Flamanville 3 fed its first electrons into the grid on December 21, 2024. Nuclear operation began following the operating license authorization of May 7, 2024. Fuel loading started on May 8, 2024, and startup tests were completed on April 27, 2026. The delay stems from insufficient project preparation, governance and monitoring deficiencies, loss of expertise and quality culture in the sector, as well as technical, contractual, and subcontracting problems.
The Essential Points
- The Flamanville EPR entered commercial service in 2026 with a nine-year delay from the initial 2017 date, not for technological reasons, but due to lack of continuity in expertise, a maintained subcontracting chain, and long-term program management.
- The PPE3 (February 2026) sets a target of 60% decarbonized energy by 2030 and 70% by 2035; the first EPR2s will not be operational until after 2038, according to World Nuclear News.
- Vogtle in the United States and Japanese revivals show that different management architectures can reduce, without eliminating, these institutional frictions.
- The central constraint of nuclear power in a democracy is the mismatch between investment cycles of twenty to thirty years and political mandates of four to five years.
- France has the technical resources to succeed in its EPR2 program; the question is whether it will equip itself with governance structures capable of sustaining it over time.
Nine Years of Delay: Lessons from a Construction Site
France has adopted an ambitious PPE3 combining a nuclear revival, electrification, and renewable development: 60% decarbonized energy by 2030, 70% by 2035, and a series of new EPR2s whose first startup, at Penly, is planned for 2038 according to EDF. The technology is available. The capacity of institutions to sustain a program of this duration remains to be demonstrated.
Flamanville began construction in 2007. Industrial startup was initially planned for June 2012, then startup was rescheduled for 2017. The reactor was connected to the grid on December 21, 2024, and its startup tests ended on April 27, 2026. Each delay has its own technical explanation: non-compliant welds, cracked concrete, defective components. The delays resulted from a set of interdependent systemic causes, including initial lack of preparation, management failures, and problems with expertise, quality, contracts, and subcontracting.
Between 1997 and 2007, France built no reactors. France experienced a long interruption in launching new reactors before the Flamanville 3 construction site in 2007, without the entire civil nuclear program coming to a halt. The sector experienced difficulties in the availability of welders qualified for nuclear work. Engineering firms reduced their workforce. Specialized subcontractors diversified their activities into other sectors.
When the Flamanville construction site started, the French nuclear industry had to simultaneously build and rebuild itself. The bill was heavy: the final cost of the reactor is estimated at more than 13 billion euros, compared to 3.3 billion euros initially planned, according to EDF.
This diagnosis is shared by EDF engineers themselves. Nuclear expertise is not a stock that one preserves indefinitely: it is a flow that is maintained through practice. Stopping construction is the beginning of forgetting. Flamanville’s return was thus that of an industry relearning its trade on a real construction site, with all the errors that entails.
The lesson is not new. It was already visible in Britain with Hinkley Point C, at double its initial budget. It is observed in Finland with Olkiluoto 3, put into service in 2023 with thirteen years of delay. Every country that has interrupted its nuclear program pays the same price upon resumption. Industrial discontinuity costs more than any design defect.
The PPE3 Traces a Trajectory, Not a Guarantee
The pluriannual energy programming adopted in February 2026 is an ambitious document. It commits France to an accelerated decarbonization trajectory and makes nuclear power one of its cornerstones. The official target is 60% decarbonized energy by 2030; for 2035, the PPE announces a higher share, without setting a specific numerical target for this source at 70%. The six EPR2s constitute a component of long-term renewal and strengthening of the nuclear fleet, within an electricity mix combining existing nuclear, new nuclear, and renewables.
But programming is not a funding program. And a funding program is not an industrial organization. The PPE3 says where France wants to go. It does not guarantee that the structures to get there already exist.
The EPR2s benefit from the lessons learned from Flamanville. The design has been simplified, standardization increased, and interfaces between trades clarified. These improvements are real. They should make it possible to reduce delays and costs compared to the first-generation EPR. But the first EPR2 will also, to a certain extent, be a training reactor for the sector.
The gains from standardization are fully realized only with the third or fourth unit in a series.
EDF’s current objective is startup of the first EPR2 at Penly in 2038, the following ones planned at intervals of 12 to 18 months. This fleet ages well, thanks to EDF’s major renovation program, but it is not indefinitely extensible. The schedule leaves little room for unforeseen events.
The Institutional Choices of Vogtle and Japan
The argument that nuclear power would be structurally unmanageable in a market economy or liberal democracy stumbles on two contemporary counterexamples.
In the United States, the Vogtle plant in Georgia put its two AP1000 reactors into service in 2023 and 2024. The construction site also suffered significant cost overruns: final cost around 35 billion dollars for the two units, roughly double the initial budget, and several years of delay. The experience is hardly an unqualified triumph. But the plant is in service. For Vogtle, the U.S. federal government guaranteed loans to Georgia Power, Oglethorpe Power, and MEAG Power; recovery of certain costs was governed by the Georgia Public Service Commission.
Continuity was not ensured by the virtue of the actors, but by the architecture of incentives.
Japan offers a different lesson. Following the near-total shutdown after Fukushima, Japan initiated a gradual revival starting in 2015; ten reactors had already restarted by July 2022, and the country announced the construction of new advanced models. The Japanese revival relies on operators, METI, and several coordination and training structures; it does not appear to be driven by a single agency dedicated to rebuilding expertise and subcontracting. Expertise is treated there as a prerequisite for revival, not as an automatic consequence.
