In April 2023, the French nuclear industry told its government it would need about 100,000 hires by 2033. In June 2024 it said 100,000, by 2034. In November 2025 it said about 100,000 again, this time by 2035. Three editions, one number, a deadline that moved forward a year each time. The industry's own forecast now describes the need as running "over a rolling ten years".
When an industry prints the same hiring figure three years in a row, what exactly is it promising, and to whom? This piece reads the three editions of MATCH, the forecast produced by GIFEN, the trade body of the French nuclear industry, for what each one counts. It sets them beside the strongest outside reading, from France's public audit court, and asks what a person choosing a career in the sector should take from them.
The 100,000 is not a countdown. It is a rolling ten-year need that GIFEN recalculates every year from its members' workload plans, so the headline stays flat while what lies beneath it changes: the territory counted, the number of occupations, the size of the core workforce. Read edition by edition, the forecast says less about a single target and more about where the work is: mostly technicians, mostly at suppliers, with peaks in 2026 and around 2032. The public auditor's warnings show what the forecast cannot see.
This is the first CareerOn piece on GIFEN. It rests on GIFEN's own documents, each read in full and cited by page, and on one outside document from a different publisher, the Cour des comptes. No figure is taken from press coverage of these reports. Where secondary accounts disagree with the documents, we follow the documents.
Three editions, one number
The first edition was delivered to ministers on 21 April 2023; they had commissioned it on 20 January (Ministry press release, 21 April 2023). Built from work begun in 2020 with more than 100 companies and professional bodies (GIFEN, MATCH summary, April 2023), it covered 20 operational activity segments and 84 key occupations. It counted a little over 125,000 full-time-equivalent jobs in that core perimeter in 2023, about 45% at the nuclear operators and 55% at their suppliers, and projected a need of a little over 155,000 by 2033: growth in the volume of work of about 25% in ten years, before any productivity gain (p13). "Extrapolated to all 220,000 jobs in the nuclear industry," it concluded, "the foreseeable need would be about 100,000 full-time-equivalent recruitments" (p13).
The second edition was presented on 18 June 2024 and is reported in GIFEN's 2024 activity report. Its perimeter was now France and the United Kingdom, across the same 20 segments but 97 "standard occupations". It foresaw activity growth of 15 to 20% by 2030, before productivity gains, then a plateau to 2034. Its key figure: "100,000 FTE: foreseeable need for new work capacity" (p24).
The third edition is dated November 2025. It maps 1,830 companies and estimates that they employed 247,000 full-time equivalents in 2024 (p4). Across the whole industry, it says, the foreseeable recruitment need "remains of the order of 100,000 FTE over the next ten years, that is, of the order of 10,000 FTE a year on average", half to replace departures and half for growth (p5). In its conclusion it names the mechanism itself: a need that continues "at a pace of about 10,000 FTE a year over a rolling ten years" (p9).
Three editions print the same 100,000 while the horizon, perimeter and occupations move
| April 2023 | June 2024 | November 2025 | |
|---|---|---|---|
| Horizon | 2033 | 2034 | 2035 |
| Headline need | about 100,000 recruitments (p13) | 100,000 FTE of new work capacity (p24) | about 100,000 FTE over 10 years (p5) |
| Perimeter | France | France and United Kingdom (p24) | France and United Kingdom, plus accessible activity (p2) |
| Occupations studied | 84 key occupations | 97 standard occupations (p24) | about 100 core occupations |
| Whole-industry jobs | 220,000 (p13) | Not printed | 247,000 FTE in 2024 (p4) |
| Core workforce | a little over 125,000 FTE (p13) | Not printed | 163,000 FTE; 2024 estimate 135,000 (p5) |
| Stated margin | Not printed on the pages cited | Not printed | plus or minus 15% (p6) |
Sources 1 Groupement des industriels français de l'énergie nucléaire · 2 Groupement des industriels français de l'énergie nucléaire · 3 Groupement des industriels français de l'énergie nucléaire
How the number is made
MATCH is not a survey of vacancies. It starts from the workload plans of the seven national project owners, ANDRA, CEA, EDF, Framatome, Naval Group, Orano and TechnicAtome, together with about 200 key supply-chain companies and nine professional organisations (GIFEN, MATCH 2025 key results, p2). Those plans are converted into the work each activity segment will require, segment by segment, in full-time equivalents. The core result is then extrapolated to the whole industry. Each year the exercise is revised, and the ten-year window moves forward with it.
