Why Utilities Are Finally Warming Up to Nuclear After Decades of Skepticism
After forty years of running scared from the atom, America’s electric utilities are back at the table — and this time, they’re not just talking.
For most of the past four decades, asking a utility executive about nuclear power was a bit like asking someone who once burned their hand to describe their love of bonfires. The instinct was to back away slowly. Three Mile Island in 1979. Chernobyl in 1986. The spectacular, soul-crushing cost implosions at plants that took decades to build and billions more than promised. Utilities learned a brutal lesson: nuclear is the kind of technology that sounds great until the bill arrives. So they built gas plants instead, signed wind contracts, and left nuclear to the governments and the dreamers.
That calculation is changing. Not dramatically, not overnight, but with the kind of deliberate, unmistakable momentum that tends to precede a genuine shift in a multi-trillion-dollar industry. The Tennessee Valley Authority became the first American utility to file a construction permit application for a small modular reactor in May 2025. Duke Energy is partnering with GE Vernova Hitachi to advance the BWRX-300 design for its Carolinas footprint. And Constellation Energy is literally flipping the lights back on at Three Mile Island — arguably the most famous symbol of nuclear’s failure — to sell power to Microsoft. You probably want to sit with that irony for a moment.
Something real is happening. But why now, after so many false starts? The answer involves artificial intelligence, a hard economics lesson the industry finally absorbed, and a new generation of reactor designs that actually seem designed with utilities’ nightmares in mind.
The long winter, and why utilities froze
It is worth being honest about how bad it got. Nuclear’s collapse in the United States wasn’t just a PR problem — it was a financial catastrophe that nearly swallowed several utilities whole.
The most recent cautionary tale is still fresh. Georgia Power spent $30 billion and eleven years building two new reactors at Plant Vogtle, completing them in 2023 and 2024. That sounds like progress until you notice the 25% rate increase that landed on Georgia customers’ doorsteps, and the two public service commissioners who subsequently lost their jobs at the ballot box. The political fallout was real. As Utility Dive reported, elected officials discovered that voters do not reward nuclear enthusiasm — they punish the electricity bills that come with it. ⚠️
Before Vogtle, there was the SCANA disaster in South Carolina, where the utility and Westinghouse made false claims of progress on twin reactors, eventually resulting in criminal charges and the project’s abandonment after billions were already spent. The wound from that one still hasn’t fully healed in the industry.
The deeper history is more damaging still. Three Mile Island’s partial meltdown in 1979 triggered a regulatory crackdown that made new nuclear almost impossible to approve economically. Then Chernobyl in 1986 turned the public mood to permafrost. By the 1990s:
129 nuclear plants had been approved for construction before TMI; fewer than half were ever completed 🚫
No new utility-scale reactors broke ground in the U.S. for over 30 years
Construction costs ballooned as safety regulations multiplied, layering complexity onto complexity
Gas plants, by comparison, could be built in two or three years at a fraction of the upfront cost
Utilities are rational creatures. They respond to incentives and fear losses. For forty years, nuclear offered more of the latter. So they walked away. 🚶
The electricity demand earthquake changed everything
Ask a utility executive today what’s keeping them up at night, and the answer isn’t climate policy or fuel prices. It’s load growth — specifically, the absolutely staggering electricity appetite of AI data centers. ⚡
The numbers are almost surreal. OpenAI has suggested it may need 20 gigawatts of power to accommodate ChatGPT’s 700 million weekly users. Amazon signed a deal with Talen Energy for 1,920 megawatts of nuclear power from the Susquehanna plant through 2042 to support its AWS data centers. Google committed to buying power from Kairos Power’s small modular reactors, targeting 500 MW online by 2030. Microsoft is funding the restart of Three Mile Island for the same reason. These aren’t press releases — they’re power purchase agreements worth billions. 💰
The reason these companies are chasing nuclear specifically, rather than just buying more solar, comes down to one word: firmness. Data centers can’t run on weather. They need electrons at 3 a.m. on a cloudy Tuesday in January, not just on sunny summer afternoons. Renewables are wonderful and cheap, but they require storage backup to be truly dispatchable — and grid-scale storage at the terawatt-hour scale isn’t commercially available yet. Nuclear runs around the clock at 90%+ capacity factors, regardless of what the wind is doing. For hyperscale computing, that reliability isn’t a nice-to-have. It’s the whole point.
Think about it this way: utilities suddenly have customers — very large, very solvent customers — who want nuclear power and will sign 20-year purchase agreements to get it. That changes the financial math fundamentally. The old problem was that utilities had to take all the risk of building an expensive plant and hope ratepayers would eventually benefit. Now there are corporate buyers who will essentially pre-purchase the output, de-risking the investment before the first shovel breaks ground.
Does the current AI energy boom feel durable to you, or does it seem like a hype cycle that might cool? The utilities are clearly betting on the former, but it’s a question worth asking. 🤔
Small modular reactors rewrite the utility risk model
Here’s where the technology story intersects with the business story. Traditional large reactors — the 1,000-plus-megawatt behemoths like Vogtle — require massive upfront capital, take a decade to build, and carry enormous financial risk if anything goes wrong. That profile is almost perfectly designed to terrify a utility’s board of directors.
