Big Tech's Billion-Dollar Rush on Nuclear Uprates

Zach Martin

Welcome to NUCLEAR CAPITAL on KMKT I R Hub Radio. I'm Zach Martin, and with me is Emily Martin. Today's angle is narrow on purpose. We're not rehashing the whole nuclear renaissance. We're looking at how Big Tech money is landing on the existing U.S. reactor fleet through uprates, as of early October 2026.

Emily Martin

And a quick note before we start. Every company we mention today, Constellation, Vistra, Amazon, Google, Meta, and others, is discussed for informational purposes only. Nothing here is investment advice or a recommendation to buy or sell anything.

Emily Martin

Zach, last time we broke down the Google and Constellation deal on its own, as one anchor contract of 890 megawatts. This week it stopped looking like a single deal. In about a week we got Amazon at Calvert Cliffs, Google across eleven units, and a Department of Energy loan to Vistra. That's a pattern.

Zach Martin

It is, and I'd like to start with the simple version. Think of a farmer who needs more grain this season. He can plant a new field, which takes years. Or he can get more out of the acres he already owns. Uprates are the second option. Hyperscalers need firm power now for AI data centers, and advanced reactors and small modular reactors aren't expected to deliver meaningful commercial gigawatts until the 2030s.

Emily Martin

Google's own blog says it well. It says the fastest new megawatt is often the one that can be provided by existing infrastructure. There are 94 operating reactors in the U.S., and nuclear generates nearly 20 percent of the country's electricity. Those plants already have the substations, the transmission lines, the water, and the interconnection approvals.

Zach Martin

Right, and that interconnection point matters in PJM. Google's announcement says uprates deliver capacity gains without the multi year timelines or interconnection bottlenecks of greenfield construction. When the queue is the constraint, an asset that's already plugged in has a real head start.

Emily Martin

Let's walk through the deals. Tuesday, October 6th, Google and Constellation announced a 20 year power purchase agreement supporting 890 megawatts of new nuclear capacity. It comes from uprates at 11 units across six sites in Illinois, Pennsylvania, and New Jersey. Constellation is putting in more than $4.3 billion of its own investment. The first uprate is expected in 2028, and Google says all 890 megawatts arrive before the end of 2032.

Zach Martin

And there's a second piece people skip over. A separate 15 year supply agreement for another 2,700 megawatts across Constellation's operating PJM fleet. That one keeps existing assets economically viable. It isn't new capacity. Hold that thought, because I want to come back to it.

Emily Martin

Then Amazon. Constellation announced September 30th a 20 year agreement for 690 megawatts from Calvert Cliffs in Maryland, including a 190 megawatt uprate. That's roughly a 10 percent increase on a plant of about 1,790 megawatts. It backs more than $3 billion of Maryland infrastructure investment, with the new capacity expected between 2030 and 2032.

Zach Martin

And the third piece is the Department of Energy. On October 5th its Office of Energy Dominance Financing, which used to be the Loan Programs Office, announced a conditional commitment of up to $4.2 billion to Vistra. It covers 433 megawatts of uprates at Beaver Valley in Pennsylvania, and Perry and Davis-Besse in Ohio. It's also supposed to preserve nearly 4 gigawatts of existing nuclear. Note the word conditional. Vistra still has to satisfy technical, legal, environmental, and financial conditions before anything is funded.

Emily Martin

Fair. And that loan ties back to Meta's 2.6 gigawatt agreement with Vistra from January. Perry brings 213 megawatts of uprate, Davis-Besse 80, Beaver Valley 140. Add it up and you get 433. So within a single week you have a private balance sheet, a hyperscaler contract, and federal debt all pointed at the same asset class.

Zach Martin

Here's where I want to push. Everyone is calling this a bridge. But do the arithmetic on the new megawatts. Google's 890, Amazon's 190, Vistra's 433. That's about 1.5 gigawatts. That's our math, not a number from any press release. PJM is short about 6.8 gigawatts. So we're talking a bit more than a fifth of the gap, and the timelines run out to 2032.

Emily Martin

But nobody serious says uprates close the gap by themselves. The claim is about speed and certainty per dollar. Google says this latest deal takes its enabled new nuclear capacity from uprates and restarts past 1.5 gigawatts, and that by the end of the decade that's nearly 25 percent of the Department of Energy's goal of 5 gigawatts from uprates and restarts by 2030. That's one company moving a national target.

Zach Martin

Nearly 25 percent of a goal is still a goal, Emily. And I'll press the definitional point. Google says the 890 is entirely new capacity. I believe the physics. But with Amazon, only 190 of the 690 megawatts is new. The other 500 is power from a plant that already runs. Is that expansion, or is it a hyperscaler buying a seat at an existing table?

Emily Martin

It's both, and that's the point. Look at what Constellation said. The revenue certainty lets it pursue relicensing for another 20 years. The licenses at Calvert Cliffs expire in 2034 and 2036. If Amazon's commitment is what keeps a 1,790 megawatt plant running into the 2050s, then the existing power is the foundation for the new power. You can't uprate a plant that's shutting down.

Zach Martin

Okay, that's a real concession on my side. The existing megawatts and the new ones are bundled for a reason. The financing logic depends on the old base. But it does change how I'd describe these deals. They're as much life extension contracts as growth contracts.

Emily Martin

I'd accept that framing. Life extension plus incremental capacity. And it explains why a hyperscaler would sign 20 years when its own chips turn over every few.

Zach Martin

Let's talk about how the money actually works, because that's the capital markets story. A 20 year contract from a company like Google or Amazon is a long dated, high quality revenue stream. An independent power producer can finance billions of capital spending against it without stretching its own balance sheet to the limit. Constellation's $4.3 billion and the roughly $3 billion at Calvert Cliffs are examples.

Emily Martin

And the Vistra piece adds the public leg. Federal debt is cheaper and longer than most private options. Blend a hyperscaler offtake with a Department of Energy loan and you've got shared risk. The tech company takes the demand risk off the table, and the government supplies patient capital. That structure didn't really exist for nuclear a few years ago.

Zach Martin

Though I'd add a caution on markets. Utility Dive reported Constellation shares were up almost 14 percent in morning trading on the Google announcement. That tells you how much investors value this kind of contract, but a one day move isn't a forecast, and we're not here to judge any stock. The discipline question is whether the returns hold when real construction work meets real schedules.