That method explains why the headline barely moves. If the programmes in the plans stay broadly the same and the window advances by a year, the ten-year need stays of the same order: some hires made in the past year drop out of the count, and a new year of need enters at the far end. GIFEN says as much in its 2025 conclusion: the update "does not fundamentally change the input data for training and attractiveness actions" (GIFEN, MATCH 2025 key results, p9). A flat headline is therefore what a stable rolling forecast should produce. It is not, by itself, evidence that hiring is behind.
What does move is underneath. The 2025 edition puts the current core workforce at 163,000 full-time equivalents, against an estimate of 135,000 in its 2024 figures, and attributes the rise "partly to a beginning of growth in headcount and partly to a refinement of the modelling" (p5). The perimeter widened to include the United Kingdom in 2024 (GIFEN, activity report 2024, p24). The number of occupations studied went from 84 to 97 to about a hundred. Because the counting changed each time, we do not compute a growth rate from one edition to the next. GIFEN does not either.
The forecast also states its own margin. The 2025 figures, it says, are "calculated with a margin of uncertainty of the order of plus or minus 15%" (p6). Its total for the industry is printed two ways in the same document: the industry "could employ about 300,000 FTE from 2032" (p4), and should go "to about 300,000 FTE in 2035" (p9). We print both as written.
Where the work is
For a reader choosing a path, the useful signal is not the total but its composition. Technician occupations account for two-thirds of recruitment needs, the 2025 edition says, across trades such as maintenance technicians, formwork carpenters, steel fixers and design draughtsmen. Among engineers, the needs are concentrated in project management, mechanical engineering, nuclear safety, civil engineering, instrumentation and control, and certified non-destructive testing. Works supervisors and site managers alone represent nearly 8%1 of the recruitment need for 2025 to 2035 (p7).
GIFEN's 2025 activity report adds where those hires sit: 40%1 at the project owners and 60%1 at suppliers, with two-thirds in technical occupations, and half of all recruitment for growth and half to replace departures.
Two-thirds of the need is technician work, and most hires sit at suppliers
| Measure | As printed | Document |
|---|---|---|
| Technician occupations | two-thirds of recruitment needs | MATCH 2025, p7 |
| Works supervisors and site managers | nearly 8% of the need, 2025 to 2035 | MATCH 2025, p7 |
| Hires at suppliers | 60% | Activity report 2025 |
| Hires at project owners | 40% | Activity report 2025 |
| Replacement and growth | half each | MATCH 2025, p5 |
Sources 3 Groupement des industriels français de l'énergie nucléaire · 6 Groupement des industriels français de l'énergie nucléaire
Timing matters as much as volume. The project owners' purchasing forecasts across the 20 segments rise by nearly 30% over five years, and the model shows load peaks "in 2026 and then around 2032", which GIFEN attributes to the ramp-up of the EPR2 reactor programme and then the launch of Orano's Aval du Futur fuel-cycle programme (p4). Segments do not move together. Civil-engineering purchasing is expected to more than double between 2025 and 2035, with headcount that could triple at the peak. Electrical work is expected to rise from 2028, by about 20% to 2032. Heating, ventilation and air conditioning is expected to stay stable until 2027 and then rise by 70%1 to 2032 (p5–6).
GIFEN's own account of the 2025 results, dated 15 December 2025, does not present the 100,000 as news. It calls the recruitment estimate "confirmed" at 100,000 full-time equivalents across the industry over the next ten years, and it adds that the load peaks "could be smoothed in the future" (GIFEN, 15 December 2025). In the forecaster's own words, then, the headline is a standing figure and the calendar is still moving.
The first edition made the same point from the supply side: more than 40,000 new full-time-equivalent jobs by 2030 inside supplier firms, mainly mid-sized and small companies, implying an average flow of about 6,000 a year at suppliers across the core segments (p70). That figure is sometimes repeated as the whole industry's yearly pace. It is not.
What the forecast cannot see
The strongest case against reading MATCH as a disciplined rolling plan comes from the Cour des comptes, in its January 2025 report on the EPR programme. The court describes MATCH as a tool for steering the match between needs and resources, and restates its 2023 result as about 60,000 full-time-equivalent recruitments in the core perimeter, half for growth and half for departures (p68). It then reports what a forecast built on workload plans cannot capture.