Small modular reactors are engineered to fix exactly those problems. The World Nuclear Association describes SMRs as representing “a rebalancing of historic economies of scale towards economies of series production” — factory-built, modular components that get shipped to site and assembled rather than constructed from scratch over years. The leading designs include:
BWRX-300 (GE Vernova Hitachi): 300 megawatts, targeting coal plant replacement sites, uses 60% less concrete and steel than traditional reactors, with a claimed 24-36 month construction timeline 🏗️
SMR-300 (Holtec): Two units planned for Michigan’s Palisades site, pursuing an unusual “one-stop-shop” model where Holtec acts as vendor, constructor, and operator
Xe-100 (X-energy): An 80-megawatt module typically packaged in groups of four, backed by Amazon’s $500 million investment
For utilities, the appeal is structural. Lower capital per unit means less financial catastrophe if a project runs over budget. A 300-megawatt plant going wrong is painful; a 1,400-megawatt plant going wrong is existential. TVA understood this, which is why it structured its advanced nuclear deals explicitly “to make sure that risk isn’t on our balance sheet,” according to CEO Don Moul. That sentence is doing a lot of work. It signals a new era of utility engagement with nuclear: interested, but insisting on better risk allocation. And that’s probably healthy. 🔬
The policy environment finally caught up
Nuclear has struggled for decades not just with economics but with regulatory timelines that add years — and billions — to project costs. The U.S. Energy Information Administration notes that SMR licensing is now advancing in 15 countries, with a 65% increase in pre-licensing activities since 2023. America is part of that wave, and its policy framework has finally started to reflect that.
Two successive administrations, Biden and Trump, have taken steps in the same direction:
The ADVANCE Act, passed with bipartisan support, is designed to streamline Nuclear Regulatory Commission approvals and reduce duplicative review
Trump signed four executive orders in May 2025 targeting 400 gigawatts of nuclear by 2050 — nearly four times current capacity 🚀
The Department of Energy launched a $900 million funding program for SMR deployment, awarding $400 million to TVA for the BWRX-300 at Clinch River and $400 million to Holtec for its Michigan project
TVA signed the first U.S. utility power purchase agreement for a Generation IV reactor in August 2025, a three-way deal with Kairos Power and Google for the Hermes 2 pilot plant 📋
The IEA’s path to a new era for nuclear projects that SMR deployment could reach 190 gigawatts globally by 2050 if construction costs come down to parity with large-scale reactors. That’s a big if. But the policy infrastructure to support that trajectory — tax credits, loan guarantees, production tax credits set at $30 per megawatt-hour — is now in place in a way it simply wasn’t five years ago.
The bipartisan nature of this shift is worth noting. Nuclear energy is one of the very few policy areas where you can get both parties to show up and agree. That doesn’t happen unless the underlying incentives are genuinely compelling. ✅
It’s still not a sure thing
Anyone writing a triumphalist piece about nuclear’s comeback is probably getting ahead of the evidence. The skeptics have legitimate points, and a serious article owes them some space.
NuScale — the only SMR company to win full NRC design approval — canceled its flagship Idaho project in 2023 after subscribers couldn’t absorb the projected costs. That was a genuine gut-punch to the industry’s credibility. The ITIF’s realist assessment of SMRs makes clear that power purchase agreements will likely be mandatory for every first-of-a-kind project, because no lender will take the risk without a committed buyer. Right now, the buyers are largely Big Tech — and if that demand ever softens, the financial scaffolding gets shakier. 🏗️
There are also genuine questions about how much the “small” in SMR reduces costs versus just reduces output. Factory fabrication sounds efficient in theory, but the industry hasn’t actually demonstrated it at scale yet. Only two commercial SMRs are operating anywhere in the world as of now. The global pipeline of 74 designs in development is exciting; the two that are actually running electricity into a grid are a more sober number.
The Vogtle cautionary tale also hasn’t faded. Customers got a 25% rate increase. Elected officials lost their seats. If the first few SMR projects run significantly over budget — which first-of-a-kind nuclear projects historically do — utilities will face the same political backlash all over again. The technology may be modular; the politics are not.
If you’re tracking the SMR sector, the next 18 months will be telling. Watch TVA’s Clinch River construction permit process, the Holtec Palisades restart, and whether Kairos Power’s Hermes plant actually generates power on schedule. Those are the real tests. 🔍
The reasons utilities are warming to nuclear are real, structural, and driven by genuine demand signals rather than ideology. But “warming up” is not the same as “problem solved.” The industry still needs to deliver on its promises, on something approaching on time and on budget, for the current enthusiasm to compound rather than collapse.
The question worth sitting with: is this the generation of utility executives who finally cracked the nuclear problem — or are we watching the early stages of a familiar cycle of optimism that the first cost overrun will deflate?