Why Did Plant Vogtle Cost So Much? Inside the Nuclear Overrun

Plant Vogtle's two new reactors cost more than about $35 billion, including financing, against an original estimate of about $14 billion, because construction began before the AP1000 design was finished, the US nuclear supply chain and workforce had to be rebuilt from scratch, the fixed-price vendor went bankrupt, and every year of delay added interest on billions already spent. Units 3 and 4 in Georgia entered service in July 2023 and April 2024, about seven years late. They work well and should run for most of this century, but they are among the most expensive power plants ever built in the United States.

The contrast that defines the case came fourteen months after Unit 3 started. On 20 September 2024 Constellation Energy agreed with Microsoft to restart Unit 1 at Three Mile Island, closed in 2019, at a cost of about $1.6 billion (company estimate, reported), roughly $1,900 per kilowatt. Vogtle's new units cost about $15,900 per kilowatt all-in. The gap between those two numbers, an eightfold difference between the value of a reactor that exists and the cost of building a new one, is the central fact of the nuclear business.

How much did Vogtle cost, and how late was it?

The cost basis matters. The $35 billion figure covers all four owners and includes financing; narrower measures that exclude some financing or cover only Georgia Power's share are lower, and most reported totals exceed $30 billion. Georgia Power owns about 45.7%; the rest belongs to Oglethorpe Power, a generation cooperative (30%), MEAG Power, a municipal agency (22.7%), and Dalton Utilities (1.6%). Georgia Power's own share of capital cost was forecast at about $10.2 billion (company).

Why did Vogtle go so far over budget?

The post-mortems on Vogtle, by MIT and the Georgia PSC staff, point to the same causes as France's own audit of its Flamanville reactor. They compound.

The design was not finished

Southern Company's utilities signed the construction contract with Westinghouse and Stone & Webster in April 2008. The Georgia Public Service Commission certified the project in March 2009, and the Nuclear Regulatory Commission issued combined construction and operating licences on 9 February 2012, the first for a new US nuclear plant since 1978. But amendments to the AP1000 design certification continued until 2011. Construction started before the design was complete, so every change cascaded through the schedule. Early problems included rebar placement that did not match the licence and late, non-conforming modules from a new factory in Lake Charles, Louisiana (reported).

The industry had forgotten how to build reactors

The United States had not completed a new reactor since Watts Bar 1 in 1996. Vogtle had to rebuild a supply chain for heavy components and train a craft workforce in nuclear-grade quality. Under nuclear rules, a defect found late is very expensive to fix, and quality findings stopped work more than once.

A fixed price from a vendor that could not pay

Westinghouse, then owned by Toshiba, converted the Vogtle and V.C. Summer contracts to fixed-price terms in 2015–16 and took over the construction business of its partner. It was losing billions on both projects. On 29 March 2017 it filed for Chapter 11. A fixed price from a vendor without the balance sheet to honor it, as the research puts it, "is not risk transfer; it is risk deferral." Toshiba paid about $3.68 billion under its parent guarantee to the Vogtle owners, in full by December 2017 (reported). That covered only part of the overrun.

Time is interest

Every year of delay added interest and overheads on billions of dollars already spent, before any revenue arrived. The book's chapter on nuclear power shows the effect with a stylized 2,200 MW plant at an overnight cost of $8,000 per kilowatt. The numbers are round, not Vogtle's.

With the same overnight cost, stretching construction from six to fourteen years at 8% adds about $9 billion before the plant earns anything. Real overruns are worse, because the overnight cost rises at the same time as the schedule slips.

The market moved underneath the project

The case for Vogtle was made when gas was expensive. Henry Hub averaged about $8.9 per MMBtu in 2008, and a carbon price looked likely. Shale then pushed gas to about $2.75 in 2012, and federal carbon legislation never passed. The reactors were justified against a fuel that became cheap while they were being built.

Why was Vogtle finished when V.C. Summer was abandoned?

South Carolina's V.C. Summer used the same AP1000 design and the same vendor. When Westinghouse failed, SCANA and Santee Cooper abandoned the two half-built units on 31 July 2017 after spending about $9 billion. Customers kept paying for them through rates for years, and SCANA was later bought by Dominion Energy in 2019.

Southern Company went the other way. Its subsidiary Southern Nuclear took over project management, collected the Toshiba guarantee quickly and hired Bechtel to manage construction. On 21 December 2017 the Georgia PSC voted to continue, while signaling that not every cost might be recoverable. The co-owners voted again to continue in 2018 after another cost increase.

