France spends more than its neighbors to fund public research. It files more university patents than all its European partners. And yet French companies buy less science from its laboratories than German companies do: six times less according to data from the early 2010s, and the ratio has barely moved since. This is a paradox that no injection of funding resolves, because it is not a problem of volume but of architecture.
Subsidizing the exit when the link is missing at the entry
In 2023, companies operating in France financed roughly 6% of research conducted in government agencies and universities—the domestic R&D spending of government administrations, known as DIRDA. In Germany, this share reached roughly 15% about fifteen years ago, against barely 2% in France, a gap that twenty years of incentive schemes have not closed. The share of government agencies and universities is higher in France than in the OECD average, so the spending they carry out is ultimately of similar magnitude as the OECD average, but funded very differently.
The French response to this slippage is well known: subsidize transfer on the way out. The Carnot institutes, created in 2006, are the most structured example. In 2024, 39 structures benefit from the Carnot label, coupled with financial support of 116 million euros calculated according to the volume of contracts billed to their socioeconomic partners, which reached 636 million euros in 2023.
This is a logic of payment for results: the more a laboratory sells research to companies, the more public money it receives. The principle is sound. The result, measured against the economy it is meant to nourish, is modest.
In 2023, according to a survey by the CURIE network, all public research valorization activities generated 822 million euros in research contracts, with 1,128 contracts signed and 735 priority patent applications. For a system employing more than 110,000 researchers in full-time equivalent, this is a flow that says something about the distance between the academic world and the industrial fabric. In 2023, 1,669 patent applications resulting from French public research were filed at the European Patent Office, positioning France in third place in the OECD. Third place for filing, but without these patents becoming massively turned into products.
Thirty years of schemes addressing an architectural problem
The causal chain of the blockage has been documented for a long time. The State spends, laboratories publish and patent, intermediate structures try to make the connection, companies do not buy enough. This is not a problem of bad faith. It is a problem of demand.
In the United States, public procurement of innovation toward small businesses and startups has been mandatory since 1982. The SBIR program (federal small business innovation fund) requires each federal agency spending more than 100 million dollars on outsourced R&D to reserve at least 3.2% of that budget for contracts with innovative SMEs. Funds are obtained by taking a percentage of the outsourced R&D budgets of eleven federal agencies, representing approximately 2.5 billion dollars allocated each year. The mechanism does not subsidize the existence of an intermediary: it creates solvable demand that forces small companies to seek their science in laboratories, and laboratories to organize themselves to deliver it.
In France, the equivalent does not exist. Public procurement represents 111 billion euros (figures from 2020) and constitutes a major lever for supporting innovative industries in many countries, but this tool remains very little used in France. Only 1.75 billion euros was dedicated to startups under public procurement in 2022. The Senate noted that 70% of public buyers declare they are not trained in innovative procurement. It is a marginal budget line where it is a central mechanism elsewhere.
The consequence is that transfer happens mostly in one direction: from public funds to structures hoping to convince companies. The other direction, from companies ordering science from laboratories because they need that science to fulfill a public contract, is systematically underdeveloped. Philippe Aghion, winner of the 2025 Nobel Prize in Economics, states it clearly when comparing France to the United States: in terms of valorizing patents and academic work, the United States strongly encourages innovators; they have technology transfer offices that France does not have.
France 2030 money without the demand to capture it
The France 2030 plan represented a massive effort: 38 billion euros invested over three years, with 155,000 jobs directly created and 7,500 projects supported, including more than 6,000 patents filed; fourteen of the sixteen performance indicators are on track with the trajectory. The Deeptech plan, carried out by Bpifrance within this framework, displays creation results: 2,830 active startups, 50,000 direct jobs, 5.4 billion euros in revenue in 2025.
But the pool of young companies has reached critical size and must now materialize through the emergence of industrial and technological leaders: and that is precisely where the system gets stuck. The DGE estimated in 2025 that deeptech startups will need to mobilize at least 30 billion euros of public and private financing by 2030 to stay on track. And the Court of Auditors, in its assessment of France 2030 at the end of 2024, noted that the targeting of public investments should be tightened and that the intervention doctrine should focus more on the most effective instruments by limiting recourse to subsidies.
The problem is structural. You can create deeptech startups with subsidies. You cannot make them grow into industry without markets. And French public markets, too fragmented, too scattered between the State (8% of the total), local authorities (80%), and network operators (12%), are not steered according to innovation objectives. This is a continent of spending (170 billion euros in markets above 90,000 euros in 2023, or twice as much as ten years ago) without a strategy for directing innovation from public research.
Procurement as a mechanism, not as a reward
The arbitration for 2027 is at hand. France can continue financing transfer intermediaries (Carnot institutes, support schemes, maturation funds) whose positive evaluation remains conditional on the commercial goodwill of companies. Or it can change who decides the demand for science, by introducing a mandatory innovation procurement obligation targeting SMEs and startups resulting from public research, modeled on the American mechanism.
