America's Nuclear Comeback: Between Vogtle's Lessons and the Small Reactor Promise

On April 29, 2024, Georgia Power's Plant Vogtle Unit 4 entered commercial operation, becoming the second new nuclear reactor to begin operating in the United States in the 21st century. Its twin, Unit 3, had started commercial operation nine months earlier. Together, they represent 4,536 megawatts of around-the-clock, carbon-free generating capacity — the largest nuclear facility in the country and the first new commercial reactors built in America in thirty years. The achievement deserves acknowledgment. So do the numbers that came with it.

The original construction budget for Vogtle Units 3 and 4, when Georgia Power and its partners made the final investment decision in 2012, was approximately $14 billion. The actual cost was $36.8 billion. The scheduled completion date was 2017. The actual completion date was 2024. The project's contractor, Westinghouse, filed for bankruptcy during construction. Two of Georgia Power's co-owners withdrew from the project entirely. The Southern Company, which owns Georgia Power, absorbed billions in cost overruns that were ultimately passed through to ratepayers.

Vogtle, in other words, is simultaneously America's most impressive recent energy infrastructure achievement and its most cautionary tale about large-scale nuclear construction in the current regulatory and labor environment.

What Changed, and What Hasn't

Nuclear energy's fundamental value proposition has not changed: a reactor produces large amounts of electricity continuously, regardless of whether the sun is shining or the wind is blowing, for 60 to 80 years. In a grid increasingly dependent on variable renewable generation, that characteristic — technically called "dispatchable firm capacity" — is worth more than it was a decade ago, when planners had not yet confronted the full implications of a high-renewables grid at scale. The North American Electric Reliability Corporation now explicitly identifies the retirement of firm capacity as a primary reliability risk. Nuclear plants, by definition, do not have that problem.

What has also changed, more quietly, is public opinion. A 2025 Gallup survey found that 61% of Americans support nuclear energy — near a record high. More striking is the partisan distribution: support among both Republicans and Democrats has climbed from 43% in 2020 to approximately 60% in 2024-2025. Nuclear has become, improbably, one of the least partisan topics in American energy policy.

What has not changed is the economics of large, light-water reactor construction in the United States. Vogtle's cost trajectory — from $14 billion to $36.8 billion — reflects structural challenges that are not unique to a single project: a degraded nuclear construction workforce, a regulatory framework designed for serial production that now applies to one-of-a-kind projects, and supply chains that have been dormant for decades. Building another AP1000 at another greenfield site today would not automatically be cheaper than Vogtle. The lessons from the project are valuable. They do not erase the cost reality.

The SMR Thesis

This is precisely the context in which small modular reactors have attracted serious attention. The SMR thesis, stripped to its essentials, is that the problems of large nuclear construction — bespoke engineering, site-specific regulatory review, massive upfront capital — can be addressed by designing smaller reactors that can be standardized, factory-manufactured, and deployed in multiples at a single site. The economics of manufacturing, rather than the economics of construction, would govern cost.

NuScale Power received NRC design certification for its 77 MWe module in January 2023 — the first SMR to receive U.S. regulatory approval — and followed with an upgraded design approval in June 2025. TerraPower, backed by Bill Gates, broke ground on its 345-megawatt Natrium reactor in Kemmerer, Wyoming in June 2024, targeting the site of a retiring coal plant. X-energy, backed by a $500 million investment round led by Amazon's climate fund in October 2024, is developing a high-temperature gas-cooled reactor designed to provide both electricity and process heat for industrial decarbonization.

"The nuclear industry is not repeating the last generation's mistakes. The question is whether the new generation of designs can actually deliver on the cost and schedule projections that make the economics work — projections that, historically, have rarely survived first contact with actual construction." — Senior energy analyst, Columbia University Center on Global Energy Policy

A Cautious Optimism

We do not think nuclear is the right answer to every grid problem, and we are skeptical of the more bullish projections for SMR deployment timelines. The NuScale project originally planned for Idaho National Laboratory was canceled in 2023 when projected costs escalated sharply — a reminder that the SMR cost revolution has yet to be demonstrated in actual construction. TerraPower's Natrium reactor will not be operational before 2030. X-energy's first commercial plant is further out still.

What we do believe is that the grid reliability case for nuclear has strengthened materially, and that dismissing nuclear as an economic non-starter because of Vogtle is as intellectually sloppy as dismissing solar because of Solyndra. The policy question is not whether to prefer nuclear over renewables — it is whether the regulatory, financing, and supply chain conditions can be created to make new nuclear economically viable alongside an expanding renewable fleet.

The answers to those questions will emerge over the next decade as the first SMR projects either deliver on their promises or, as has happened before, miss them. The grid needs that question answered. With data center demand growing at a pace that no reasonable renewable-plus-storage combination can fully satisfy on its own, the margin for error in U.S. capacity planning is narrowing. Nuclear, for all its difficulties, remains one of the few technologies capable of providing the sustained, weather-independent capacity the grid will need.