Executive brief
The French nuclear sector says it must recruit 100,000 full-time equivalents over ten years to reach 300,000 FTE by 2035, from 247,000 jobs counted at end-2024, and that two thirds of those hires sit between CAP and Bac+3, while 48% of today's employees are engineers or managers (Revue Générale Nucléaire, reporting the third edition of GIFEN's Match programme, 28 October 2025). That single mismatch is the story. A sector whose current workforce is nearly half graduate-level is about to hire mostly technicians, and half of the intake is replacement, not growth. This dossier reads the sector's own workforce study against what one employer (Framatome, whose 27,500 m² Lyon campus holds close to 3,000 engineering, project and supply-chain staff) is actually advertising, and states which career conclusions the evidence supports and which it does not.
I. The measured demand, with its perimeter attached
The number that circulates as "100,000 nuclear jobs" is a recruitment need, not a net creation. GIFEN's delegate general stated on 28 October 2025 that half of those future recruitments serve the increase in activity and the remaining 50% replace retirements and departures to other sectors. A reader planning a career should treat the replacement half as the more reliable half: it is driven by demography already recorded, not by programme decisions that can slip.
The 2025 edition also restates the base. GIFEN's president put the ecosystem at 247,000 jobs at end-2024, with a visible increase of around ten percent in full-time equivalents since the start of the decade, and cautioned that this figure is difficult to compare precisely with earlier estimates. The first Match report, presented in early 2023, estimated 220,000 FTE, of which 125,000 were highly technical core-target posts; at end-2024 those core jobs represent 163,000. Part of the rise is genuine dynamism and part is a consolidation of data that now includes the United Kingdom perimeter accessible to the French sector. We print both readings and refuse to derive a growth rate across the two editions, because the perimeter changed between them.
One exclusion matters for anyone reading start-up headlines: GIFEN states that the needs of SMR and AMR developers are not taken into account in this work. Small modular reactor demand is therefore outside every number in this dossier.
II. The qualification inversion
Two published figures, placed side by side, define the decade's opportunity:
The sector’s own workforce arithmetic, with its vintage attached
| Measure | Published value | Vintage |
|---|---|---|
| Jobs counted in the ecosystem | 247,000 | End-2024, stated 28 October 2025 |
| Core highly technical posts | 163,000, against 125,000 estimated in the first edition | End-2024, against early 2023 |
| Total recruitment need to 2035 | 100,000 FTE, half of it replacement | Stated 28 October 2025 |
| Share of hires at technician level, CAP to Bac+3 | Two thirds | Stated 28 October 2025 |
| Share of current employees who are cadres or engineers | 48% | Stated 28 October 2025 |
The practical reading: a sector that has been hiring graduates for decades now needs hands. If two hires in three are below Bac+3 while nearly half the standing workforce is graduate-level, then the bottleneck is not the engineering pipeline: it is welding, piping, assembly, inspection, planning and site coordination. For a student choosing between a five-year engineering track and a two-year technical qualification with a nuclear specialism, the sector's own arithmetic favours the second in volume terms. It does not follow that it favours it in pay or progression; no published series in our sources supports that, and we do not assert it.
100,000 FTE
the recruitment need to 2035 stated by the sector federation, of which half replaces retirements and departures rather than adding jobs
A recruitment need is not net job creation, and the replacement half is the more reliable half because it rests on demography already recorded rather than on programme decisions that can slip. No per-employer share of this figure is published, and none is apportioned here.
Sources 5 Société française d'énergie nucléaire · 9 France Travail
III. Demand is not simultaneous
The same presentation separated immediate from deferred growth, which is the detail most coverage drops. GIFEN's delegate general illustrated an immediate need in civil engineering to accompany the EPR2 programme, saying the number of jobs devoted to nuclear activities in that segment could as much as triple in the coming years. By contrast, electromechanical assembly requires fewer resources in the short term, with headcount needs growing 20% between 2028 and 2032.
That is a timing instruction. Civil engineering (earthworks, concrete, formwork, rebar, site quality) is hiring first because a reactor is a construction project before it is a machine. Electromechanical assembly hires later, when there is a building to install equipment in. A twenty-year-old choosing now is choosing between a market that is tight today and one that will be tight around the end of the decade. Both are real; they are not the same decision.
IV. Where the jobs physically are
Concentration decides search strategy. GIFEN identified 1,830 companies at end-2024, implying roughly 2,000 structures in the sector, and reported that micro-firms and SMEs are 91% of the structures but hold only 11% of the jobs, while about 95% of nuclear workers are in the 600 GIFEN member companies. Nuclear now sits among the three largest French industrial sectors, alongside automotive and aerospace.