Two things kept Vogtle alive. A 2009 Georgia law let Georgia Power collect financing costs from customers during construction, starting in 2011, so interest was paid as it was incurred rather than compounded. And the Department of Energy guaranteed about $12 billion of the owners' debt by 2019. That captive, patient capital is what a first-of-a-kind project needed, and it is what Summer's owners lacked.

Who paid for Vogtle?

This is how regulated utilities work. A utility earns a return on capital the regulator lets into its rate base, and laws that allow construction costs to be recovered before completion move the risk of overruns onto customers. That is the whole argument over such laws in one sentence. Our guide to reading a utility rate case explains the prudence and "used and useful" tests that decide how much of an overrun customers carry.

Is Vogtle worth it now?

The answer has changed with the market. In 2023 gas was cheap again, and a new combined-cycle plant would have produced far cheaper energy. But the reactors make firm, carbon-free power that can run for sixty to eighty years, and the existing US nuclear fleet runs at capacity factors of about 90–93% (EIA). From 2024, technology companies began paying premium prices for exactly that product. Constellation signed Microsoft for the Crane restart and Meta for Clinton's roughly 1,121 MW. In PJM, capacity prices rose from about $28.92 to about $329.17 per MW-day between 2024/25 and 2026/27, as our explainer on the PJM capacity auction shows. The long-run value of Vogtle's output rose. Its cost did not fall.

Unit 4 was built faster and cheaper than Unit 3 in its final phases (reported), which is the learning effect every first unit lacks. The industry now argues that the next AP1000s will be cheaper. In October 2025 the US government signed a term sheet with Westinghouse for at least $80 billion of reactors, and in June 2026 DOE issued a conditional $17.5 billion loan for long-lead AP1000 equipment (reported). Southern Company had not committed to new units by late 2026.

What do successful nuclear programs do differently?

Europe has its own Vogtles. EDF's Flamanville 3 reached the grid on 21 December 2024, seventeen years after construction began; Britain's Hinkley Point C was estimated at about £35 billion in 2015 money, with its first unit expected in 2030 (EDF, February 2026). The counter-examples share four features: a standard design built repeatedly, a stable team moving from one unit to the next, state-backed finance at low rates, and a regulator familiar with the design. Korea's KEPCO built all four units at Barakah in the United Arab Emirates, about 5.6 GW, and had them operating by September 2024. France built about 56 standardized reactors in about fifteen years under its 1974 plan.

Small modular reactors promise the same benefits through factory production. The first test was not encouraging: NuScale's first project was cancelled on 8 November 2023 after its target price rose from about $58 to about $89 per MWh. The book's study of nuclear power, fuel and small modular reactors covers the newer designs.

Which factor of the power equation failed?

The Power Equation holds that an energy company creates durable value only when four factors hold at once: supply at a cost the market will pay, a route to market, capital matched to the asset, and a licence to operate. Vogtle turned on the first factor. The reactors work, they had a guaranteed route to market through Georgia Power's rate base, and they eventually had matched capital through state law and federal guarantees. But their cost, more than double the budget and seven years late, was paid by customers and shareholders who would have been better off with almost any alternative at the time it was incurred, and the company that promised to deliver them at a fixed price went bankrupt.

Vogtle also shows the licence factor in a regulated business: finishing the plant did not end the fight over who pays, as the 2025 Georgia election showed. For the parallel story of a project that failed on cost in a different industry, see our case study on what happened to Northvolt. For a wider list, see the biggest energy bankruptcies and what they teach, and for where nuclear sits in the industry, our guide to how the energy industry makes money. Company detail is in the Southern Company and Westinghouse Electric profiles.

The takeaway

Vogtle cost more than twice its budget because it was a first-of-a-kind build on an unfinished design, by a rebuilt industry, under a fixed-price contract the vendor could not honor, with interest compounding through every delay. It was finished, unlike V.C. Summer, because Georgia law and federal guarantees provided patient capital that put most of the risk on customers. The reactors now produce valuable firm power, but their cost was a transfer from Georgia ratepayers and Southern shareholders that a better-sequenced program would have avoided. The test for every new reactor is whether it can close the gap between what existing nuclear is worth and what new nuclear costs, the question at the heart of Chapter 11 of The Power Equation.