In the laboratory, the material works. At the prospective customer’s site, everything remains to be proven: will it last ten years, will it be cheap enough, can it be manufactured in quantities beyond a few grams? For a deeptech startup, this distance between discovery and use is measured in testing campaigns, hard-to-find talent and cash burn. The challenge is not simply to raise large sums. It is to find, at each stage, funding patient enough to wait for the next piece of evidence. Looking ahead to September 2026, this question is becoming as much an issue of industrial sovereignty as an entrepreneurial challenge.
A journey, not a giant leap
The “valley of death” suggests a single obstacle. There are several. A technology must work outside the laboratory, meet a need that customers are willing and able to pay for, potentially clear regulatory hurdles, and then be produced with consistent quality. A quantum sensor, a chemical process and an innovative therapy follow neither the same timetable nor the same funding routes.
The tools discussed here draw on programmes and trends documented before September 2026; their eligibility requirements should be checked when applying. The perspectives offered are not an assessment of the situation as of that date. In France, the Deeptech Plan led by Bpifrance since 2019, followed by France 2030, has established a continuum of support. But no programme can replace a strategy linking each expense to a specific uncertainty.
Fund evidence, not a broad promise
The first task is to turn the scientific roadmap into fundable milestones. “Improve the prototype” is not enough. Demonstrating performance in an industrial environment, measuring reproducibility or validating a manufacturing cost gives a funding provider something more concrete to assess. Each stage must include a budget, a deadline, a success criterion and a decision on what to do if it fails.
Consider a company developing a filtration membrane. Its initial funding might be used to reproduce results across several batches. The next round funds a module tested with actual water from an industrial company. A pilot line then follows to verify yield and service life. Building that line before characterising fouling would amount to funding a factory on the basis of a still-fragile hypothesis.
The right management metric is risk eliminated per euro spent. Technology readiness levels, or TRLs, provide a common language, but they do not tell the whole story: an advanced prototype may remain unsellable if its installation disrupts the customer’s production. Commercial, regulatory and industrial readiness must therefore also be tracked.
Intellectual property: an asset and a source of friction
Before convincing an investor, a company must first establish what it owns. An invention arising from public research often belongs to the institution or to several co-owners. The startup must then negotiate a licence, notably with a technology transfer acceleration company, known as a SATT, or the relevant technology transfer office. This work deserves an early start.
Exclusivity, territories, fields of use, royalties, access to know-how: these clauses determine the licence’s real value. Funding providers will also examine development obligations, termination conditions and the treatment of future improvements. An agreement that is acceptable at launch can become a hindrance during a funding round or a partnership negotiation if it leaves too much uncertainty.
Another essential distinction: holding a patent does not guarantee freedom to operate. Other rights may block commercialisation. Specialist analysis should inform market choices. Finally, publishing too early can compromise protection. Dialogue between researchers and entrepreneurs must be organised, without confusing useful protection with the costly accumulation of patents.
Public funding to absorb early-stage uncertainty
Grants and repayable advances are particularly relevant when scientific risk remains high. The i-Lab and i-Nov competitions, Bpifrance funding, certain France 2030 calls for proposals and European instruments can finance different stages. The EIC Accelerator notably offers support combining grants and investment, subject to the applicable rules.
Their value goes beyond avoiding dilution: this funding can make possible an experiment that private capital would consider premature. But it comes with eligible expenditure requirements, supporting documentation and timetables. Rules on combining funding sources vary. An award does not necessarily mean the money is received immediately; co-financing or bridge funding may still be essential.
The research tax credit can also contribute to funding, but should not be treated as a grant available from day one. The company must document the eligibility of its work and anticipate delays. The danger is building a strategy around calls for proposals: multiplying applications can stretch a small team thin and fund peripheral work rather than the decisive evidence.
Venture capital must buy productive time
Equity investors finance what public support covers poorly: staffing, business development, intellectual property, unforeseen costs and preparation for the next stage. But companies must choose partners capable of understanding scientific cycles. A fund accustomed to software may underestimate the time needed for industrial qualification or a regulatory procedure.
The discussion should cover the reserves available for subsequent rounds, the fund’s investment horizon and its ability to support the company through a delay. Raising money too early at an ambitious valuation can become a trap if the evidence is slow to emerge. Raising too little means having to seek more money before reaching a credible milestone. The appropriate amount therefore includes a buffer for experimental failure, not just the ideal scenario.
The industrial partner: an accelerator, with conditions
An industrial company brings resources that money alone cannot quickly buy: data, a test site, production expertise and access to buyers. The partnership becomes particularly useful when the pilot also tests integration constraints and purchasing criteria. A technically successful trial with no identified customer budget remains a demonstration, not a route to sales.
The contract must specify who pays for the trials, owns the results and can reuse them. Overly broad exclusivity can close off the market; dependence on a single partner weakens future negotiations. Where possible, exclusivity is better limited to a particular use or period, in exchange for concrete commitments. And a paid pilot generally provides a stronger commercial signal than a letter of interest.
Scaling up without building everything in-house
Scaling up requires costing equipment, qualification, maintenance, scrap and working capital needs. The first order can even intensify cash flow pressure: purchases and production must come before payment. Outsourcing, a shared production line or licensed manufacturing can limit the initial investment, provided critical know-how is preserved. Debt becomes more accessible when assets, contracts and future cash flows provide sufficient visibility.
What now? Looking ahead to September 2026, the challenge may lie less in multiplying funding schemes than in coordinating them around evidence that can be shared. For entrepreneurs, the discipline remains simple to state: protect before disclosing, test before building, put agreements in place before making promises. Deeptech will not eliminate long development cycles. It can, however, avoid paying too much for the wrong steps.
