The Duane Arnold Deal Is Just the Beginning of the Nuclear-Tech Pivot

A Restart With a Template

The United States Department of Energy has approved a $1.9 billion loan to NextEra Energy for refurbishing the Duane Arnold Energy Center in Iowa. Google is the end buyer of most of that power. This is only the second loan of its kind the agency has issued in recent months—the first went $1 billion to Constellation Energy for a Three Mile Island reactor—and it carries a signal more important than the dollar figure: the federal government is now explicitly subsidizing nuclear restarts to serve tech-sector demand.

That distinction matters. This is not a broad energy subsidy. It is a targeted industrial policy that ties public money to private load. The Trump administration, which inherited this policy trajectory from its predecessor, has made clear it views revived nuclear capacity as a strategic resource for powering the AI infrastructure buildout. Two loans, two tech giants, two shuttered reactors. The pattern is deliberate.

Why Duane Arnold Is Different

The Duane Arnold deal is smaller than the Three Mile Island arrangement. The Iowa plant will deliver 615 megawatts after refurbishment, up 14 megawatts from its pre-mothball output. That capacity will mostly feed Google’s planned data center cluster in the region. A mere 50 megawatts will go to the local cooperative—roughly 18 percent of Iowa’s electricity demand growth since 2021, a metric that tracks the quiet acceleration in state load long before ChatGPT made headlines.

But size is not the point. The significance lies in what the deal proves. A nuclear reactor that sat idle for four years—shuttered after storm damage in 2020 when natural gas prices made revival uneconomic—can now come back online with federal backing and a corporate off-taker signed in advance. The risk profile has flipped. What looked like a stranded asset in 2021 looks like a de-risked power purchase agreement in 2026.

The Bigger Pattern

Duane Arnold is the third data-center-linked nuclear restart in active discussion. Constellation’s Three Mile Island Unit 1 will return in 2028 with 835 megawatts under a Microsoft agreement. Meta’s deal with Constellation for the Clinton Clean Energy Center in Illinois secures the output of a 1.1-gigawatt plant that was on the brink of closure; Clinton’s electrons will feed the Illinois grid while Meta claims the clean-energy certificates to offset emissions from sites like its planned Hyperion data center in North Carolina—a facility expected to consume more electricity than the state of South Dakota.

Taken together, these three arrangements represent the easiest nuclear recoveries available. Duane Arnold, Three Mile Island, and Clinton are all plants with recent operating histories, existing grid connections, and proven regulatory pathways. Utility Dive has suggested there may be one or two additional candidates in the United States, but facilities like California’s San Onofre have been offline longer and would require substantially more capital and regulatory effort to revive.

The low-hanging fruit is being picked. The question that follows is what happens next.

The Real Winners and Losers

The winners are obvious. NextEra and Constellation have a new revenue stream from assets that were at risk of permanent retirement. Google, Microsoft, and Meta have secured long-duration, carbon-free power that solar and wind cannot guarantee on their own. The Department of Energy has a portfolio of projects that demonstrate its lending authority in action. And Iowa, Pennsylvania, and Illinois each gain industrial anchors and jobs tied to reactor refurbishment.

The less obvious loser may be the broader ratepayer base. Federal loans reduce the financing cost for these restarts, but they also commit public resources to projects whose primary beneficiary is a handful of technology companies. The 50 megawatts set aside for Duane Arnold’s local cooperative is a concession, not a model. Most of the power from these reactors will travel to data centers, not households. The political economy of who pays for clean firm power and who actually receives it will become a friction point as more deals of this type are structured.

What Comes Next

The 2029 restart date for Duane Arnold gives us a timeline to watch. If the project proceeds on schedule, it will establish a repeatable playbook: identify a recently idled reactor, secure a tech off-taker, obtain a DOE loan, complete refurbishment, and resume commercial operation. Constellation’s Three Mile Island restart in 2028 will test whether the model scales. Meta’s Clinton arrangement will test whether it works for plants that are still technically operating but economically fragile.

Beyond those three, the pipeline is thinner. New nuclear construction in the United States has moved slowly for decades, and small modular reactor promises have yet to translate into commercial operation. Restarting existing plants is the fastest path to additional carbon-free baseload capacity, and the tech industry’s demand for it is accelerating. Electricity consumption from data centers is projected to nearly triple by 2035, according to industry estimates. That demand will not be met by solar and wind alone, and it will not be met by natural gas alone if the sector is serious about emissions targets.

The Duane Arnold loan is a $1.9 billion bet on that reality. The bet is that the template works, that more plants can be revived along similar lines, and that the convergence of federal energy policy and corporate power procurement can close the gap between AI’s appetite for electricity and the grid’s ability to supply it cleanly. Whether that bet pays off for everyone beyond the three companies and two utilities directly involved remains the question the next three years will answer.