Executive brief
Inocel is a Grenoble-area fuel-cell company whose engineering sits in Saint-Égrève and whose factory sits 400 kilometres away in Belfort. That split, not the cell itself, decides where the jobs land. In March 2024 the company announced a €64 million financing made of equity, debt and grants, and published a dated industrial plan against it: pre-series units in 2024, the Belfort production line commissioned at the end of that year, test tooling brought in-house, then a production ramp in the first half of 2025. The company today states a factory of 15,000 m² scaling to 90 MW by 2027 and a long-term capacity of up to 3 GW by 2032. This dossier reads that plan against what the company itself publishes — and finds three specifications that disagree with each other on its own website: 15 patents or 25, up to 65% efficiency or up to 60%, full power in under 2 seconds or under 1.5. We print all of them with their pages. Nine claim families we could not source are named and refused at the end.
I. The mechanism: a company split by function, not by geography
Most industrial dossiers begin with a product. This one begins with an address list, because the address list is the strategy. The company states three facilities and assigns each a different job:
| Site | Declared role | Declared content |
|---|---|---|
| Saint-Égrève (Isère, Grenoble metropole) | Headquarters & R&D | Stack development, balance-of-plant design, software & electrical |
| Belfort (Bourgogne-Franche-Comté) | Gigafactory, 15,000 m² | Industrialisation team, customer services; capacity scaling to 90 MW by 2027, up to 3 GW long-term by 2032 |
| Mougins (Alpes-Maritimes) | Advanced Technology | Complex system prototyping, powertrain |
Read that table as a careers document and it says something a press release never says out loud. The intellectual property — the stack, the balance of plant, the control software — is being built in the Grenoble metropole. The volume — the assembly, the line, the test benches, the customer-facing service organisation — is being built in Belfort. Prototyping and powertrain integration sit on the Côte d'Azur. A candidate who wants to design a fuel cell and a candidate who wants to manufacture one are applying to the same company in two different regions, and the second population is the one that grows with the ramp.
This is the structural point that survives whatever happens to any single figure below: Grenoble's stake in this company is a design stake, not a tonnage stake. The megawatts are declared for Belfort. Coverage that reports "Grenoble hydrogen gigafactory" gets the geography wrong, and a reader who moves cities on that basis pays for the error.
II. The €64 million, and what was promised against it
On 22 March 2024 the company announced a financing of €64 million. Two details of that announcement matter more than the headline number.
First, the composition. The financing is described as made of equity, debt and grants ("capitaux propres, dettes et subventions"), with the company's longstanding shareholders participating. A €64 million round of pure equity and a €64 million package that blends a shareholder cheque with bank debt and public subsidy are different objects with different risk. The first prices the company; the second part-finances a factory and part-reimburses a research programme. We do not report the blend as a valuation event, and we do not know the split — the split is not published, and it is refused below rather than estimated.
Second, the milestone list. The March 2024 announcement, as relayed by the national hydrogen industry association on 29 March 2024, attached the money to four dated commitments: finalise industrialisation of the product and deliver pre-series units in 2024; commission the Belfort factory's production line at the end of 2024; internalise its test tooling at the same moment; then ramp production in the first half of 2025 and accelerate commercial development. The same source records that many commercial contracts were "under discussion" — a phrase we treat as an absence of an order book, not as evidence of one.
That list is the single most useful artefact in this dossier, because it is falsifiable. A dated plan can be checked. Two and a half years later, the company's own public pages describe a factory that is scaling to 90 MW by 2027 — a forward target, stated in 2026, for a line the 2024 plan said would be commissioned by the end of 2024 and ramping by mid-2025. We do not conclude from that wording that the plan slipped; a company can be ramping and still publish a 2027 capacity target. We conclude only what the record supports: the achieved output is not published. No megawatts shipped, no units delivered, no revenue. That absence is the finding, and Section V refuses to fill it.
III. Public money is in the room, and it is regional
The company declares one named research programme: HyPace — Hydrogen Power Advanced Cell & Generation — described as an R&D project dedicated to high-power fuel cells and next-generation containerised hydrogen generators, supported by the Auvergne-Rhône-Alpes Region and co-financed by the European Union. The amount is not published on that page and we do not invent one.
Two consequences follow for a reader in the region. First, the containerised generator — the product family Section IV describes — is not a private roadmap; part of it is a publicly co-funded research programme, which is why its documentation is unusually explicit for a scale-up. Second, the regional authority co-funding high-power fuel-cell R&D is the same authority that co-funds much of the local semiconductor and battery research base. The hydrogen line in Auvergne-Rhône-Alpes is not adjacent to the electronics line; it draws on the same regional instrument.
The deeper lineage is older than the company. The company states its technology rests on more than 25 years of research conducted by the CEA, and that its 300 kW fuel cell is "built on over twenty-five years of research and development conducted at the CEA". The origin story it publishes is consistent with that: a 2020 convergence between a Dakar Rally ambition and an industrial founder's read of high-power decarbonised energy, formed into the GEN Z project in collaboration with the CEA; incorporation in 2022, an advisory board, a CEO appointment, and a first round of €16.5 million. In other words, this is a technology-transfer company. The research subsidy that produced the stack was spent in Grenoble over two decades before the company existed — which is precisely why the design jobs are in Saint-Égrève and not in Belfort.
