Amazon and Constellation sign 20-year nuclear power purchase agreement for 690 MW at Calvert Cliffs
The agreement covers 690 MW from the 1,790-MW Maryland plant, including 190 MW from an uprate the contract is backing, plus a separate retail supply deal for Amazon's PJM operations.
Constellation has signed a 20-year power purchase agreement with Amazon for 690 megawatts from the Calvert Cliffs Clean Energy Center in Maryland, 190 of which come from an uprate the contract is backing. The plant's 1,790 megawatts generate roughly 80% of Maryland's clean energy, enough to power about 1.3 million homes and, on that accounting, the largest source of carbon-free power in the state; Amazon's share is about 39% of current nameplate. The same agreement pays for improvements and renovations at the site, and because the incremental capacity sits inside Amazon's 690 megawatts rather than beside them, the 190 megawatts are the piece of the contract that requires construction.
The physical arrangement barely changes. Constellation says all of Calvert Cliffs' output continues to flow to PJM, the largest grid operator in the country, which dispatches wholesale electricity across all or parts of thirteen states and the District of Columbia, and PJM sits between the two contracts: the plant's electrons go into the pool and Amazon's facilities draw from it. What the buyer has secured is a long-dated claim on output and clean-energy attributes, not a co-located supply deal of the sort hyperscalers have struck at new data campuses; a second contract, a retail supply agreement with Constellation, covers Amazon's own operations inside PJM, and the companies say the two together help it manage energy costs at facilities in the region.
No price, no delivery date and no cost for the uprate work appears in the announcement, which leaves the central capital questions open: what the 190 megawatts cost to build, when they start producing, and what Constellation earns on them once the contract is signed. A 20-year offtake is the instrument that would make that spending financeable for a merchant generator, and the term is the number that matters most in the release.
A 20-year offtake is the instrument that would make that spending financeable for a merchant generator, and the term is the number that matters most in the release.
The buyer's stated rationale is as much about the grid as about the balance sheet. Kerry Person, Amazon Web Services' vice president of global operations and data center delivery, framed the agreement as sustaining the continued operation and expansion of Maryland's largest source of carbon-free energy, supporting hundreds of jobs and bringing significant new generation to the regional grid, and Constellation's own language goes further in saying the deal shows how large energy users can underwrite new, reliable generation while strengthening the electric system.
Eighty percent of a state's clean energy from one site is a concentration that gives the contract a public dimension most corporate power deals lack, because the buyer, the operator and the state are all exposed to the same reactor. The coverage does not say how the retail supply agreement interacts with the PPA, or how much load Amazon carries in PJM.
Constellation's nuclear queue
Calvert Cliffs is the third hyperscaler-adjacent nuclear contract in this sequence: Meta's agreement, dated to last year, gives the company access to 1,121 megawatts of zero-emissions power for its AI operations beginning in June 2027 and, by Constellation's account, helps keep an Illinois reactor online, while Walmart followed earlier this year with a long-term contract covering nuclear energy, environmental attributes and capacity for its Illinois operations. Amazon's is the only one of the three described as adding capacity; the other two sustain reactors that already run. The headline volumes are not directly comparable, since Meta's 1,121 megawatts is new supply beginning in 2027 while part of Amazon's 690 already exists, but the direction of the buying is the same.
That pattern supports a position this publication has argued: the transition premium is accruing to firm, dispatchable assets while intermittent generation gets re-underwritten as development inventory. A 20-year wind or solar contract does not convert a project into a contracted asset in the same way; at Fagerasen, when Amazon took 199 megawatts of Swedish wind, the merchant tail stayed with the seller, the piece the buyer declined to take. At an operating reactor the output already exists, and what 20 years fix is the price of it and the ownership of the attributes.
These agreements also shift power rights without a fund standing between the parties, the shape three corporate contracts took in a single week in September. Here the duration an infrastructure vehicle would ordinarily hold sits with the offtaker, and the credit behind the revenue line is a hyperscaler's. Whether that displaces fund capital in contracted generation or gives owners a deeper second bid is not something a single PPA settles, though the nuclear queue is starting to look less like a run of one-offs.
Maryland's 80% share of clean energy comes from the same plant either way; what has changed is who agreed to buy the next 190 megawatts of it. The coverage gives no date for the uprate, while the only delivery date in the material belongs to Meta's June 2027 start, and the next test of the model is whether a hyperscaler will put the same 20 years behind a reactor that has never produced a megawatt.
Save this analysis and keep the funds you follow together in My Desk.
Sign in to save articles or follow funds.