Twenty years of career — teacher, technician, system administrator, Head of SRE — converge on a single empirical finding: the structural shortage in Iron professions is not a demographic accident. It is the mechanical consequence of amputating half the available human capital.

When a technical field durably excludes 50% of its potential talent pool, it does not merely reduce its number of candidates. It cuts itself off from a diversity of approaches, profiles, retraining paths — and it locks itself into a spiral where the shortage itself becomes the argument justifying outsourcing to managed solutions, proprietary consoles, hyperscalers.

This study documents three things: the real scale of the deficit, quantified and sourced; the structural parallel with other chronically tight sectors (metallurgy, construction); and the concrete protocol — regulatory, financial, pedagogical — to reverse the trajectory before the breaking point is reached.

These figures are not an activist estimate. They are INSEE and France Travail data, verified and sourced.

In the training programmes that feed these fields, the gap widens before entry into the labour market: 17% of girls in computer engineering programmes, 6% in IT vocational tracks. The field does not lose its women along the way — it never recruits them.

The parallel between digital infrastructure and physical transformation industries is not a circumstantial analogy. It is a structural homology, documented by the figures of the country's most exposed professional branch: metallurgy.

This tension forced the French Union of Metallurgy Industries to launch, as early as 2023, the "Tu as ta place" ("You have your place") plan — six lines of action, an explicit and public objective: moving from 23% to 33% women in the industry by 2033. This is not a CSR posture. It is an industrial war plan triggered by a shortage that threatens production.

This parallel establishes one precise and verifiable thing: the durable exclusion of half the talent pool mechanically produces the same effect — a shortage that self-aggravates until a massive mixity plan corrects it.

Three elements, and three only, structure the mechanism. None is sufficient alone.

These three elements only work together. Training without mentorship reproduces the isolation documented in field dropouts. Mentorship without a contractual framework remains informal and unfundable. Apprenticeship without neutral fundamentals reproduces dependency on the tools of the vendor hosting the placement. The following sections detail the exact regulatory vehicle that makes this triptych fundable and enforceable.

Two contractual vehicles coexist, with a determining age distinction:

The recommended 12-to-24-month duration fits strictly within the legal framework of both mechanisms. What remains to be established is the funding — the subject of the next section.

The lock most often invoked against long-training mechanisms — the absence of an identified funding vehicle — does not exist here.

The OPCO Atlas fund publicly confirms funding for the "Systems and networks: Cloud systems and network administrator, Helpdesk technician, DevOps" track, with a public search engine by RNCP code giving the exact level of coverage.

This reframing is not cosmetic. It changes the identity of the role: the senior becomes the last custodian of a rare skill, whose transmission is an act of professional preservation — not an additional burden imposed by a top-down diversity policy.

The transmission protocol structures this relationship: a formal pairing for the duration of the apprenticeship, precise and verifiable competence objectives, explicit recognition of the mentor role in the senior's own professional evaluation.

Integrating an apprentice into an SRE team under permanent strain carries a real risk: the overload that turns mentorship into a burden, and failure into premature abandonment.

This structuring protects both the apprentice — never exposed alone to a critical outage without a safety net — and the existing team, whose mentoring load is calibrated and decreasing over time rather than constant.

A perfectly designed training and mentorship mechanism fails if the downstream labour market continues to algorithmically filter the profiles it produces. Automated CV screening systems (ATS) are calibrated on proprietary terminology — "AWS Certified", "Azure Administrator" — which mechanically eliminates profiles trained on neutral fundamentals, regardless of their actual competence.

The operational answer: systematically replace proprietary terminology with agnostic nomenclature in job postings — "Systems and Network Administrator" rather than "AWS Cloud Engineer", "Infrastructure Engineer" rather than "Azure DevOps Engineer". This reframing relies directly on the RNCP codes themselves, which are, by design, vendor-agnostic.

This nudge requires no negotiation with ATS vendors. It acts upstream, on the very wording of job offers — a lever entirely controlled by the hiring company.

The HR rewriting nudge, deployed at the scale of a single company, remains an isolated gesture. Deployed at the scale of a professional branch, it becomes an enforceable standard. Collective bargaining agreements and job classifications defined by professional branches carry a normative force that the internal practices of a single isolated company lack.

This lever goes beyond the scope of an individual study. It calls for social dialogue work — with digital employer organisations (Numeum) and representative trade unions — to have agnostic nomenclature recognised as a branch standard, not as an isolated and fragile initiative.

IT consultancies (ESN) — formerly known as SSII — employ a significant share of France's system and network engineers. Their commercial apparatus, today oriented toward reselling hyperscaler-certified skills, can be turned around. An ESN that builds a pool of profiles certified in neutral fundamentals — LFCS (Linux Foundation Certified Sysadmin), RHCSA (Red Hat Certified System Administrator) — holds a differentiating commercial argument against a public procurement market increasingly sensitive to digital sovereignty issues.

This mechanism connects directly with the training scheme described in the previous sections: apprentices trained on RNCP36370 or TP AIS are precisely the pool that ESNs, under public procurement pressure, will have an interest in recruiting and showcasing.

Three factors converge to create a tightened time window:

The INSEE and France Travail figures cited in this study are verified and sourced. One additional data point deserves to be named for what it is: an empirical estimate, not an official statistic.

This distinction between verified data and field observation does not weaken this study's thesis. It strengthens it: official figures already suffice to demonstrate a major structural deficit. The empirical estimate adds a dimension that aggregated statistics do not always capture — the conviction of a practitioner who has watched, over two decades, the talent pool empty out without ever refilling.