Concretely, this means writing into law that any public program financing outsourced R&D beyond a threshold (ministries, agencies, major network operators) reserves a binding fraction of its purchases for companies whose technology comes from public research and is documented by a patent or license. This is not a criterion in a call for bids: it is an architectural obligation that creates solvable, regular, and predictable demand where there is none.
The mechanism changes three things simultaneously. It shifts the purchasing decision: today, the choice to buy from a deeptech startup falls to each buyer, alone facing their criminal liability. Tomorrow, the legal framework sets them an obligation of results. It shifts the basis of transfer funding: instead of subsidies conditional on hypothetical contracts, a fraction of existing public procurement spending directly finances uses. And it changes the relationship between laboratories and companies: instead of a laboratory seeking to sell a solution to a company that does not need it, a company that needs a solution seeks the laboratory that can supply it.
Public procurement presents many advantages over subsidies: it strengthens the position of emerging players facing their investors, constitutes a client reference that counts with private clients and other states, and drives economic activity in France. Dani Rodrik, in his work on conditional industrial policy, emphasizes precisely this point: an industrial policy that creates demand is more robust than one that finances supply, because it subjects innovation to the real test of the market rather than the judgment of a selection committee [11]. Aghion cites DARPA (which is first and foremost a military procurement program) as a model for reconciling industrial policy and competition policy.
After this change, public laboratories are no longer waiting for a hypothetical client: they know that the State and its operators have the obligation to be one. This is not a guarantee of commercial success (the private market remains the final judge) but it is the creation of a first systematic client that the French system has never had.
Sources
[1] SIES (MESR), “Domestic Research and Experimental Development Spending in France: Detailed Results 2023 and First Estimates 2024,” Information Note No. 25-12, November 2025, https://www.enseignementsup-recherche.gouv.fr
[2] CURIE Network, “Survey on Public Research Valorization: 2024 Report (2023 Data),” January 2025, https://www.curie.asso.fr/reseau-curie/valorisation/la-valorisation-en-chiffres
[3] General Inspection of Finance / General Council for Economics, “Evaluation of the Carnot Institutes Scheme,” December 2024, https://www.igf.finances.gouv.fr and https://www.economie.gouv.fr/cge/evaluation-du-dispositif-des-instituts-carnot
[4] SIES (MESR), State of Higher Education, Research and Innovation in France No. 19, fact sheet “Financing and Execution of R&D in Companies in France,” 2024, https://publication.enseignementsup-recherche.gouv.fr/eesr/FR/T068
[5] France Stratégie / DG Treasury, “French Public Research in International Comparison,” working note, https://www.tresor.economie.gouv.fr
[6] Bpifrance, “Deeptech Barometer 2025: Six Years of the Deeptech Plan,” March 2025, https://presse.bpifrance.fr and https://www.bpifrance.fr/le-plan-deeptech
[7] Directorate General for Enterprises (DGE), “Disruptive Innovation Facing the Scale-Up Challenge,” report, March 2025, https://www.entreprises.gouv.fr/files/files/Publications/2025/Etudes/202503-rapport-innovation-rupture-defi-passage-echelle.pdf
[8] French Government, “France 2030 Assessment: Interministerial Innovation Committee,” April 2025, https://www.info.gouv.fr/actualite/bilan-de-france-2030-ce-quil-faut-retenir
[9] Court of Auditors, “2024 Budget Execution Note: Invest for France 2030 Mission,” April 2025, https://www.ccomptes.fr/sites/default/files/2025-04/NEB-2024-Investir-pour-France-2030.pdf
[10] Senate, “Converting the Innovation Experiment: An Imperative for Reindustrializing France,” Report No. 655, 2021-2022, https://www.senat.fr/rap/r21-655/r21-655_mono.html
[11] Dani Rodrik, Shared Prosperity in a Fractured World, Princeton University Press, 2025.
[12] Philippe Aghion, “An Intellectual Trajectory: Grand Interview with the 2025 Nobel Prize Winner in Economics,” Le Grand Continent, November 2025, https://legrandcontinent.eu/fr/2025/11/02/philippe-aghion-prix-nobel-deconomie-2025-grand-entretien/
[13] Senate, Inquiry Commission, “The Urgency to Act to Avoid Going Off Course: Steering Public Procurement in Service of Economic Sovereignty,” Report No. 830, July 2025, https://www.senat.fr/rap/r24-830-1
[14] Small Business Administration (United States), SBIR Program, Annual Data, https://www.sbir.gov
[15] CURIE Network, White Paper on Valorization: Valorization Assizes, 2024, https://www.curie.asso.fr/sites/default/files/inline-files/Livre%20blanc%20de%20la%20valo%20VF%202024_0.pdf
[16] HCERES / OST, “France’s Technological Position Through Patents Filed at the European Patent Office,” EESR No. 19, Table 34, 2024, https://publication.enseignementsup-recherche.gouv.fr/eesr/FR/T218



