The container, not the cell: which hydrogen skills actually get hired near Grenoble

I. Why the container, not the cell, sets the skill demand

Start from the published division of labour, because it is unusually explicit. In the partnership announced on 13 January 2025, Inocel provides the fuel cell and Equans provides the containerisation; where a customer wants the whole hydrogen chain — production, storage, generation — the two combine their expertise and lean on partner networks. The stated envelope is 300 kW to several megawatts, with named applications: quayside power for ships, marine propulsion and onboard generation, storing intermittent renewable output as hydrogen, grid balancing and load-shedding, and backup for critical industrial and tertiary sites including data centres, hospitals and banks, plus mobile generators replacing diesel units on off-grid sites.

Every one of those applications is an installation, not a component sale. That is the mechanism behind the skill demand. Between a 300 kW-class stack and a hospital that stays lit sit: DC-DC and DC-AC conversion, a coolant loop sized for continuous full load, a control system that arbitrates between cell, battery and load, a hydrogen supply and its safety case, switchgear and protection coordination, civil works, commissioning and a maintenance contract. The stack is one line item in a bill of materials whose delivery risk lives in the other lines.

So the honest answer to "what skills does high-power hydrogen need in this region?" is not "electrochemistry". It is: the skills of electrical and thermal integration, plus the regulated-safety literacy that lets an integrated object be accepted on a customer's site. The companion industry dossier establishes the corporate geography; this one establishes what a person actually has to be able to do.

II. Five capability layers, read off the published product

We derive the layers from published specifications rather than from a generic competency framework. Each row names the evidence it rests on.

Layers 2 to 5 are the majority of the delivered value and the majority of the hiring surface. They are also, critically, transferable: a power-electronics engineer who has converted 300 kW for a traction inverter, a thermal engineer from a data-centre cooling background, a controls engineer from process industry, an ATEX safety engineer from chemicals — each is closer to employable in high-power hydrogen than a generalist who has read about fuel cells. In a region whose engineering base is semiconductors, nuclear and batteries, that is good news, and it is the single most useful career conclusion in this dossier.

III. The two-region problem, stated plainly

The company declares Saint-Égrève as headquarters and R&D — stack development, balance-of-plant design, software and electrical — and Belfort as a 15,000 m² factory holding the industrialisation team and customer services, with capacity scaling to 90 MW by 2027 and up to 3 GW long-term by 2032. A third site at Mougins is declared for advanced technology and powertrain prototyping.

Map the five layers onto those addresses and a career reality appears that no job board states:

• Layers 1, 2 and 4 — stack, power electronics, embedded controls — sit predominantly where design sits: Saint-Égrève, in the Grenoble metropole.
• Layer 3 straddles both: packaging is designed with the product and re-learned on the line.
• The manufacturing expression of every layer — process engineering, line ramp-up, production test, quality, supply chain, field service — is declared for Belfort, a different region 400 km away by road as of 2026.

For an Isère-based reader the implication is concrete. If your skill is design, the local address exists. If your skill is production, the local address does not — and applying to "a Grenoble hydrogen company" for a manufacturing role means relocating to Bourgogne-Franche-Comté. This is exactly the kind of fact that a glossy regional narrative blurs and that a careers publication owes its readers.

IV. The overlooked skill: making a specification agree with itself

Here is a finding we did not expect to make in a skills dossier. The company's own public pages publish, for the same product, 15 patents on one page and 25 on another; efficiency "up to 65%" and "up to 60%" in the same page, according to those pages as of September 2026,; and full power in "< 2 s" and "in less than 1.5 seconds" in the same page. The generator page also rates a product family in kVA while the stack is rated in kW — different quantities — and asserts, as of September 2026, "99.99% power availability" with no measurement basis, test population or observation window.

None of that makes the technology weaker. All of it makes a serious buyer slower. In critical-infrastructure sales — hospitals, ports, data centres, banks, exactly the customers named in the partnership — the procurement question is not "is this impressive?" but "which number is contractual?" A specification that contradicts itself transfers work to the customer's engineer, who must then ask which page governs.

