A few years ago, declaring yourself a nuclear optimist in clean energy circles was a good way to get dismissed. The meltdowns at Fukushima were still fresh. Vogtle's cost overruns were becoming a cautionary tale. Renewables were getting cheaper by the quarter. Nuclear felt like a technology that history had already judged.
That consensus has cracked open in a striking way. Governors from both parties are now signing nuclear expansion bills. Federal regulators approved the first new reactor design in years. Microsoft signed a deal to restart Three Mile Island. Hyperscalers are quietly racing to secure nuclear power agreements because their AI data centers are hungry for reliable, around-the-clock electricity that solar and wind simply cannot provide on their own. Nuclear has gone from political liability to political asset, and the shift happened faster than almost anyone predicted.
What Actually Changed
The short answer is demand. The longer answer involves a complicated mix of climate urgency, grid reliability concerns, and the raw electricity appetite of artificial intelligence infrastructure.
U.S. data center power demand is projected to roughly double by 2030, according to Goldman Sachs research. Grid operators in PJM, MISO, and other regional markets are warning that retiring coal and gas plants are creating capacity gaps that wind and solar, for all their cost advantages, cannot fill without massive battery storage buildouts that don't yet exist at scale. Nuclear, which runs at around 90 percent capacity factor year-round regardless of weather, suddenly looks a lot more attractive to grid planners who have to keep the lights on.
The political chemistry shifted too. Climate-focused Democrats who once opposed nuclear have softened, partly because the math on decarbonization keeps getting harder without it. Republicans, meanwhile, have found nuclear to be a rare energy issue they can embrace enthusiastically: domestic, reliable, and not a solar panel made in China. That unusual alignment created space for the Advance Act, which passed with overwhelming bipartisan support in 2024 and directed the Nuclear Regulatory Commission to streamline its licensing processes.
The License Approvals Making Headlines
The NRC's approval of Kairos Power's fluoride salt-cooled reactor design earlier this year marked a genuine milestone. Kairos isn't just a paper reactor; the company is actively building a demonstration unit in Tennessee and has a contract with Google to supply power from commercial plants in the early 2030s. The fluoride salt design runs at lower pressure than traditional light-water reactors, which its proponents argue reduces the risk of the kind of catastrophic pressure-driven accidents that define public fear around nuclear power.
TerraPower, Bill Gates' nuclear venture, broke ground on its Natrium reactor demonstration project in Wyoming. Natrium pairs a sodium-cooled fast reactor with a molten salt energy storage system, which lets the plant modulate its output to complement variable renewables. The Wyoming site is deliberate: it's a coal community that agreed to host the project, and the state government has been aggressive in positioning itself as a nuclear-friendly destination.
X-energy, NuScale, and several other small modular reactor developers are at various stages of licensing, financing, and site selection. NuScale had a high-profile stumble when its Utah project collapsed over cost escalation in 2023, and that failure still haunts SMR skeptics who argue these designs haven't actually proven they can be built on budget. The criticism is legitimate. But the pipeline of competing designs is wider now than it was five years ago, and the companies learning from NuScale's mistakes represent a second generation of attempts.
The Money Coming In
Federal dollars have been a significant accelerant. The Department of Energy's loan guarantee programs, expanded under the Inflation Reduction Act, have made financing available for advanced nuclear projects that private capital alone would not touch at this stage. The IRA also extended and expanded the Production Tax Credit for nuclear, giving existing plants a lifeline and new plants a clearer revenue model.
Private investment has followed. Oklo, backed by Sam Altman, went public via SPAC and has a pipeline of power purchase agreements with data center operators. Helion, working on fusion rather than fission, raised $500 million and signed a deal with Microsoft. Fusion remains decades away from commercial reality by most honest assessments, but the willingness of sophisticated investors to write large checks signals how seriously the electricity supply problem is being taken.
Utilities are behaving differently too. Dominion Energy has applied to extend the license of its North Anna plant. Pacific Gas and Electric successfully extended Diablo Canyon's operating license after California reversed course on its planned closure. That reversal was politically remarkable: California, the state that once led the charge against nuclear power, decided it needed the 2.2 gigawatts Diablo Canyon provides as it works to hit aggressive clean electricity targets.
The Serious Problems That Haven't Gone Away
Nuclear's renewed political popularity doesn't make its economics easier. Vogtle Units 3 and 4 in Georgia came online years late and roughly $17 billion over their original budget, eventually costing about $35 billion total. That experience wasn't unique to Georgia Power or to Westinghouse's AP1000 design; it reflected the systemic atrophy of nuclear construction expertise in the U.S. over 40 years of near-zero activity. The supply chain for reactor components is thin. The workforce of welders, pipefitters, and nuclear engineers who know how to build these plants is small and aging.
Proponents argue that SMRs solve this by enabling factory manufacturing and modular assembly, reducing on-site construction complexity. That argument is theoretically sound. Whether it holds up when actual projects are actually built is the open question that the industry has not yet answered in practice. NuScale's failure was instructive precisely because the cost projections that made the project viable kept revising upward as detailed engineering work proceeded.
Waste remains genuinely unresolved. The U.S. still has no permanent repository for high-level nuclear waste. Yucca Mountain has been politically dead for years, and the interim storage sites holding spent fuel across the country are running out of room. New reactor designs produce different waste profiles, sometimes less, sometimes more exotic, but the absence of a long-term disposal solution is a real liability that doesn't disappear with optimistic press releases.
Safety concerns, while statistically manageable when put in context of deaths per terawatt-hour compared to fossil fuels, carry unique psychological weight. A gas plant explosion is terrible. A nuclear incident creates contaminated land that is unusable for generations. That asymmetry in consequence shapes public perception in ways that clean energy advocates have never fully figured out how to address.
What This Means for the Grid and for Consumers
The practical implications depend heavily on timeline. Existing nuclear plants being kept open or license-extended are the most immediately meaningful piece of the picture. Every gigawatt of existing nuclear capacity that stays online is a gigawatt that doesn't need to be replaced by gas, batteries, or new construction. Diablo Canyon alone represents roughly 8 percent of California's electricity generation.
New plants, whether large conventional reactors or SMRs, are a 2030s story at the earliest for most projects, and realistically a late 2030s story for widespread deployment. That's not a reason to dismiss them; the decisions made now determine what the grid looks like in 2040. But consumers expecting near-term rate relief or near-term carbon reductions from new nuclear construction will be disappointed.
The corporate power purchase agreement trend is worth watching closely. When Google, Microsoft, and Amazon sign direct agreements with nuclear operators, they're effectively taking on some of the demand risk that makes utility-scale nuclear financing difficult. If hyperscalers become anchor customers for new nuclear capacity, they could change the economics in ways that traditional utility financing never could.
A Genuine Inflection Point
Nuclear power is not going to replace wind and solar. The clean energy grid of 2050 almost certainly involves large amounts of both, along with storage, transmission upgrades, and demand flexibility. The more interesting question is whether nuclear can carve out a specific, valuable role as firm low-carbon capacity that keeps the grid stable when the sun isn't shining and the wind isn't blowing.
The political conditions for that to happen are better now than they've been in 30 years. The financial interest is real. The technology pipeline is deeper. None of that guarantees success; the history of nuclear in America is littered with projects that made sense on paper and collapsed under the weight of execution. But dismissing nuclear's comeback as a mirage would require ignoring a substantial amount of evidence that something real is happening here, even if the full shape of it won't be clear for another decade.