For a candidate, that means the efficient search is not a wide sweep of small suppliers. It is a short list of large employers and their named plants. Framatome publishes 7,500 employees in France, a group of more than 22,000 worldwide, a Fuel business unit of about 5,000 and an Installed Base business unit of more than 6,000 with experience across more than 380 reactors (company pages, undated; consulted 4 September 2026). Its Lyon campus at Gerland (27,500 m², close to 3,000 employees) holds engineering, project management, sales, supply chain, R&D and support functions, and the company states its proximity to the Bugey, Saint-Alban, Cruas and Tricastin stations and to the Le Creusot and Saint-Marcel component plants.
The regional shape follows from that: Lyon is where you are hired to specify and plan; Burgundy and Normandy are where you are hired to make and install. Both belong to the same programme and neither substitutes for the other.
Where the jobs physically sit, and why a wide search wastes time
| What is counted | Published value |
|---|---|
| Companies identified in the sector | 1,830 at end-2024, implying roughly 2,000 structures |
| Share of structures that are micro-firms and SMEs | 91% of structures, holding 11% of the jobs |
| Share of nuclear workers inside the federation’s member companies | About 95%, in 600 companies |
V. What the vacancies themselves disclose
Rather than infer pay from surveys, we read what the employer prints. Framatome's own posted vacancies in 2026 carry references, locations, experience bands and salary bands. Three examples, quoted as published:
- Experienced design engineer: fuel rod thermomechanics (ref. 2026-25569), permanent contract, Lyon (69), studies/design and engineering, 5–10 years of experience, €40,000–50,000.
- EPR2 Edvance project planner (ref. 2026-27170), permanent contract, Lyon (69), project, contract and business management, 2–5 years, €45,000–55,000.
- Special machines market manager (ref. 2026-26505), permanent contract, Lyon (69), studies/design and engineering, 2–5 years, €40,000–55,000.
Two observations survive scrutiny. First, the planning role is paid in the same band as the design roles at lower seniority: in a programme measured by contractual milestones, scheduling competence is priced like technical competence. Second, the fuel and installed-base descriptions attached to these adverts name their own scale (a Fuel unit of about 5,000; an Installed Base unit of more than 6,000 across 380+ reactors), which tells a candidate which internal population they would join.
These are three published adverts, not a salary survey. We do not average them, we do not project them onto other employers or other regions, and we do not present them as a market rate.
VI. The public route into the sector
The state and the sector run a shared entry channel. France Travail, co-organising with the Université des Métiers du Nucléaire, announced the fourth edition of the Semaine des métiers du nucléaire from 9 to 13 March 2026, with more than 400 events across the territory, under the headline that the sector seeks 100,000 talents by 2035. GIFEN itself, created in 2018, federates more than 400 members including EDF, Framatome, Orano and the CEA (note submitted to government, April 2023).
The practical value of that week is not inspiration: it is access. Four hundred events organised by an employer federation whose members hold 95% of the sector's jobs is the cheapest introduction to hiring managers available to someone without a network. It runs once a year, and the dates are published in advance.
VII. A defensible plan, and its limits
What the evidence supports, in order:
- Target the technician band deliberately. Two hires in three are CAP to Bac+3. That is the sector's own statement about the next decade, not a guess.
- Choose civil engineering for the near term and electromechanical assembly for 2028–2032. The sector published both timings separately; treat them as separate markets.
- Apply to the large employers first. 95% of workers are in 600 firms; 91% of firms hold 11% of jobs.
- For engineering candidates, take scheduling and configuration control seriously. A programme judged on contractual milestones prices those skills alongside design, as the posted Lyon bands show.
- Consider instrumentation with a cyber discipline. Framatome's Instrumentation and Control unit now carries a cybersecurity business line with three named subsidiaries: a stated, dated change in what that unit does.
What we refuse to claim: any Framatome hiring target for 2026 or later; any per-employer share of the 100,000; any salary trend, median or progression curve; any wage comparison between nuclear and other sectors; any regional employment figure for Auvergne-Rhône-Alpes nuclear activity; any job number for the Bugey EPR2 project; any SMR workforce figure, which Match excludes by construction; any claim that the 2023 and 2025 Match editions can be compared as a growth rate, which GIFEN's own president cautioned against; and any statement that technician demand implies technician pay premiums, for which we found no dated source.
Every figure above is attributed to a named publisher with its date. Where two vintages disagree, both are printed.