At the end of April 2026, an average of 1,017 people were at work each day on the cliff at Penly, on the Normandy coast, where EDF plans to build the first two of six EPR2 reactors (p5). The state and EDF expect about 10,000 at the peak of the build (p12). The decision to build has not yet been taken. EDF's board is due to take it before the end of 2026.
When France decides to build its next reactors, what does it decide about people, and what has already been decided before the cheque is signed? This piece reads the decision as its sponsors present it, then reads the site where it will be carried out, through the reports the regional prefecture and EDF publish together every six months. It sets both beside the public audit court, the industry's own forecast and EDF's works council, and asks what a person choosing a career in the sector should take from them.
The investment decision is argued in money: a price, a state loan, a forty-year contract. On the site, the binding variable is a workforce that must grow about tenfold, and its hardest trades are only now being trained. Most hires so far have been managers and engineers; the share is falling, slowly, as the first steel fixers and formwork carpenters arrive. The decision fixes a price. The site has to supply the crews, and that work began before anyone agreed to pay for it.
This is CareerOn's first piece on EDF's new-build programme. It rests on eight dated documents, each read in full. No figure is taken from press coverage. Where a press account and a document differ, we follow the document, and where a document is not public, we say so rather than report what others say it contains.
The decision, as it is argued
On 18 December 2025, EDF presented its board with a forecast cost for six EPR2 reactors at Penly, Gravelines and Bugey: €72.8 billion in 2020 euros (EDF, 18 December 2025). The board approved €2.7 billion for the programme in 2026. The first reactor at Penly is to enter service in 2038, the following ones every 12 to 18 months. The state's support, submitted to the European Commission on 19 November 2025, has three parts: a subsidised loan covering at least half of construction costs, a contract for difference running forty years, and a sharing of risks between the state and EDF.
On 12 March 2026, the President's nuclear policy council "took note of the conclusions" of an audit of that estimate carried out early in 2026 by the state's new-build delegation, the DINN (Élysée, 12 March 2026). EDF, the council said, "committed to hold the construction cost target of €72.8bn2020 and the programme schedule". It asked EDF to carry out the audit's recommendations "by the end of 2026". It also stated the loan at "60% of the total amount of the programme". The two documents describe the loan with different bases and three months apart; we print both as published and do not reconcile them.
The audit itself is not public. A ministerial answer to a senator, published on 4 September 2025, set out its place in the sequence: once EDF had finalised its estimate, "this estimate will be the subject of an audit by the DINN", and "the final investment decision is expected for the end of 2026" (Sénat, 4 September 2025). We have the acknowledgement of the audit, not the audit, and we do not paraphrase its contents.
The price has moved before. The Cour des comptes reports that EDF's estimate for the same six reactors rose from €51.7 billion to €67.4 billion in 2020 euros at the end of 2023, with the Penly pair at €23.6 billion, and that the first Penly unit moved from September 2036 to July 2038 (Cour des comptes, January 2025, p32). We do not compute a rate of increase from those figures to the €72.8 billion: the court itemises what its estimate contains, and the December 2025 release does not, so we cannot show that the perimeters match.
The EDF release, the Élysée statement and page 32 of the Cour report set out money and dates, without a workforce count in those passages. That is not a claim about the full Cour report. Delivering EDF’s schedule, a first reactor in 2038 and one every 12 to 18 months after it, requires the right trades on the right part of the site in the right year. That assumption is tested in a different set of documents, a few kilometres from the reactors.
The cited decision passages set out money and dates, not workforce counts
| EDF, 18 Dec 2025 | Élysée, 12 Mar 2026 | Cour des comptes, Jan 2025 | |
|---|---|---|---|
| Cost | €72.8bn (2020 euros), six reactors | EDF committed to hold €72.8bn (2020 euros) | €51.7bn to €67.4bn (2020 euros), end-2023 estimate (p32) |
| First reactor | Penly, 2038 | first commissioning by 2038 | slipped from September 2036 to July 2038 (p32) |
| State loan | at least half of construction costs | 60% of the total programme | Not printed on the page cited |
| Workforce, hiring or training | Not printed | Not printed | Not printed on the page cited |
Sources 2 Électricité de France · 3 Présidence de la République · 5 Cour des comptes
A thousand before the decision
The regional prefecture and EDF publish a joint barometer of the Penly works, twice a year. The December 2025 edition counted an average of 883 people on site at the start of that month (Grand Chantier barometer, December 2025, p5). By the end of April 2026 the average was 1,017, with peak days near 1,200 (Grand Chantier barometer, June 2026, p5). They were reshaping the cliff into three terraces at 44, 64 and 84 metres, filling about 20 hectares of new platform won from the sea with 1.7 million cubic metres of chalk, and digging the channel where the reactor buildings will stand (p8).