So there is a fifth-and-a-half capability, rarely written into a competency framework and decisive in regulated sales: technical documentation discipline — datasheet governance, one rated operating point per quantity, units stated correctly, availability figures defined by their measurement basis, revision dates on public specification pages. In a Grenoble-area labour market full of engineers who can compute a number, the scarce person is the one who can make a company publish the same number twice. Candidates who can evidence that discipline — from nuclear qualification files, from medical-device technical files, from semiconductor process specs — should say so explicitly, because a scale-up moving from prototype to contract is exactly when that skill starts to bind.

V. Where these people come from, and what to build

Three adjacent pools in Auvergne-Rhône-Alpes and its neighbours already hold most of the required capability, and the transfer path is short:

From electrified mobility and traction. High-current power electronics, hybridisation control and thermal packaging for a vehicle powertrain are the same disciplines, at similar power levels, as a 300 kW-class stationary converter. The transfer gap is hydrogen-specific safety and stationary duty cycles, not fundamentals.

From process and nuclear engineering. Gas handling, pressure equipment, zoning, permitting, commissioning and operator qualification are mature disciplines in the region's nuclear and chemical base. The transfer gap is the electrochemical behaviour of the stack, learnable on the job by someone who already knows how to write a safety case.

From data-centre and critical-facility engineering. The named customers include data centres, hospitals and banks; people who already design N+1 backup power, switchgear coordination and cooling for continuous load speak the buyer's language natively. This pool is systematically underused by hydrogen recruiters and is arguably the best-fit source for layer 5.

What the region should build against this is narrower and cheaper than a new degree programme: a short, shared hydrogen-safety and system-integration module that a working power-electronics, thermal or controls engineer can complete without leaving their job, plus a technical-writing and datasheet-governance component that treats specification integrity as an engineering deliverable. That is a modular add-on to existing engineering curricula, not a new institution.

We deliberately do not quantify how many such people are needed. No published figure exists for this company's headcount, vacancies, or hiring plan, and inventing a demand number would be the exact failure this publication exists to avoid.

VI. What a reader should do this quarter

1. Position by layer, not by industry. Do not present yourself as "interested in hydrogen". Present yourself as a 300 kW-class power-electronics, thermal, controls or hydrogen-safety engineer. The layers are what get hired.
2. Choose your region before you choose the employer. Design in Saint-Égrève, manufacturing in Belfort, prototyping in Mougins — the company states this itself. Decide which one your skill maps to and apply accordingly.
3. Prepare the integration answer. Expect to be asked how a stack becomes a site: conversion, cooling, controls, hydrogen supply, protection, commissioning. A candidate who can walk that chain outperforms one who can only discuss membranes.
4. Bring documentation evidence. A qualification file, a datasheet you governed, a specification you cleaned up. Section IV explains why this is worth more than it looks.
5. Track shipped output, not capacity targets. 90 MW by 2027 and up to 3 GW by 2032 are declared ambitions. Megawatts shipped are not published. Time your move on evidence of production, not on evidence of intent.

VII. Refused in print

This dossier contains no figure for any of the following, because none is published in a dated, attributable source:

• salaries, salary bands or compensation for any role at any site;
• current headcount, at the company or at any site, and any headcount growth figure or hiring target;
• the number of open vacancies, or any breakdown of them by discipline (the company's public careers link resolved to a "page not found" document when we checked it);
• any job-creation figure attached to the €64 million financing announced in March 2024 or to the Belfort factory;
• the number of training places, apprenticeships or alternance contracts offered by this company or any regional programme serving it;
• any count of hydrogen jobs in Auvergne-Rhône-Alpes or in France, and any skills-shortage percentage;
• any measured attrition, time-to-hire or vacancy-duration figure;
• the amount of the regional and European co-financing behind the company's declared HyPace research project.

What remains is a structural argument that does not need those numbers: the delivered product is a container, the container is mostly electrical and thermal integration wrapped in a safety case, the design roles are declared in the Grenoble metropole and the manufacturing roles are declared 400 km away by road as of 2026, and the adjacent regional pools — traction, nuclear, process, data centre — already hold most of the capability required. That is a plan a reader can act on this quarter, and every load-bearing sentence in it points at a page we can name.