France’s deeptech sector is showing solid indicators: more than 2,800 startups tracked, 410 new ventures created in 2025, and a fourth-place worldwide ranking among deeptech ecosystems when it comes to fundraising. Europe now attracts nearly 20% of global deeptech investments. Yet behind these figures lies a reality far more heterogeneous than sector-based approaches typically reveal.
This is the starting point of Deeptech Dynamics 2026, the first edition of the study published by CentraleSupélec (through its 21st accelerator) with the support of Bpifrance and France Deeptech.
The main finding: classifying deeptech startups by sector (health, space, energy, quantum, etc.) no longer suffices to explain their real innovation trajectories or their market access strategies.
A quantitative and qualitative methodology
The study rests on a dual approach. The quantitative component analyzes a cohort of 242 deeptech startups that raised more than €15 million between January 1, 2010 and December 31, 2025, spanning stages from Seed to Series D. Health biotechs were excluded from the scope. Data come from the Dealroom database.
The qualitative component is built on about forty interviews with startup founders (15) and managing directors of specialized investment funds (25).
Deeptech innovation vs advanced engineering: a fundamental distinction
The study starts by drawing a distinction often poorly understood: that between advanced engineering and deeptech innovation.
Advanced engineering consists of developing new knowledge from identified concepts, for markets where the value is already established (example: NVIDIA’s Rubin platform for AI data centers).
Deeptech innovation involves exploring new concepts while simultaneously bringing forth the markets associated with them, implying a transformation of the existing value chain (example: NVIDIA’s COSMOS platform for robotics and autonomous vehicles).
Deeptech innovation can be seen as a process of double expansion: on one side, product architectures that may stay the same or undergo deep transformations; on the other, value with uses that can stay close to the existing or introduce major ruptures.
It is the combination of these two axes that underpins the Deeptech Quadrant.
The Deeptech Quadrant: four archetypes of startups
From this analytical framework emerge four major categories of deeptech startups, each representing distinct innovation challenges and very different market-access trajectories:
| Archetype | Definition | Example | Key data |
| Performers | Improve the performance of dominant solutions in an existing industry, with a conserved product architecture. | Expliseat (ultra-light aircraft seats) | 86 startups — €60M avg. — Series B in 6.0 years |
| Transformers | Develop a new architecture leading to a new standard within an industry, redefining power dynamics and business models. | Genomines (phytoextraction of nickel) | 98 startups — €52M avg. — Series B in 5.5 years |
| Explorers | Explore different uses for polyform innovations across new markets, testing multiple sectors simultaneously. | Unseenlabs (RF space detection) | 22 startups — €38M avg. — Series B in 6.0 years |
| World Builders | Create entirely new industry corridors from new architectures, building product, market and regulatory framework in parallel. | Alice & Bob (quantum computing) | 36 startups — €205M avg. — Series B in 5.0 years |
This framework breaks with a strictly sectoral reading. It reveals that deeptech does not follow a single trajectory but rather differentiated paths depending on the intensity of value-chain transformation and the scale of value exploration.
Radically different go-to-market strategies
One of the essential contributions of the study is to show that each archetype requires a market-access strategy that is specific, often incompatible with generic playbooks usually promoted in the startup ecosystem.
Performers: rapid execution and early customer validation
For Performers, the classic MVP approach remains pertinent. The challenge is to quickly demonstrate technical superiority and return on investment in real-world conditions, with pilot customers and prescribers.
Expliseat embodies this profile. The aerospace startup designed an ultra-light aircraft seat that reduces seat weight by 35% compared with standard seats, with more than 40 customers in 15 countries. Its trajectory illustrates rapid execution in an existing market: EASA certification obtained between 2011 and 2014, scaling toward single-aisle aircraft from 2019, followed by international expansion.
Transformers: co-construction with design partners
For Transformers, traditional approaches do not work. Value only emerges if clients are willing to transform their processes or their business lines.
The study recommends an inception strategy: co-create the offering with design partners, industrialize progressively, and guide the regulatory framework. Genomines, which extracts nickel by phytoextraction (a low-carbon biological method), illustrates this archetype with a €45M fundraising in 2025 to finance international expansion.
Explorers: platform logic and gradual navigation
Explorers cannot deploy a classic MVP, nor can they focus on a single target market.
The study advocates a strategy of “co-exploration”: design a flexible product that covers the essential base functionalities, adaptable to different use cases, and advance market by market starting with the less risky mid-market segments.
Unseenlabs, a specialist in space-based RF detection applied to maritime surveillance, initially targeted combatting illegal fishing before expanding to military applications. Its 2024 Series C round of €85M funds a constellation of 25 satellites by 2026.
“A technology only makes sense if it finds its market. We made a clear choice to focus first on maritime surveillance and then to broaden.”
Clément Galic, CEO and co-founder of Unseenlabs
World Builders: terraforming an entire supply chain
World Builders face the most ambitious challenge: creating simultaneously the product, the infrastructure, and the market.
The study describes a strategy of “terraforming”: laying the first bricks of infrastructure, structuring the ecosystem with integrators, regulators and early adopters, and then gradually expanding.
Alice & Bob, which is developing a universal, error-tolerant quantum computer via its “cat qubits,” aims to position itself as a supplier of quantum computers for industry by 2030, from a 4,000-square-meter lab near Paris.
Product-market fit reinvented by archetype
Deeptech Dynamics 2026 proposes a reimagined concept of product-market fit (PMF) applied to deeptech. Far from universal, PMF takes a radically different shape depending on the archetype:
| Archetype | Innovation strategy | Product-market fit |
| Performers | Lean innovation: lean prototyping, rapid iterations with customers. | Test the value proposition in an existing market, without challenging the regulatory framework. |
| Transformers | Inception: introduce and deploy a new process with design partners. | Explore a pioneer frontier by activating architecture, workflows, market fragments, and regulation. |
| Explorers | Navigation: explore new applications and gradually map value potential. | Build a structured path to market rather than seeking immediate PMF. |
| World Builders | Terraformation: shape a new supply chain link by link. | Simultaneously bring forth an offering, an infrastructure, new processes and an ecosystem of actors. |
The absence of fit with an intermediary market should not be seen as a failure for an Explorer or a World Builder, but as a source of essential learnings. Evaluating these startups against linear-model standards risks prematurely concluding a lack of opportunity.
Implications for the entire ecosystem
Beyond entrepreneurs and investors, the study offers concrete recommendations for several key players:
- For the leadership teams of valuations and the OTTs: identify a project’s archetype as early as possible in the pre-maturation phase to tailor support. Pushing an Explorer or a World Builder to focus too soon on a single sector can be counterproductive.
- For University Innovation Hubs (UIHs): move beyond gatekeeping to play the role of architect, orchestrating differentiated support according to archetypes.
- For Bpifrance and the DGE: tailor the Deeptech Plan to the capital-intensive needs of Transformers and World Builders, which require substantial investments in infrastructure and industrialization before achieving meaningful commercial traction.
“Making the deeptech ecosystem more legible is today a strategic priority. It is an essential condition to mobilize all actors and attract more capital to support the structuring and scaling of the sector.”
Pascale Ribon, Deeptech Director, Bpifrance
A shared reference framework for tomorrow’s deeptech
Driven by the 21st accelerator at CentraleSupélec, which supports 50 deeptech startups each year, Deeptech Dynamics 2026 aims to become a shared reference framework for the entire ecosystem. France Deeptech plans to use it to raise awareness among investment funds and “move the needle in how they identify and classify the startups they fund and support.”