This is preparatory work. It is not the nuclear construction itself. But it is already a workforce: between 52% and 55% of it local, between 12% and 16% of it women, with 454 people recruited for the project by EDF and its contractors by the end of April 2026 (p5, p6). About 200 temporary contracts were active, 92% of them for manual workers (p6). The barometer counts 67 apprenticeship and work-study contracts signed since the start of the site (p6).
The order matters. A reactor programme decided in December would normally be expected to start hiring in January. At Penly the hiring is two years old, the cliff is half reshaped and the housing is being built, all before the decision that is supposed to commit the money. The decision does not open the site; it decides whether the site, already open, becomes a reactor.
Who has been hired
The most useful line in the barometers is a share. Until the end of 2024, managers and engineers, the "cadres" in French statistics, accounted for about 80% of hires on the project. By the end of November 2025 they accounted for 62% of 351 hires (Grand Chantier barometer, December 2025, p6). The report reads the shift as the site moving "into an operational phase", with more manual workers and technicians, and expects the trend to continue in early 2026.
It did, but slowly. By the end of April 2026, the share of managers "is stabilising at around 60%" of 454 hires (Grand Chantier barometer, June 2026, p6). The first hires of steel fixers, the workers who build the steel cages inside reinforced concrete, began in 2026 with the contractor Sendin (p6). The first masons, formwork carpenters and steel fixers are only now appearing on site (p5).
Set that beside the industry's own forecast. GIFEN's 2025 MATCH edition finds that technician occupations account for two-thirds of the nuclear industry's recruitment need over the next ten years, naming formwork carpenters and steel fixers among them (GIFEN, MATCH 2025 key results, p7). The forecast covers the whole industry, not one site, so the two shares measure different things and we do not set one against the other as a gap. They point the same way: the trades the forecast says the industry needs most are the trades Penly has only begun to hire.
A manager share of 60% is not a failure. Early on, a site of this kind needs engineers, planners, safety staff and supervisors before it needs crews. The point is the direction of travel and its speed. From 80% to 62% took about a year; from 62% to 60% took five months. The barometer will show whether the share falls faster as civil engineering starts, and that is the line to watch in the next edition.
Managers still make up about 60% of hires as the first trades arrive
| December 2025 edition | June 2026 edition | |
|---|---|---|
| Managers among hires | about 80% until end 2024; 62% of 351 hires by end November 2025 (p6) | stabilising around 60% of 454 hires (p6) |
| Active temporary contracts | 200 at end October 2025, 90% manual workers (p6) | 200 at end April 2026, 92% manual workers (p6) |
| Trades named as starting | Not printed | first steel fixer hires in 2026; masons, formwork carpenters and steel fixers mobilised (p5, p6) |
Sources 6 Préfecture de la région Normandie et EDF · 1 Préfecture de la région Normandie et EDF
The trades that do not yet exist
The barometer's training pages show what the site has to create. By the end of April 2026, 295 people had been trained locally and were working on the project (Grand Chantier barometer, June 2026, p15). Nine formwork carpenters who had taken a first course went on to work-study contracts with site employers. In May a new training centre opened in Dieppe, run for Sendin, to train steel fixers for Penly: "a key trade in the ramp-up of civil engineering". Its first group was ten people, on a course of just over four months, after which they will be employed through a temporary-work agency.
Ten steel fixers is a first cohort, not a workforce. The same document describes the scale it has to reach. To house part of the 10,000 people expected at the peak, the territory needs about 6,000 new beds; at the start of June 2026 about 5,300 had been programmed, 110 fewer than in December 2025, as some projects proved not viable (p12). The December edition adds that scenarios under study "could reach 12,000" people, and that about 60% of the peak workforce would come from outside the area (Grand Chantier barometer, December 2025, p12).
