In the laboratory, the result fits on a workbench. At a prospective customer’s site, it will have to run for eight hours straight, withstand vibrations, meet a standard and cost less than the problem it solves. Between these two worlds, deeptech consumes time and capital. In September 2026, the entrepreneurial challenge remains the same: funding not just an invention, but its transformation into a solution customers can buy. Programs and trends already documented provide solid reference points; their extrapolation here is forward-looking analysis, not a quantitative assessment of 2026.
The real valley of death is as much commercial as technical
A low-carbon material, a quantum sensor or a fermentation process does not follow the same timeline as software. Companies sometimes need to build several generations of prototypes, access scarce equipment and obtain approvals before they can issue an invoice. Above all, scientific success does not automatically answer three questions: who pays, from which budget and following what validation?
Funding must therefore support two parallel paths: technological maturity and commercial maturity. A high-performing demonstrator may find no market if no one has checked how it integrates into an industrial production line. Conversely, an enthusiastic letter of interest is not an order. Every major expense should reduce an identified uncertainty: performance, cost, manufacturing, regulation or purchasing.
Patent licensing: securing rights without buying everything
The first task for a laboratory spinout is to establish who owns what. Depending on the institutions and research contracts involved, several organizations may hold the rights. In France, technology transfer offices, SATT technology transfer companies and research organizations’ specialist units support these negotiations. Scientific publication, meanwhile, must be coordinated with protection of the invention to avoid compromising its patentability.
Buying the patents outright immediately is not necessarily in the start-up’s best interests. A license can give it the rights it needs while preserving cash. But its scope must be negotiated: fields of application, territories, exclusivity, sublicensing, future improvements and termination conditions. Expensive worldwide exclusivity becomes a burden if the initial market is limited to a very specific application.
The financial balance matters as much as the royalty rate. Upfront payments, reimbursement of patent costs, annual minimums and milestone payments can pile up before any revenue arrives. The aim is to align cash outflows more closely with value creation: phased payments, realistic development obligations and periods to remedy breaches before termination. The institution must be able to generate value from its assets without making their commercial use financially unviable.
Another crucial distinction: holding a license does not guarantee freedom to operate. Third-party patents may block the final product. A targeted analysis, updated at key stages, helps avoid discovering this obstacle after the pilot has been built. Access to unpatented know-how, data and equipment must also be arranged: the patent alone does not always contain a reproducible recipe.
Build the demonstrator that triggers a decision
The classic trap is to build the most impressive prototype rather than the cheapest one capable of convincing a customer. Before starting construction, the team should obtain a written specification from an industrial company for the required test: environment, duration, performance thresholds, safety and measurement method. Without acceptance criteria, the demonstrator risks becoming an indefinitely extended experiment.
The right question is not “can we make it bigger?” but “what evidence is missing to move to the next stage?” For a chemical process, that may be stability over several cycles; for a sensor, reliability outside the laboratory. Renting a test platform, outsourcing a production run or using a technical center can avoid the premature purchase of a dedicated facility.
Three budgets rather than one
- The proof budget funds the experiment that overcomes the main scientific or technical barrier.
- The integration budget covers adaptation to the customer’s site, testing, qualification and documentation.
- The delivery budget anticipates components, quality control, insurance, maintenance and working capital requirements.
This separation makes visible what investor presentations often underestimate: a successful pilot does not fund its own industrial scale-up. The first order may even increase cash requirements if suppliers demand deposits and the customer pays after delivery.
Assemble funding according to the risks
France has funding building blocks established before 2026: maturation support from technology transfer organizations, Bpifrance assistance, the i-Lab and i-Nov competitions, and funding linked to France 2030. At the European level, the EIC offers instruments ranging from breakthrough research to acceleration. Eligibility, schedules and terms must nevertheless be checked for each call: no funding scheme provides automatic entitlement.
Grants are particularly suitable for work still too uncertain to be financed through sales. Equity funding offers greater flexibility to hire, pursue customers and absorb delays. Debt, meanwhile, requires a credible ability to repay or appropriate guarantee mechanisms. Diversifying funding sources only makes sense if their constraints remain compatible.
A funding award does not always mean cash is available. Staggered payments, eligible expenditure, supporting documentation and co-funding requirements can create a dangerous timing gap. The research tax credit is no substitute for immediately available cash, either. Financial planning must account for a delayed scenario: an unsuccessful test, hiring difficulties or a postponed purchasing decision. Preparation for the next funding round begins before that buffer disappears.
The industrial partner: a potential customer, not the owner by default
A large corporation brings a testing ground, expertise and sometimes funding. But a poorly structured partnership can leave the start-up stuck. The contract must distinguish pre-existing knowledge from jointly generated results, specify usage rights and establish confidentiality provisions. Exclusivity deserves something in return: funding, purchase commitments or a strictly limited duration.
A paid pilot remains a valuable signal, without guaranteeing a market. The operational lead must be identified, but so must the buyer and the budget holder. The contract can provide for billable stages, acceptance criteria and the conditions for a subsequent rollout. Vague commercial prospects alone should not justify several months of bespoke development.
A team capable of choosing what it will not do
Finally, governance must connect science and business. The roles of the researcher, the chief executive and the head of industrial operations must be explicit, with the necessary authorizations when public-sector employees participate in the project. An early commercial hire does not replace technical expertise; it helps identify the first use case with paying customers, the one requiring the fewest adaptations before sales become repeatable.
What next? In the coming years, the advantage could go to deeptech companies capable of sharing testing resources and signing contracts with customers earlier. This is no promise of a shorter journey: some breakthroughs will still take a long time to industrialize. But funding a succession of proven solutions that customers can buy, rather than abstract technical perfection, gives a discovery a better chance of becoming a sustainable business.
