+ as “an integrated solution that brings together new capacity, including new generation and demand response, with energy products to help large customers meet their evolving needs.” On the earnings call the same morning, chief executive Joe Dominguez described Peak+ in broader terms: capacity solutions plus existing clean and reliable generation. The first contract, he said, “combines 270 megawatts of unit contingent energy capacity and clean energy attributes from our existing nuclear fleet with new generation supply and demand response designed to satisfy expected regulatory requirements.”

That last phrase is doing real work. “Expected regulatory requirements” is not the language of a seller describing a buyer’s sustainability target; it is the language of a seller describing a compliance problem. PJM, the grid operator covering Constellation’s nuclear fleet and much of its data center load, has been building a framework under which large new loads are expected to bring their own power — new supply matched to new demand — rather than leaning on a system whose capacity prices have already surged. Peak+ is a product designed to be sold into that expectation.

The mechanics, as far as Constellation has disclosed them, are a bundle rather than a single resource. The nuclear fleet supplies the energy and the clean attributes — the 24/7 carbon-free output that anchors the deal. New capacity supplies the peak: batteries, additional generation, or demand response that can be called on when the system is tight. Dominguez said Constellation could add batteries and generation supply at its plants, “given the vast amount of land that we own around the plants.” Constellation has not said which resources back this 270 megawatts, or how much new capacity, if any, is already attached to it.

Three products in one contract

The release’s phrasing — “energy, capacity, and nuclear attributes” — lists three distinct products, each solving a different problem. Energy is the megawatt-hours: electricity delivered around the clock from reactors that run at capacity factors above 90 percent. That flat output is the closest match in the market to a flat load. Capacity is the scarcer product, and the one this deal is really about. In PJM, capacity is a separate market from energy: generators are paid to commit to being available during the system’s tightest hours. By bundling capacity into the contract, Constellation is selling the buyer a position in the peak hours, not just the average ones. The “plus” in Peak+ is, in substance, a capacity product stapled to a nuclear energy contract.

Nuclear attributes are the third leg: the clean-energy claim attached to each megawatt-hour, which the buyer uses against its carbon commitments. Attributes from existing nuclear are not new clean generation — the reactors were already running — but they are the mechanism by which a buyer claims its consumption is matched by carbon-free output. Constellation is not selling three things because buyers like bundles; it is selling three things because buyers now have to solve three problems simultaneously.

Why the peak hours matter to a flat load

Why should a customer whose servers draw power at nearly the same rate at 3 a.m. as at 3 p.m. care about peak-hour firmness at all? The answer is that the grid charges for the peak whether or not the customer’s own load shape creates it. Capacity obligations in PJM are allocated on the basis of a customer’s contribution to the system peak, and a flat load contributes to every peak there is. A data center running at 90 percent utilization through a July heat wave is drawing at full tilt in exactly the hours when the system is most stressed.

The argument running through every data center interconnection fight this year is that new large loads should fund the capacity their arrival requires, rather than socializing the cost across existing customers. A buyer that arrives with a contract covering its own energy and capacity preempts that argument. Peak+ is, among other things, a preemptive answer to the who-pays question: the customer brings its own peak coverage, purchased from the same seller that supplies its energy.

What “unit-contingent” limits

The most important word in the disclosure may be the one that sounds like boilerplate: “unit-contingent.” In power contracting, a unit-contingent sale means delivery depends on specific generating units operating. If the designated unit is offline — for a refueling outage, a forced outage, or a derate — the seller is generally not obligated to replace the power from elsewhere in its portfolio or from the market. The buyer absorbs the outage risk.

That is a meaningful limit on the firmness the deal appears to promise. The capacity product is sold as peak-hour coverage, but a unit-contingent contract covers the peak only when the unit is running. Refueling outages are planned — Constellation reported 20 planned refueling outage days in the third quarter, against 23 a year earlier — but forced outages are not, and the fleet recorded six non-refueling outage days in the quarter. A buyer counting on 270 megawatts in a peak hour needs to know what happens when the unit behind it is down. The disclosure does not name the units, so the public cannot assess which reactors’ outage schedules and histories are in play.

None of this makes the deal a bad one. Unit-contingent contracts are standard in the industry, and they are priced accordingly: the buyer accepts delivery risk in exchange for a lower price than a fully firm, portfolio-backed sale would command. It is a twenty-year claim on the output of specific reactors, plus a capacity component whose resources Constellation has not yet specified. The firmness is real but bounded, and the bound is the unit.

The price frame — and what it is not

Constellation did not disclose the price of the Peak+ contract, and the price should not be inferred from the figures floating around the deal. What the company did put in front of investors is a framing: slides in its third-quarter presentation put premium pricing for nuclear power purchase agreements at $20 to $50 per megawatt-hour above the floor set by the federal nuclear production tax credit, and below the $110 to $130 per megawatt-hour levelized cost it estimates for new combined-cycle gas turbines in eastern PJM by 2032, assuming a $3,000-per-kilowatt build cost. The production tax credit pays nuclear units up to $15 per megawatt-hour when revenues fall between $26 and $44.75 per megawatt-hour — the floor the premium is measured against.