Read together, the pages show training and lodging being organised before the investment decision. The first steel-fixer course lasts just over four months; the housing programme can lose planned places when projects prove unviable. The €72.8 billion headline does not describe either pipeline. That does not establish that their costs are excluded from the estimate: its perimeter is not itemised in the release.
The site holds about a tenth of its peak workforce, and housing is still being found
| December 2025 edition | June 2026 edition | |
|---|---|---|
| Average workers on site | 883, early December 2025 (p5) | 1,017, end April 2026; days near 1,200 (p5) |
| Expected at peak | 10,000; scenarios under study up to 12,000 (p12) | 10,000 (p12) |
| Share expected from outside the area | about 60% of the workforce (p12) | Not printed |
| Lodging places | about 6,000 needed; programme up 81 places since summer (p12) | about 5,300 programmed of about 6,000; 110 fewer than December 2025 (p12) |
| Trained locally and working on site | Not printed | 295 (p15) |
| Steel fixer course | Not printed | first group of 10, a little over 4 months, from end May 2026 (p15) |
Sources 6 Préfecture de la région Normandie et EDF · 1 Préfecture de la région Normandie et EDF
What the decision cannot see
The strongest case against putting people at the centre comes from the Cour des comptes. Its January 2025 report traces the delays of earlier EPR projects mainly to design and to requirements that kept changing. It reports the conclusion of the November 2023 maturity review of the programme: that it is "crucial" for the programme's success that the requirements applying to EPR2 be frozen, at least for the design and construction of the first six units (Cour des comptes, January 2025, p55). On that reading, a site can be fully staffed and still be late, if the drawings it builds from keep moving.
EDF's own central works council raises the workforce question from the inside. Consulted on the investment decision on 25 September 2026, its elected members questioned "the industrial and human capacity to carry it out without losing control to subcontracting", as well as the realism of costs, schedules and expected returns, and voted unanimously to commission an independent expert review (CSEC EDF SA, 1 October 2026). Its conclusions are not yet public.
The two counter-readings do not cancel the thesis; they sharpen it. Design maturity decides whether the crews have something stable to build. Subcontracting decides who employs them and who keeps the skill when the site moves on to Gravelines. Both are questions about people and organisation that a cost estimate expresses only as a provision for risk.
How to read the next barometer
The next edition of the Penly barometer is due around December 2026, close to the investment decision. Four lines will say more than the headline count. First, the manager share of hires: read any change beside the trades actually recruited, rather than treat 60% as a start threshold. Second, the steel-fixer and formwork training cohorts: their number and size, not only their existence. Third, the bed count against the 6,000 target, and whether it keeps slipping. Fourth, the number of active temporary contracts and the share of those contracts concerning manual workers. The 92% measures that composition, not the share of all manual workers employed temporarily; neither figure identifies who will retain the skill after Penly.
What it means for your career
For candidates, the opening is in trades, and it is local first. Steel fixing, formwork, site safety and supervision are named in the documents, training is funded near the site, and the barometers report that more than half of the people working there already come from the area. A course of four months leads to a contract on a programme that, on the schedule the Cour des comptes reports, runs to 2044 across three sites (Cour des comptes, January 2025, p32).
For employers, the documents point to a contracting model that leans on temporary work and subcontractors for manual trades. That is efficient while the site is small. It is the model EDF's works council questions, because the skill a firm builds at Penly is the skill it needs at Gravelines.
For decision-makers, the numbers to keep beside the €72.8 billion are the manager share of hires, the size of each training cohort and the bed count. They are published every six months, by the state and EDF together, and they measure the part of the schedule that money alone cannot buy.
CareerOn's view is that the investment decision is the wrong date to wait for. The site already hires, already trains and already houses, and the documented steel-fixer course offers one concrete entry route. The documents do not rank shortages or compare the time needed to enter every trade. No CareerOn practice exercise yet matches site trades such as steel fixing or formwork, so no practice link is offered here. We record that as a gap in our catalogue.
EDF's board will sign the decision in money. The barometer that follows it will show whether the people arrived on time. Of the two documents, the second will tell you more about the reactor that enters service in 2038.