First, the suppliers often do not know the plan. The court cites a finding that "40% of industrial firms do not know the assumed production rate or the volume to be produced each year, and 33% do not know the maximum rate they can reach by 2030" (p74, note 64). Second, the EPR2 programme's maturity review of November 2023 "expressed reservations about GIFEN's capacity to carry the voice of the industry's suppliers" (p69). Third, the same review observed that the industrial base "has lost a specific, decisive competence: site engineering" (p75), a skill no headcount projection replaces.
The public auditor names three things the forecast has to take as given
| What the forecast assumes | What the court reports | Page |
|---|---|---|
| Suppliers know the plan | 40% of industrial firms do not know the assumed production rate or the annual volume | p74, note 64 |
| GIFEN speaks for suppliers | The EPR2 review committee expressed reservations about GIFEN’s capacity to carry their voice | p69 |
| Headcount is the constraint | The industrial base has lost a specific, decisive competence: site engineering | p75 |
Source 8 Cour des comptes
On that reading, a deadline that moves a year with each edition could also be a schedule slipping into the future, and a forecast built from the project owners' plans would not show it. The documents do not settle which reading is right, and we do not settle it for them. A reader should hold both: the headline is a stable rolling estimate, and its reliability depends on programme decisions the forecast takes as given.
How to read the next edition
If the forecast is revised every year, the useful skill is reading the revision, not memorising the headline. Four checks separate a real change from a change in counting, and each can be run on the published documents alone.
First, check the perimeter. The 2023 edition counted France. The 2024 edition counted France and the United Kingdom. The 2025 edition also counts nuclear activity open to French industry in projects such as ITER, small modular reactors and exports (GIFEN, MATCH 2025 key results, p2). Any comparison of totals across editions that ignores this compares different territories.
Second, check the unit and the base. Every headline figure is in full-time equivalents, not people, and the 100,000 is an extrapolation from a core perimeter to the whole industry. The core perimeter itself changed: 84 occupations in 2023, 97 in 2024, about a hundred in 2025. A rise in the core workforce from one edition to the next can be growth, re-counting, or both. The 2025 edition says that its own rise is both (GIFEN, MATCH 2025 key results, p5).
Third, check the horizon and the peaks. The ten-year window advances with each edition, so the year a target is attached to tells you when the forecast was made, not when the work ends. The peaks are more informative: in 2025 they sit in 2026 and around 2032, each tied to a named programme (p4). If a later edition moves a peak, the programme it is tied to has moved.
Fourth, check the stated margin. The 2025 edition gives plus or minus 15% (p6). A change between editions smaller than that margin is not, on the forecast's own terms, a change.
Run those four checks and the next edition becomes a reading of the industry's schedule, segment by segment, rather than a single number to repeat.
What it means for your career
For candidates, the forecast points to trades before titles. Two-thirds of the need is technician work, from vocational certificates up to three years of higher education, and the jobs named are concrete: maintenance, formwork, steel fixing, drafting, site supervision. The timing points to civil engineering first, then electrical and ventilation work as new-build reaches those stages.
For employers, the weight sits with suppliers: 60% of hires, by GIFEN's account (GIFEN, activity report 2025). Most firms in the industry are diversified: on average, nuclear accounts for 20% of their turnover (GIFEN, MATCH 2025 key results, p4). The court's finding that many do not know the production rate they will be asked to meet is the practical constraint on whether they can recruit ahead of demand.
For decision-makers, the forecast's own margin of plus or minus 15% and its dependence on programme timing are the two numbers to keep beside the 100,000. A rolling ten-year need is a planning instrument, not a promise of jobs on a date.
CareerOn’s view is that the most useful thing to carry from MATCH is not its total but its calendar: which trade is needed, at which stage of a build, tied to which programme. No CareerOn practice exercise yet matches the trades the forecast names, from formwork and steel fixing to site supervision, so no practice link is offered here. We record that as a gap in our catalogue rather than fill it with an exercise from another industry.
The 100,000 has been printed three times, and it will likely be printed again next year with a later date. That is how it was designed. A forecast that rolls forward is doing its job when its headline holds steady; it is telling you something new only when its peaks, its perimeter or its mix of trades move. Those are the lines worth reading first in every edition. The question for anyone planning a career in the sector is not whether the number is right, but which part of it describes their trade, and when.