IV. The product, and the three numbers that do not agree
The published system is a proton-exchange-membrane (PEM) fuel cell, the Z300 stack, integrated into a stationary system (Z300-S) and into containerised generators branded GEN-Z in three sizes: GEN-Z 300, GEN-Z 600 and GEN-Z 1300. The company publishes hard specifications for the cell, according to its fuel-cell page consulted in September 2026:
| Specification, as published | Value |
|---|---|
| Type | PEM (proton exchange membrane) |
| Power | > 300 kW |
| Dimensions | 440 × 450 × 491 mm |
| Weight | 136 kg |
| Volume | 97 l |
| Power ramp-up | < 2 s (stated elsewhere on the same page as < 1.5 s) |
| Efficiency | up to 65% (stated elsewhere on the same page as up to 60%) |
| Patents | 15 (fuel-cell page) / 25 (generator page) |
The disagreements below are quoted from those same pages as of September 2026, and are worth naming precisely, because they are not typos in a newspaper — they are two numbers published by the same company on the same site, and a reader deciding whether to believe the technology should see them.
- Efficiency. The fuel-cell page states "Efficiency up to 65%" in its key-features block and, three paragraphs later, "With an efficiency of up to 60%". Both are ceilings ("up to"), so neither is a rated operating point.
- Transient response. The same page states a power ramp-up of "< 2 s" as a specification and "reaches its maximum power in less than 1.5 seconds" as a benefit.
- Patents. The fuel-cell page states the system "incorporates 15 patents covering both hardware and software". The generator page states the technology is "Powered by 25 Patents". The count may have grown between page revisions; neither page is dated, so we cannot say which is current.
One further unit shift deserves flagging. The cell is specified in kilowatts (> 300 kW); the GEN-Z 300 generator is described as covering "power needs from a few kVA up to 300 kVA". Kilowatts and kilovolt-amperes are not the same quantity — apparent power exceeds real power at any power factor below unity — so the generator's 300 kVA rating and the stack's 300 kW rating are not interchangeable, and we do not treat them as one number. The generator page also states, as of September 2026, "99.99% Power Availability" with no measurement basis, test population or observation window; we report it as a marketing claim, not as a reliability figure.
What is unambiguous, and industrially interesting, is the design intent: high power density in a small envelope (a 300 kW-class stack in 97 litres and 136 kg), fast transient response explicitly aimed at reducing the size of the auxiliary battery, and modularity from a few hundred kilowatts to several megawatts. That combination targets a specific market — one where diesel currently wins.
V. The commercial mechanism: containerisation, and the diesel comparison
On 13 January 2025 Equans and Inocel published a formalised partnership to bring a fuel-cell energy generation and storage solution to market. The division of labour is stated plainly and is the most concrete commercial fact in the public record: Inocel supplies the fuel cell; Equans provides the containerisation, and where a customer wants the full hydrogen chain — production, storage, generation — the two combine and lean on their partner networks. The stated envelope is 300 kW to several megawatts, available from 2025, with named use cases: cold ironing for ships at quayside, marine propulsion and onboard generation, storing intermittent renewables as hydrogen for later use, grid balancing and load-shedding, backup power for critical industrial and tertiary infrastructure including data centres, hospitals and banks, and mobile generators for off-grid sites replacing diesel units and batteries.
That list explains the strategy better than any market forecast. A high-power fuel cell does not compete with a battery on cost per kilowatt-hour, and it does not compete with the grid where the grid is available. It competes with a diesel genset in exactly the places diesel still wins: where connection is absent, delayed or capped, and where runtime matters more than round-trip efficiency.
The consequence for the region is direct and under-reported: the customer that makes this economic in 2026 is not a truck. It is a data centre or a critical facility waiting for a grid connection. That is a demand curve set by electricity-network queues, not by hydrogen policy — and it is the reason the company's own product page leads with data centres, construction sites, events and remote operations rather than heavy mobility.
VI. What this changes for a reader in Isère
Three conclusions a reader can act on, each traceable to a published source above.
1. The design work is here; the volume work is not. Stack development, balance-of-plant design, software and electrical engineering are declared for Saint-Égrève. Industrialisation and customer service are declared for Belfort. If your skill is process, line ramp-up or production test, the declared employer address is Belfort — a different region, a different labour market and a different commute.
2. The company's own value chain names your customer. The Equans partnership means the product a customer buys is a container, not a stack. The integration skills that partnership consumes — high-voltage electrical conversion, cooling, controls, hydrogen safety and site works — are as decisive to delivery as membrane chemistry, and they are the skills the companion dossier examines.
3. Judge the ramp by shipped output, not by capacity targets. A 15,000 m² building and the 3 GW long-term ambition the company states for 2032 are real facts about intent. Neither is a fact about production. Until megawatts shipped, units delivered or revenue are published, the honest verdict on the ramp is not measured — and this publication will say not measured rather than borrow a number from a forecast.
VII. Refused in print
Nine claim families are absent from this dossier because we could not source them to a dated, attributable document. They are named here so their absence cannot be mistaken for an oversight — and so that any future round that finds them knows exactly what it is filling:
- the split of the €64 million between equity, debt and grants, and the identity of the lenders or grant instruments;
- any current headcount, for the company or for any of its three sites;
- any revenue, order book, backlog or delivered-unit figure;
- any megawatts actually produced or shipped from Belfort, and any confirmation that the 90 MW-by-2027 or 3 GW-by-2032 trajectories are on schedule;
- any cost per kilowatt, price per unit, or cost comparison against a diesel genset in euros;
- any independently measured efficiency, durability or lifetime figure for the Z300;
- any salary, salary band or compensation figure for any role at any site;
- any job-creation figure attached to the €64 million financing or to the Belfort factory;
- the amount of the HyPace regional and European co-financing.
What remains after those refusals is still a substantial industrial statement: a CEA-derived, high-power PEM stack, designed in the Grenoble metropole, industrialised in Belfort, sold as a container through a large integrator, into a market defined by grid queues and diesel replacement, on money that mixes shareholders, banks and public subsidy. That is enough to plan a career around. The megawatt count is not yet public, and no reader is served by pretending otherwise.