Costs and timelines: Vogtle and the SMR bet

Safety used to be the main argument against nuclear power in the United States. Today it is cost. Reactors are cheap to run once built: fuel is a small share of the bill and they run more than 90 percent of the time. Building them is the problem. The last two finished in the U.S. took about 15 years from the start of construction in 2009 to full operation and more than doubled their budget.

Vogtle 3 and 4

Plant Vogtle near Waynesboro, Georgia, already had two reactors from the 1980s. In 2009 Georgia Power and three co-owners began building two Westinghouse AP1000s, the first new U.S. reactor construction in about 30 years. The design was supposed to fix the industry's old problems: standardized, simplified, with passive safety and modules built in factories.

	Original plan	What happened
Total cost	About $14 billion	About $31 billion for the owners (AP calculation), near $35 billion counting $3.7 billion Westinghouse paid the owners
Unit 3 in service	2016	July 2023
Unit 4 in service	2017	April 2024
Capacity	About 1,100 MWe each	About 1,114 MWe each; with four units Vogtle is the largest U.S. nuclear plant, at nearly 5 GW
Contractor	Westinghouse	Westinghouse filed for bankruptcy in 2017

Georgia Power customers pay through their bills. AP reported that the final rate increase added about $8.95 a month to a typical residential bill, on top of $5.42 a month added when Unit 3 started. Southern Company, Georgia Power's parent, estimated a shareholder loss of about $3 billion. In South Carolina, the sister project at V.C. Summer, two more AP1000s, was abandoned in 2017 after about $9 billion was spent. The NRC terminated its licenses in 2019. Customers there are still paying for reactors that will never run.

Why nuclear projects overrun

The best-known study of this is "Sources of Cost Overrun in Nuclear Power Plant Construction Call for a New Approach to Engineering Design" by Jessika Trancik, Jacopo Buongiorno and colleagues at MIT, published in Joule in 2020. Looking at U.S. plants over decades, they found the largest driver of rising costs was indirect or "soft" costs: engineering, planning, supervision and the costs of crews waiting during delays. Safety regulations added cost, they found, but were one factor among several. And their most surprising finding was that later plants of an existing design often cost more, not less, than the first, the opposite of the learning curve industry expected.

Vogtle shows most of the usual problems:

• Design not finished before construction: work began while parts of the design were still changing, which forced rework.
• Supply chain atrophy: after decades without new U.S. reactors, few suppliers had current nuclear quality experience. Modules built off site arrived with defects.
• Workforce: crews had to learn nuclear-grade construction on the job; at the peak, thousands of workers were on site.
• Contract structure: fixed-price deals pushed risk onto Westinghouse until it went bankrupt, then risk came back to the utility and its customers.
• Financing: every year of delay adds interest. On a $30 billion project at 6 percent, a year of delay can add well over a billion dollars in interest alone.

Pro-nuclear analysts also point to regulation. They argue that the NRC's process, frequent rule changes during construction and the cost of documenting every safety-grade component add cost without matching safety benefit. Anti-nuclear and consumer groups answer that the overruns at Vogtle were mostly construction management failures, and that the problem is the technology's complexity. Both camps agree first-of-a-kind projects are where the money gets lost.

The SMR bet

Small modular reactors are the industry's answer to Vogtle. The arguments on each side are worth laying out in their strongest form.

What SMR advocates argue	What skeptics argue
Smaller projects mean less money at risk per project; a utility can add capacity in 300 MW steps instead of betting billions on one unit	Reactors historically got cheaper per kW by getting bigger. Smaller units lose economies of scale in containment, staff and security
Factory production brings learning: each unit cheaper than the last	Factory savings need a steady order book of dozens of units; there is no such order book yet, and the MIT study found little historic learning
Passive safety and smaller cores reduce the need for expensive safety systems and large emergency zones	Early SMR estimates, like NuScale's in 2023 and Darlington's, came in at $18,000 to $20,000 per kW, higher than Vogtle
Tech companies will pay premium prices for firm, carbon-free power, which absorbs first-of-a-kind costs	If tech buyers pay the premium, fine; if costs land on ordinary utility customers, they carry the risk
Sites at retired coal plants reuse grid connections, water rights and trained workers	Advanced designs need HALEU, which is scarce, and many have never operated at commercial scale

Groups that make the skeptical case include the Institute for Energy Economics and Financial Analysis (IEEFA), which tracked the NuScale cost increases, and the Union of Concerned Scientists. The advocacy case comes from the Nuclear Energy Institute, think tanks such as the Breakthrough Institute, and DOE's own "Pathways to Commercial Liftoff" report, which argues that a committed order book of five to ten deployments of at least one reactor design is what it would take to drive costs down, and sets a goal of on-time, on-budget delivery within plus or minus 20 percent.

Where does the evidence point? Clearly on one thing: first-of-a-kind reactors in North America have been expensive, and no SMR has yet shown the cost declines its backers predict, because none has been built in series. Whether the learning shows up is the test of the next decade. Executive Order 14302 (May 2025) directed DOE to help get 10 new large reactors with complete designs under construction by 2030 and 5 GW of uprates at existing plants, a signal that Washington is betting on both large and small reactors.

Who carries the risk

For residents and ratepayers, the money question is usually more important than the technology question. Ask who pays if it runs over.

• Regulated utility customers: in states that allow it, utilities can charge customers during construction (construction work in progress, or CWIP), as Georgia did. That cuts financing costs but shifts risk to ratepayers.
• Public power: TVA, municipal utilities and co-ops pass costs to their members. The Utah cities in the NuScale project were exposed until they withdrew.
• Tech company buyers: a fixed-price PPA puts some risk on the buyer, but the details matter: who pays if the plant is late?
• Taxpayers: DOE loans, grants (like the up to $400 million in cost-shared funding DOE selected TVA for at Clinch River) and tax credits for nuclear production and investment spread cost to the public.
• Investors: merchant developers and venture-backed startups carry risk directly, which is why their schedules depend on raising capital.