Amazon signs 20-year PPA for Calvert Cliffs nuclear uprate
The 690 MW purchase supports a 190 MW addition due online between 2030 and 2032 and Constellation's bid to relicense the 1,790 MW Maryland plant.
Amazon will buy 690 megawatts of carbon-free power from Constellation Energy's Calvert Cliffs nuclear plant under a 20-year contract that adds 190 MW of capacity and supports more than $3 billion in Maryland infrastructure investment. The new capacity comes online between 2030 and 2032, and Constellation says the 1,790 MW plant, which produces about 80% of Maryland's carbon-free energy, can be relicensed for another 20 years while the company develops new clean energy plants at the same site.
Announced alongside the PPA was a related retail supply agreement that will support Amazon operations across PJM, the grid operator for 13 states and the District of Columbia. The two contracts do different work: one holds output from a single asset under a single buyer for two decades, the other serves Amazon's buildings across the market where that power sits. Constellation president and chief executive Joe Dominguez called the PPA evidence of what private money can do with critical infrastructure, saying the commitment "supports the long-term operation of Calvert Cliffs for generations to come and creates a strong foundation for future investment in both the facility and in advanced nuclear technologies." Kerry Person, Amazon's vice president of AWS global operations and data center delivery, described the agreement as sustaining Maryland's largest source of carbon-free energy, supporting hundreds of jobs and bringing new generation to the regional grid.
What a 20-year buyer buys
Uprates and license extensions sit at the far end of the duration curve in power generation: merchant markets price electricity in horizons far shorter than these assets take to pay back, so a reactor owner weighing a relicensing filing carries that mismatch on its own balance sheet. The revenue certainty of a 20-year buyer is what converts a relicensing decision into a financeable capital project. In this instance the counterparty is a corporate procurement book with a 2030 target attached to it rather than a government program.
The other template for nuclear risk capital is the $1.9 billion Department of Energy loan reported in September that put the federal balance sheet in the first position on restart risk at the Duane Arnold plant. At Calvert Cliffs the money is arranged around an offtake instead, and the announcement describes no federal participation, a difference that matters for how the next tranche of nuclear capacity gets financed—a federal credit desk can only process so many applications, while a hyperscaler with a decarbonization mandate and load growth to feed can sign repeatedly.
Amazon's procurement scale makes it a credible counterparty for a 20-year obligation: the company contracted 10.22 GW of clean energy in 2025, one of the largest corporate volumes in the world according to BloombergNEF, and has invested in more than 700 renewable energy projects representing over 40 GW of carbon-free capacity. Its 2019 pledge covers all electricity consumed across global operations, data centers, corporate buildings, stores and fulfillment centers, matched with renewable energy by 2030. The nuclear deals are described by the companies as a step beyond renewable energy toward carbon-free power, a distinction that shows up in the executive titles attached to the announcement, where data center delivery sits alongside the sustainability language.
From annual matching to firm output
The arc of Amazon's offtake book has been visible in our coverage all year. In August, four Swedish wind PPAs signed by the company served as the revenue backstop for three Mirova-owned Swedish wind farms and a fourth still under construction—the same instrument working on a different technology, with the contract doing the work that a bank covenant would otherwise do. Nuclear changes the shape of the output: a wind contract matches an annual volume against an annual consumption figure, while a reactor delivers firm megawatts around the clock, which is the profile a data center portfolio actually draws.
Load is driving that shift, and Climate Week's opening day put the loudest number of the session at 118 GW of projected data-center demand, against a day whose only funded commitment was $300 million of Nigerian public money for off-grid power. Corporate buyers are not uniform in how they respond—Microsoft cut its carbon-removal purchases by 80% through mid-July, shrinking a market it had anchored—but power procurement is being treated differently from credit purchasing, because a data center cannot be served by an offset.
Calvert Cliffs becomes a test case because of its schedule. The uprate is scheduled to land between 2030 and 2032, well inside the window Amazon has set for its own target, and the relicensing Constellation says the deal enables has not been granted. Between now and the first new megawatt, the contract has to survive equipment lead times, a licensing docket and whatever PJM's capacity market does in the interim. The dollars are committed; the steel is not.
For the transition finance market, the more portable question is whether the structure travels. A 20-year PPA from a creditworthy buyer is a financing instrument that does not depend on a green bond label, a sovereign guarantee or a development bank's syndicate—the attributes that have carried most nuclear and grid capital to date. If utilities can sell two decades of firm output to corporate buyers directly, the underwriting constraint on life extensions and uprates shifts from the public balance sheet to the procurement departments of a handful of technology companies. That is a narrower set of counterparties than a bond market, and a more concentrated one.
The revenue certainty of a 20-year buyer is what converts a relicensing decision into a financeable capital project.
Save this analysis and keep the funds you follow together in My Desk.
Sign in to save articles or follow funds.