NUCLEAR CAPITAL
All Episodes
Uprates, Not Unicorns: Google Bets $4.3B on Nuclear Power

Uprates, Not Unicorns: Google Bets $4.3B on Nuclear Power

0:00|0:00

Zach and Emily break down Google's massive 20-year power purchase agreement with Constellation Energy, funding 890 megawatts of incremental nuclear uprates across 11 reactors. Discover how hyperscalers are bypassing new reactor buildouts to de-risk investments and secure firm clean power from existing infrastructure.


Chapter 1

The Check Is for the Fleet

Zach Martin

Eight hundred ninety megawatts. More than 4.3 billion dollars. Those are the two numbers I want you to hold onto today, because they describe a check that was actually written for power that gets delivered this decade. Welcome to NUCLEAR CAPITAL on KMKT I R Hub Radio. I'm Zach Martin.

Emily Martin

And I'm Emily Martin. It's Thursday, October 8, 2026, and the headline is that Google and Constellation Energy announced a 20 year power purchase agreement covering about 3,590 megawatts. Of that, about 890 megawatts is incremental capacity, meaning new output created by uprates at 11 existing reactors in Illinois, Pennsylvania, and New Jersey. Constellation said it expects to invest more than 4.3 billion dollars to do it.

Zach Martin

And the title of the show tells you where we're going. Uprates, not unicorns. The market spent years waiting for the first of a kind reactor to prove itself. This week the money went to the fleet we already have. Before we go further, a quick note. Every company we mention today is discussed for informational purposes only. Nothing here is a buy or sell recommendation, nothing here is investment advice, and past price moves are not indicative of future results. Okay, Emily, let's separate the pieces of this deal, because 3,590 and 890 are not the same number.

Emily Martin

Right, and that's the first thing people blur. The 890 megawatts is the new part. It comes from upgrades that make operating reactors produce more. The remaining roughly 2,700 megawatts comes from existing plants under a related supply arrangement. So you can read the deal as two pillars. One is a customer locking in power from plants that are already running. The other is a customer funding additional megawatts that don't exist today. The first targeted uprate is around 2028, and reporting says the full 890 megawatts is scheduled before the end of 2032.

Zach Martin

Think of it like a restaurant that's already packed every night. You can spend years permitting and building a second location, or you can add a shift and upgrade the kitchen equipment in the building you own. The uprate is the second option. And we covered the mechanics in our earlier episode, Nuclear Uprates: Big Tech's Shortcut to Clean Grid Power, so I'll keep this short. Work on an operating reactor can include steam turbine upgrades, steam generator modernization, digital controls, and fuel and core reload changes. Then there's the license work, and that part matters for the money.

Emily Martin

It does. Every uprate needs formal approval from the Nuclear Regulatory Commission, through license amendment requests. So an uprate isn't a science experiment, but it's also not a weekend project. It's engineering plus paperwork plus outage scheduling. The upside is that you're working on a site that already has a license, a workforce, a grid connection, and a neighborhood that has lived with it for decades. The source coverage also notes the companies say this lines up with the federal UPRISE program, which is about growing output from existing U.S. nuclear infrastructure.

Zach Martin

Here's where I want to slow down and get capital markets literate, because the contract structure is the story. A generator selling into PJM without a long contract is exposed to wholesale power prices. Those prices move. Now put a 20 year agreement with a very large technology buyer on top of a multi billion dollar capital program. What changes?

Emily Martin

What changes is the question a lender or an investment committee asks. Without a contract, the question is what power prices will do over the life of this project, and nobody can answer that with confidence. With a 20 year offtake, the question becomes whether this counterparty will keep paying, and whether the plant will deliver. Constellation said it expects to invest more than 4.3 billion dollars, and the Zacks write up put it plainly: the contracted revenue supports the planned upgrades. Google gets long term firm supply, Constellation gets revenue certainty. I'd add one honest caveat. We haven't seen the contract terms, the pricing, or the financing structure. So when people say this de risks the project, that's our interpretation of how long contracts generally work, not a disclosed fact about this deal.

Zach Martin

Good flag. And there's a second strategic piece on the same day. Constellation also selected Google Cloud and Gemini Enterprise under an expanded five year technology alliance, to develop what they call an AI for energy blueprint. So Google isn't just the customer for the electrons. It's also becoming a technology supplier to the generator. I'd treat that as a relationship signal, not a financial driver, because the reporting doesn't give us dollar figures on it.

Emily Martin

Agreed. Now, why PJM? PJM serves about 67 million people, according to Google's own quote in the announcement, and the Zacks piece describes it as a market where demand has been rising and available generating capacity is tight. If you're a hyperscaler that needs firm power around the clock, you can wait in line for something new, or you can pay to make existing plants bigger. This deal says the second path is available right now. Also, an uprate is incremental capacity at a site that's already interconnected. The announcement frames it as expanding total generation available across PJM. In plain terms, the 890 megawatts is new supply, not just a relabeled contract.

Zach Martin

And the Amazon parallel shows up right on schedule. Around September 30, Amazon and Constellation announced a 20 year agreement supporting roughly 190 megawatts of additional capacity at Calvert Cliffs in Maryland. That's tied to more than 3 billion dollars of plant improvements and license renewal efforts, and the target is operation between 2030 and 2032. Different size, same pattern. One hyperscaler in late September, another in early October, both attaching long contracts to operating plants.

Emily Martin

Put those side by side and you get something that looks less like a one off and more like a template. Though I'd be careful there too. Two deals with one counterparty on the generator side, both announced within about a week, is a pattern worth watching, not yet a market structure. Which brings us to something I want to talk about in the next segment: how the stock market reacted, and what it was actually saying.