These two cases have limitations. Vogtle remains a costly construction site in a country where few other states have followed suit. Japan has not yet put a new reactor into service since Fukushima. However, they show that the institutional frictions of nuclear power stem from organizational design problems and can be resolved.
Democracy Facing Thirty-Year Investments
The real challenge of the EPR2 program is political in the most fundamental sense. A nuclear reactor decided on today will be put into service under several successive governments, will traverse political alternations, economic crises, reversals of opinion. The implicit contract that society makes with itself when it launches such a program lasts longer than any electoral mandate.
This mismatch between political time and infrastructure time is not unique to nuclear power. Large industrial infrastructure poses the same problem everywhere: decisions about economic sovereignty that truly matter unfold over decades, not over legislative terms. But nuclear power embodies it with particular acuity, because the consequences of interruption midway are particularly costly.
France has a tradition of continuity in energy policy that allowed it to build the bulk of its fleet in the 1970s-1990s. This continuity rested on a specific political configuration: a cross-party consensus on energy independence, a publicly-owned EDF with considerable resources, and a public little mobilized on the issue. These conditions no longer obtain in the same way. The debate around nuclear power is more lively, funding more contested, EDF restructured after years of tensions.
France can equip itself with an institutional architecture capable of sustaining a thirty-year program: others have done so in varied political contexts. South Korea has maintained a continuity of nuclear industry that today allows it to export its APR1400 reactors to Eastern Europe and the Emirates. Finland created, with Posiva, a model of nuclear waste governance managing a horizon of one hundred thousand years with an institutional continuity that few states can claim over ten years.
These are architectural choices.
What Institution for a Program That Outlasts a Government
For the EPR2s, several governance scenarios are possible, and their difference is not merely cosmetic.
The first, close to the current configuration, entrusts management to EDF within the framework of a subsidized public loan, a contract for difference, and a sharing of risks between the State and EDF. This model has the advantage of simplicity and continuity with what exists. The continuity of the program remains exposed to decisions by EDF, the State, and the European Commission, but it is governed by a scheme of financing and risk-sharing between EDF and the State.
The second scenario, discussed in specialized circles but not yet formalized, creates a dedicated agency for the EPR2 program, endowed with pluriannual funding voted by Parliament over a long duration, outside annual budget cycles. This type of agency, inspired by governance models of large aerospace programs, separates operational management from short-term political vagaries. It demands a greater initial political investment, but it offers better resilience to political alternations.
A third scenario, more ambitious, inscribes the continuity of the program in a treaty or multilateral European agreement, mutualizing part of the risk and governance with partners engaged in similar programs, Poland, Czechia, Netherlands. This scenario reduces dependence on national political continuity but introduces additional coordination complexity.
None of these scenarios is risk-free. But the first question to resolve, before even the technological design of the reactors, is that of the institutional design of the program. The deployment of decarbonized energy at large scale shows that the supply chain and organizational continuity are as determinative as the technology itself, in solar as in nuclear.
The financing framework under consideration is already defined in the PPE3: subsidized state loans covering at least half the costs and a contract for difference. The updated provisional estimate of 72.8 billion euros in 2020 has been presented and audited; what remained included notably the final investment decision, the operational finalization of the contractual framework, and the outcome of the European state aid examination. These choices, far more than the technical characteristics of the reactors, will tell whether France has learned the lessons of Flamanville.
Industrial Expertise Cannot Be Improvised to Order
Behind the institutional question lurks an even more concrete one: will the hands, the brains, and the companies capable of building these reactors be available when needed?
The French nuclear sector has begun to rebuild itself. EDF’s plan to revive expertise, undertaken even before Flamanville’s startup, aims to train an increasing number of qualified technicians and engineers. Schools and specialized training centers have been reactivated. The subcontracting chain, nuclear boiler-making, piping, specialized civil engineering, is subject to closer monitoring than during the Flamanville construction site.
But training a welder qualified for nuclear work takes years. That of an experienced site engineer takes even longer. The needs for retraining in transforming industrial sectors show that labor markets do not spontaneously adapt to the demands of major infrastructure programs: they need anticipation and public investment in training.
The question of expertise is the most reliable signal of the seriousness of a nuclear program. A government that announces six EPR2s without a detailed plan for building up sector expertise announces an ambition, not a program. Conversely, a credible and funded training plan, backed by pluriannual ordering commitments to subcontractors, is the first tangible indicator that continuity will be maintained.
France possesses the technology, potential human resources, and a nuclear industrial tradition without equivalent in Western Europe. The next institutional stage concerns primarily a political and industrial program of six EPR2s in preparation for Penly, Gravelines, and Bugey; EDF’s final investment decision was still expected before the end of 2026.
Sources
- World Nuclear News, “Nuclear central in France’s latest energy strategy”, 2026: https://www.world-nuclear-news.org/articles/nuclear-central-in-frances-latest-energy-strategy
- EDF, press releases on Flamanville 3 schedule and costs, 2024-2026 (available on edf.fr)
- Pluriannual Energy Programming (PPE3), official text, February 2026 (available on the website of the Ministry for Energy Transition)
- Georgia Power / Southern Company, reports on Vogtle Units 3 & 4, 2023-2024 (available on southerncompany.com)
- International Atomic Energy Agency (IAEA), reports on the Japanese nuclear sector post-Fukushima (available on iaea.org)
