Glonce

US nuclear uprates could add 3 to 8 GW, a fraction of AI demand

infrastructureThe Glonce editorsMIT Technology Review🔥41Glonce rating41Impact40Novelty30Relevance55Surprise20Players40Credibility85
US nuclear uprates could add 3 to 8 GW, a fraction of AI demand

In a piece from The Spark, MIT Technology Review's weekly climate newsletter, the author looks at a wave of announcements about squeezing more power out of existing US nuclear plants. Constellation, the largest nuclear operator in the US, announced plans to add nearly a gigawatt of new capacity across 11 existing reactors at six plants. It plans to spend more than $4.3 billion on the uprates, and the extra electricity would be enough to power roughly 700,000 US homes. On October 5, the Department of Energy awarded a $4.2 billion loan to Vistra for previously announced upgrades at three plants; elsewhere the article says DOE will commit over $4 billion to help finance upgrades and modernization for Vistra plants in Pennsylvania and Ohio. Those projects will add 433 megawatts, and Meta has signed on to buy the electricity in a long-term deal.

An uprate is an equipment upgrade that lets a plant reach a higher capacity. US operators have done them since the 1970s, but they became less common as electricity demand plateaued and cheap natural gas put financial pressure on nuclear plants. Some plants closed, and at others spending more to produce more power made little sense. The US Nuclear Regulatory Commission approved over 80 uprates between 2001 and 2010, and 33 between 2011 and 2020. AI-driven demand has changed that: shuttered plants such as Palisades and Three Mile Island are racing to reopen, and existing plants are looking to uprate again.

Some uprates are minor, involving changes to instrument setpoints or calculations. The more major overhauls can boost reactor power by as much as 20% and require replacing equipment. Generally, an uprate uses fuel that is more highly enriched, or a higher volume of fresh fuel, to make more steam for the turbine. To handle the extra energy, the main turbines and equipment such as pumps, motors, generators and transformers are upgraded. The most recent major uprate approved in the US covered three reactors at Browns Ferry in Alabama: a $475 million project that expanded the plant's capacity by 14% between 2018 and 2019.

Big Tech money is behind the new push. Google agreed to buy electricity from the Constellation plants, and the 20-year agreement was a major factor in making the uprates happen. Constellation CEO Joe Dominguez told the New York Times that these are "capital intensive, labor intensive projects that require a different price point than the market trades at, frankly." On a per-kilowatt basis the Constellation project is cheaper than building a new reactor, but still more expensive than a new utility-scale solar farm with energy storage or a combined-cycle gas plant.

The scale is the catch. The US nuclear fleet has roughly 98 gigawatts of total capacity. Depending on who is asked, between three and eight additional gigawatts are on the table from uprates, and the NRC already expects applications to add 2.4 gigawatts through 2032. Against that, PJM, the grid operator covering much of the East Coast, expects about 30 gigawatts of additional peak demand because of AI by 2030. The author says all possible nationwide uprates would cover just a fraction of that. Uprates are called a relatively easy win when zero-emissions power is needed, but meeting the remaining demand with nuclear will eventually require building new reactors.

Key facts

  • Constellation plans nearly a gigawatt of new capacity across 11 reactors at six plants, for more than $4.3 billion, enough for roughly 700,000 US homes; Google's 20-year purchase agreement was a major factor.
  • On October 5 the DOE awarded Vistra a $4.2 billion loan for upgrades at three plants; the projects add 433 megawatts and Meta has signed a long-term deal to buy the power.
  • The NRC approved over 80 uprates in 2001 to 2010 and 33 in 2011 to 2020; major overhauls can raise reactor power by as much as 20%.
  • Of a roughly 98 GW US fleet, between three and eight additional gigawatts are available from uprates, depending on who is asked; the NRC already expects applications for 2.4 GW through 2032.
  • PJM expects about 30 GW of extra AI-driven peak demand by 2030, so uprates cover only a fraction and new reactors would eventually be needed.

Why it matters

Existing reactors are the quickest source of extra clean power for data centers, and the money now flowing to them comes from Big Tech buyers. Constellation's plan rests on a 20-year Google agreement, and Meta has signed on to buy power from the Vistra plants. The article's point is about limits: uprates are a relatively easy win, but the 3 to 8 gigawatts on the table sit beside PJM's expected 30 gigawatts of extra AI-driven peak demand by 2030. Meeting the rest with nuclear will eventually mean building new reactors.

Who it affects

Nuclear operators such as Constellation and Vistra, which are spending or borrowing billions on upgrades. Tech companies such as Google and Meta, whose long-term power purchases underpin the projects. The NRC, which reviews uprate applications and already expects applications to add 2.4 gigawatts through 2032. Grid operators such as PJM, which face about 30 gigawatts of additional AI-driven peak demand by 2030. Constellation's extra output would power roughly 700,000 US homes.

How to use it

The article is a briefing on the numbers, useful for anyone sizing up power supply for AI. The reference points: a roughly 98 GW fleet; 3 to 8 GW potentially available from uprates; 2.4 GW already expected in NRC applications through 2032; nearly a gigawatt from Constellation; 433 megawatts from the Vistra projects. On cost, the Constellation uprates are cheaper per kilowatt than a new reactor but dearer than new solar with storage or a combined-cycle gas plant.

How solid is it

The piece is an analytical newsletter item from MIT Technology Review, and most figures are stated plainly: Constellation's capacity and spending, the October 5 DOE loan, NRC approval counts, the Browns Ferry project and the fleet size. The 3 to 8 gigawatt range is explicitly 'depending on who you ask', and the article does not say how much of it is realistic. The article gives the Vistra financing as a $4.2 billion loan in one place and 'over $4 billion' in another. The per-kilowatt cost comparison is given in words, without figures. Dominguez's quote is secondhand, taken from his remarks to the New York Times.

Risks and caveats

The article does not say when the Constellation uprates will be completed or come online. It does not name the three Vistra plants, give the terms of the DOE loan, or give the length or price of the Meta deal with Vistra. Uprates cost more per kilowatt than new solar with storage or gas, as Dominguez's remark about needing a different price point suggests. The 700,000-homes figure applies only to Constellation's uprates, not to Vistra. And the central caveat: even all possible nationwide uprates would cover just a fraction of AI-driven demand growth.

“These are capital intensive, labor intensive projects that require a different price point than the market trades at, frankly.”

— Joe Dominguez, Constellation's CEO, to the New York Times