Bezos… DOOMED?

Every major hyperscaler has a nuclear story right now. Most of those stories will not deliver a single megawatt before 2032. Google’s is different, and the difference is the asset at the center of it: reactors that are already licensed, already generating, and already connected to the grid.

On October 6, 2026, Google and Constellation Energy announced a 20-year power purchase agreement covering 890 megawatts of new nuclear output. The “new” nuclear here is not a new reactor. It is uprates: new equipment and higher thermal and electrical efficiency at reactors that already exist, pulling more megawatts out of the same fleet. Constellation plans those uprates across 11 units at six sites in Illinois, Pennsylvania and New Jersey, with the first expected by 2028. The project represents more than $4.3 billion of new investment and creates approximately 7,200 new construction jobs during the building period.

Why the Delivery Clock Matters

Corporate nuclear procurement sorts cleanly by delivery date. Contracts with operating reactors deliver now, and uprates add capacity over several years. Restarts wait on federal licensing, while new small modular reactors wait on a construction decision. That distinction is not a footnote. It determines when electricity actually flows.

Amazon’s latest move makes the contrast vivid. Amazon, American Electric Power and X-energy announced a joint development agreement on October 19, 2026 to explore development of a 960-MW advanced nuclear project at the former Breed Generating Station coal plant site in Sullivan County, Indiana. The facility would use 12 of X-energy’s Xe-100 small modular reactors, each designed to produce 80 MW. The companies gave no timing details around the project. X-energy’s SEC filings say its first commercial delivery is planned for the early 2030s. “Exploring development” and “early 2030s” are a long way from 2028.

Constellation’s 20-year PPA with Meta for its Clinton Clean Energy Center in Illinois has the agreement starting in 2027 and covers 1,121 megawatts of emissions-free nuclear energy output. That is real capacity from an operating plant, a structure closer to Google’s deal than to Amazon’s greenfield bet, but it adds no new megawatts to the grid. Google’s uprates do.

The Scarcity Argument for CEG

The uprate megawatts carry a second advantage that rarely appears in deal announcements: they enter regional transmission markets directly. The 20-year PPA enables investments at 11 Constellation-owned nuclear units, increasing thermal and electrical efficiency to unlock additional reliable, firm power to serve the grid, and Constellation and Google say this capacity is new to the PJM Interconnection grid. Behind-the-meter generation, by contrast, stays local. It does not strengthen the grid serving more than 67 million consumers in PJM’s footprint, and it cannot be dispatched to where demand is highest.

Constellation holds a structural position no other company can replicate quickly. It has now inked long-term nuclear PPAs with Google, Amazon, Meta and Microsoft since late 2024. Each deal draws on the same finite pool: an existing licensed fleet that took decades and hundreds of billions of dollars to build. No startup and no import contract recreates that asset base on a useful timeline.

Constellation is not the only power producer benefiting from that finite supply. PJM capacity auctions clearing at the price cap two years running have turned dispatchable power owners into a new class of growth asset.

Morgan Stanley raised its price target on Constellation to $378 following the Google announcement. The deeper point is simpler: nuclear uprate capacity is now a scarce, biddable asset alongside GPU allocation. Constellation owns the units. The hyperscalers are bidding for the output. That is the dynamic a long-term investor should track, not the press release announcing yet another reactor that will be debated, licensed, and built over the next decade.

What Could Go Wrong

Uprates are not risk-free. Regulatory approval from the Nuclear Regulatory Commission is required for each unit, and engineering work at operating plants can surface surprises that delay timelines or inflate costs beyond Constellation’s $4.3 billion estimate. The unresolved framework around co-located, behind-the-meter load in PJM remains an overhang. And in Constellation disclosures about the Crane Clean Energy Center restart, PJM interconnection studies have identified contingent transmission upgrades with projected in-service dates that could extend as late as December 2030. A competitor with a faster regulatory path, or a prolonged low power price environment, could also compress the economics of a 20-year fixed contract.

None of those risks changes the fundamental picture. The existing nuclear fleet is the scarce asset of the decade. Constellation’s 11-unit uprate program under a 20-year Google contract is the clearest proof that the market already knows it.