That framing is Constellation’s sales pitch, not a contract price, and the distinction is worth stating plainly. The $20-to-$50 band describes premium nuclear PPAs generally, as a 2032 estimate; the slides give no separate price for Peak+ itself, and the band is a premium over a floor, not a full price, so it cannot be read directly against the gas benchmark either. It shows how Constellation wants the product positioned: above the subsidized floor, below new gas.

The quarter around the deal

Peak+ did not arrive in a quiet quarter. Constellation reported third-quarter adjusted operating earnings of $5.06 per share, up from $3.04 a year earlier, on GAAP net income of $5.19 per share against $2.97 in the prior-year quarter. Revenue reached roughly $9.6 billion, up from $6.6 billion, driven by favorable market conditions and the addition of Calpine, acquired in January 2026. Full-year adjusted earnings guidance was raised and narrowed to $12.50 to $12.80 per share.

The contracting picture is the more important context for this desk. Dominguez told analysts that Constellation signed three additional long-term agreements totaling about 4,560 megawatts since the prior quarter, and more than 8,500 megawatts of long-term agreements in 2026 to date, extending well into the 2040s. The Peak+ deal is 270 megawatts of that total — about 3 percent of the year’s contracting, and about 1 percent of a roughly 22-gigawatt nuclear fleet.

The same release recapped the two deals that frame Peak+ as the third act of a single strategy. In September, Constellation signed a 20-year, 690-megawatt agreement with Amazon, enabling more than $3 billion of investment at Calvert Cliffs — including a 190-megawatt capacity expansion expected online between 2030 and 2032. In October came the Google package this desk covered in detail: a 20-year PPA backing 890 megawatts of uprates across eleven units in Illinois, Pennsylvania, and New Jersey, enabling more than $4.3 billion of investment, with the first uprate expected in 2028, alongside a separate 15-year load-following agreement with Google for 2,700 megawatts in PJM. The release also disclosed a conditional DOE loan guarantee of up to $5.0 billion for uprates and a $715 million deal to buy the 609-megawatt Rhode Island State Energy Center gas plant from Shell Energy North America.

What the deal proves — and what it does not

Start with what it proves. First, there is a buyer: Dominguez described the customer as an investment-grade customer in the data economy market, buying at PJM’s Northern Illinois Hub. An unnamed buyer is still a buyer, and a twenty-year commitment from an investment-grade counterparty is the kind of revenue certainty that lets a generator plan around its fleet rather than around the spot market. Second, it proves the product exists as something a customer will sign, not just a slide in an investor deck. Third, it proves the negotiation has moved. A year ago the story was volume; now it is structure: which hours, which units, which risks, which regulatory expectations.

What it does not prove is nearly as important. One 270-megawatt deal does not prove Peak+ scales; Constellation declined to project how much of the product it would sell. It does not prove the capacity component works as advertised, because the company has not said which batteries, demand response, or new generation back this contract, or whether any of it exists yet. And it does not prove the unit-contingent structure will satisfy the regulators whose expectations the product is designed to meet — “expected regulatory requirements” are, by definition, requirements not yet finalized.

What to watch

Three things will determine whether Peak+ becomes a template or a footnote. The first is the second deal: who signs next, at what size, and whether the customer is named. The second is the capacity component. When Constellation discloses the batteries, demand response, or generation paired with a Peak+ contract — and whether it is built on land the company already owns around its plants — the product’s real firmness becomes measurable rather than asserted. The third is the regulatory test: whether a Peak+ contract counts as bringing your own power in the eyes of PJM and the state commissions. If it does, expect the product to spread quickly; if it does not, expect it to be restructured.

The fourth thing to watch is the uprate pipeline behind it. The Google and Amazon deals fund roughly 1,080 megawatts of new nuclear capacity between them, and the conditional $5 billion DOE loan guarantee is sized to support more. Peak+ sells the existing fleet; the uprates grow it. Whether the two products converge — uprate output sold through Peak+ structures — is the natural next step.

The bottom line

Constellation’s first Peak+ transaction is small — 270 megawatts, one unnamed investment-grade customer in the data economy, starting in 2028. What makes it worth a data center desk’s attention is not its size but its shape: energy, capacity, and attributes in a single twenty-year contract, built around existing nuclear, paired with new capacity and demand response, and aimed squarely at the peak hours and the regulatory expectations that now govern large-load contracting.

The hyperscaler power deal has acquired a new layer of detail. Volume was the first negotiation. Now the negotiation is about which hours the power is guaranteed, which units it comes from, and who bears the risk when one of them is down. Peak+ is Constellation’s answer to those questions. Whether buyers and regulators accept the answer will be decided by the deals that follow.