Uprates Before Reactors: The Fastest Nuclear Megawatts

Zach Martin

Welcome to NUCLEAR CAPITAL on K M K T I R Hub Radio. I'm Zach Martin, and I'm here with Emily Martin. Today's episode is called Uprates Before Reactors: Google, Constellation, and the Fastest Nuclear Megawatts. Here's the setup. The artificial intelligence load is here. Data centers are asking for power right now. The new reactors, the small modular ones, the advanced designs, are not here yet. So the question for today is a plain one. What if the fastest nuclear megawatts are the ones already spinning?

Emily Martin

That's the question, Zach, but I'm going to be a little skeptical from the start. Every few weeks there's a headline with a big megawatt number in it, and my job on this show is to ask what's real and what's paper. When I say real, I mean steel, licenses, and a grid connection. So let's start with the basics. What is an uprate? Because I think a lot of listeners hear that word and picture a new reactor.

Zach Martin

It's the opposite, and that's the whole point. An uprate means getting more electricity out of a reactor that's already built and already licensed. Think of it like tuning an engine you already own instead of buying a new car. The Nuclear Regulatory Commission describes three flavors. The smallest is called a measurement uncertainty recapture, usually under two percent. Better flow measurement lets the plant operate closer to its true limit, instead of padding the margin. The middle one is a stretch power uprate, typically two to seven percent, done mostly by adjusting instrument setpoints and controls within the original design margins.

Emily Martin

And those two are cheap and small. But the Google deal isn't that kind, is it?

Zach Martin

No. The big one is the extended power uprate. The NRC says those have been approved for increases as high as twenty percent. That's real hardware. Turbines, generators, large transformers, bigger feedwater pumps and motors, and sometimes a slightly different fuel reload. But notice what you're not building. You're not building a new reactor vessel. You're not building a new containment structure. The thermal power goes up, and the balance of the plant has to be able to carry it.

Emily Martin

So compare that to the alternatives. A new small modular reactor is a new project on a new site, or at least a new licensing path. A restart is bringing back a plant that's been shut down. An uprate is a licensed plant, with an operator, a workforce, cooling water, and a connection to the grid, getting a bigger engine. Is that the simple version?

Zach Martin

That's it. And the grid matters, because this is all happening in PJM, the PJM Interconnection. It's the largest grid operator in the country, and Google's own executive described it as serving sixty seven million people. It's also where a lot of the data center demand is landing. Utility Dive reported that PJM is short roughly 6.8 gigawatts of new capacity and has been struggling to find resources. So a megawatt added inside PJM isn't just a nice press release. It lands where the shortage is.

Emily Martin

Okay, then let's walk the numbers, because this is where people get confused. Early October 2026, Google and Constellation announce a deal. The headline figure I keep seeing is about thirty five hundred megawatts. Is that all new nuclear?

Zach Martin

No, and that's the most important separation in the whole story. The total is about 3,590 megawatts. It's two pieces. The first piece is 890 megawatts of incremental capacity, new output created by uprates at eleven Constellation owned units, across six sites in Illinois, Pennsylvania, and New Jersey. The second piece is a separate fifteen year energy supply agreement for another 2,700 megawatts from Constellation's existing operating fleet in PJM. Eight ninety plus twenty seven hundred gets you thirty five ninety.

Emily Martin

So only about a quarter of that headline is new.

Zach Martin

About a quarter, yes. And I want to be careful with how I describe the other piece. Constellation's announcement says the 2,700 megawatt agreement is structured so the operating plants keep delivering energy and capacity to the PJM market, and that it doesn't take resources off the grid. So Google is buying existing electrons, and the value to Constellation is revenue certainty. The 890 is the part that adds to the supply stack.

Emily Martin

And the uprate piece is the twenty year power purchase agreement.

Zach Martin

Right. Twenty years for the uprate capacity, fifteen years for the existing output. The company says it represents more than 4.3 billion dollars of new investment by Constellation. It describes sustaining roughly 4,400 existing jobs and creating approximately 7,200 construction period jobs. And the first uprate is expected to be delivered by 2028. The schedule I've seen has all eleven finished before the end of 2032.

Emily Martin

Let me push on that timing, because it's the thing I care about most. 2028 is the first one. That's not next quarter. The 890 megawatts is a ramp over several years, not a switch that flips on a Tuesday.

Zach Martin

Completely fair. It's a staircase. Eleven units means eleven separate projects, each with its own regulatory paperwork and its own outage window. The announcement calls the 890 the equivalent of a new reactor, or the output of several small modular reactors. That's a fair comparison for size. But it arrives in steps, from 2028 on, not all at once.

Emily Martin

There's also a technology side. Constellation picked Google Cloud and Gemini Enterprise. Is that a side note or does it matter for the capital story?

Zach Martin

It's a five year technology alliance, and the stated aim is an AI for energy blueprint. The areas named are site selection, power flow modeling, permitting and interconnection planning, outage and turnaround management, and securing the operational technology networks. Here's why I'd pay attention. Outage management is exactly where uprate schedules live or die, and I'll get to that. But I'd keep it in proportion. That's a stated plan, and I haven't seen results, so I'd call it a promise to watch rather than a proven schedule saver.

Emily Martin

Agreed. Now, you said this wasn't the only one. A week earlier, Amazon.

Zach Martin

Late September 2026. Amazon and Constellation announced a twenty year power purchase agreement tied to Calvert Cliffs in Maryland. The deal covers 690 megawatts, including about 190 megawatts of additional capacity. Constellation targets that new capacity for operation between 2030 and 2032. It supports more than three billion dollars in plant improvements, and it supports Constellation's effort to renew the operating licenses for another twenty years.

Emily Martin

So it's the same pattern. A big contracted number, a smaller piece that's truly new, and a date that's further out. Five hundred megawatts is existing output, one ninety is new, and the new part arrives somewhere between 2030 and 2032.

Zach Martin

Same pattern, and it's a useful one. The license extension is part of the value, because a plant that's licensed to run longer is a plant you can justify spending real money on. Spending three billion dollars on a plant that might close in a decade is a different conversation than spending it on a plant that's planning to run for decades more.