"Anatomy of the Loss" documented how cloud certifications replaced infrastructure training in skills frameworks. A certified AWS engineer who has never seen a server room is hired instead of a technician who has maintained systems for ten years. The market follows certifications. Certifications follow hyperscalers. Hyperscalers have every interest in the market depending on them.

"The Infrational Crisis" documented how "Infrastructure as Code" and "DevOps" emptied words of their substance. The bare-metal SRE becomes "DevOps engineer". The systems administrator becomes "cloud engineer". Titles change, physical layer competences disappear from job descriptions, and HR recruits on the new titles without understanding what was lost in conversion.

There exist in France and Europe dozens of associations, think tanks, conferences and working groups on "digital sovereignty". They publish reports. They organise roundtables. They produce recommendations. Structurally, they change nothing. Digital sovereignty is not decreed in a conference room — it is built in server rooms, by trained engineers kept in activity. Without this body, all the reports in the world will not produce a functional sovereign datacentre.

If the current trend continues without state intervention, OVHcloud, Ecritel, Scaleway and their equivalents will no longer find the competences needed to operate their infrastructures. They will be forced to either massively import skills from third countries, or sub-contract their operation to providers, some of whom will be foreign. Digital sovereignty will have been lost not by political decision but by the silent exhaustion of a professional body that nobody thought to protect.

Doctor, pilot, architect, notary — these titles are protected by law. Practising them without holding them is a criminal offence. This protection creates market pressure toward quality: clients wanting these competences must use qualified professionals. The same mechanism applied to "critical infrastructure engineer" or "certified systems and networks technician" would produce an immediate revaluation of the body.

To build a public building, a contractor must present certifications (Qualibat, Qualifelec). To participate in a public aeronautical maintenance contract, DGAC approvals are required. Why is a provider operating the infrastructure of a hospital, a town hall, or a ministry not required to present equivalent infrastructure competence certifications? The absence of this obligation in digital public procurement is a gaping hole in sovereignty policy.

BTS Digital Systems, BTS SIO, DUT Networks & Telecoms — these training programmes exist. They are underfunded, undervalued in school guidance, and progressively cloud-oriented at the expense of the physical layer. The state that decides infrastructure training is strategic injects resources, revalues the pathways, and explicitly integrates bare-metal into the frameworks — as it integrated aeronautical mechanics into aeronautical technical training.

Every infrastructure engineer who leaves the market to retrain as a "cloud engineer" is an engineer who will not return to bare-metal. France and Europe's dependence on American hyperscalers deepens in direct proportion to the shortage of sovereign infrastructure competences. This dependence has an economic cost (egress fees, contractual lock-in, hyperscaler pricing power) and a geopolitical cost (CLOUD Act, legal extraterritoriality, vulnerability to unilateral American decisions) that the state bears without ever calculating it explicitly.

OVHcloud, Ecritel, Scaleway, Infomaniak, Clever Cloud — these actors exist. They have datacentres. They have offerings. They do not have enough qualified staff to operate them at the necessary scale. This is not a problem of commercial will — it is a problem of competence pool. If the state decided tomorrow that all public administrations must host their data with a European sovereign actor, these actors could not absorb the demand for want of trained infrastructure technicians and engineers. A preference policy only makes sense if the body to operationalise it exists.

National critical infrastructures — energy networks, telecommunications networks, hospital systems, transport networks — run on computing systems that infrastructure engineers maintain. The shortage of these engineers is a national security vulnerability that neither CISOs nor cybersecurity consultants can compensate. A CISO without an infrastructure engineer to implement their recommendations is a doctor without a nurse to administer the care.

A market rich in qualified infrastructure engineers is a market where clients can choose not to use hyperscalers. The shortage of bare-metal competences is structurally favourable to hyperscalers — it makes cloud even more unavoidable because there are no longer the competences to do otherwise. Hyperscalers do not actively train bare-metal engineers. They train "cloud architects" certified on their own platforms. This is not malevolence — it is economic rationality.

Training an infrastructure engineer takes 2 to 5 years. The employer who invests in this training runs the risk that the engineer leaves for a competitor or a hyperscaler once trained. This free-riding risk reduces private investment in infrastructure training below the socially optimal level — exactly the same mechanism that justifies state intervention in the training of doctors, pilots and nuclear engineers.

Digital sovereignty professional associations and conferences produce visibility and networking. They do not produce technicians. Visibility without a trained professional body is a conference on medical desertification without training doctors. Speech is necessary. It is not sufficient. And when it becomes the substitute for structural action, it is counterproductive — it gives the impression that something is being done when nothing is.

A title registered in the RNCP, accessible by curriculum and by VAE, recognised in collective agreements with a protected salary scale. This title protects competence — not the pathway. It is accessible to the self-taught Blue through VAE as much as to the engineering school graduate. It creates market pressure toward recognition of bare-metal competences.

Any provider operating the infrastructure of a public body — administration, hospital, local authority — must demonstrate engineers holding a recognised infrastructure competence title. Like Qualibat for construction. Like DGAC approvals for aeronautical maintenance. This obligation creates a market for competence and a competitive disadvantage for providers who have substituted cloud generalists for infrastructure engineers.

BTS SIO, BTS Digital Systems, DUT Networks & Telecoms must explicitly integrate the physical layer into their frameworks: cabling, hardware, bare-metal systems, physical network diagnosis. The Foundation of Iron — 26 weeks from electricity to networking — is a turnkey programme available under an open licence. It can be adapted in 12 months into a 2-year BTS curriculum.

The historical pathway exists and worked for fifteen years. France Travail directly funds accredited training organisations on shortage-area training programmes. Bare-metal infrastructure is a documented structural shortage area. The AFC convention is sufficient — no legislation needed, no exceptional budget. An administrative decision.

Integrate an explicit preference for sovereign European actors (OVHcloud, Ecritel, Scaleway, Hetzner, Infomaniak) into the award criteria of digital public procurement. This preference creates infrastructure employment in France, funds sovereign actors, and reduces dependence on the trio. A state that preaches digital sovereignty while hosting its data on AWS is not practising sovereignty — it is talking about it.

The state publishes each year the medical demographics, the state of medical deserts, training needs. It publishes nothing equivalent on digital infrastructure competences. Creating this reporting obligation — number of active SRE engineers and systems administrators, evolution over 5 years, cloud reconversion rate, shortages by region — makes visible what is invisible. And what is not measured is not protected.