Deep Tech Innovation: Support Archetypes
Deep tech needs an architecture, not scattered programs. This article maps 10 institutional models that states use to systematically finance, scale and govern deep-tech innovation.
Deep-tech innovation has become the central battleground of economic power, national security and civilisational resilience. Artificial intelligence, quantum technologies, advanced materials, new energy systems, synthetic biology and robotics are no longer speculative curiosities. They are rapidly becoming the backbone of industrial competitiveness and state capacity. Yet the way most countries support this transformation is still fragmented: individual grants, isolated accelerators, occasional “moonshot” announcements. The result is a patchwork of initiatives rather than a coherent architecture able to consistently turn science into deployable capabilities.
Deep tech is structurally different from the digital innovation wave that came before it. It relies on long research lead times, capital-intensive infrastructure, complex regulatory landscapes and highly specialised talent. The risk profile is dominated by deep technical uncertainty and system-level integration challenges, not just product–market fit. This means that generic startup instruments, designed for consumer or SaaS ventures, under-serve deep-tech founders and under-price the public interest in strategic domains like energy, health and security.
For states and regions that want to be more than passive consumers of foreign technology, the central question is therefore not “how many startups can we create?”, but “what institutional machinery do we need to repeatedly generate, finance and scale deep-tech capabilities?”. The most successful ecosystems have converged, often independently, on a set of recognisable institutional archetypes: mission-oriented R&D agencies, blended finance vehicles, development banks, university commercialisation pipelines, specialised incubators, founder programmes, and more.
This article argues that these archetypes can be treated as modular components of a national or regional deep-tech strategy. Instead of copying isolated success stories (“we should have a DARPA” or “we need an EIC-style fund”), policymakers should think in terms of a system design: which modules are present, which are missing, how they connect, and how they can be standardised. By looking across countries and sectors, we can extract the common design principles that make these institutions work, regardless of local political or administrative culture.
We begin with the breakthrough engine of the system: ARPA-style mission agencies that deliberately fund high-risk, high-reward portfolios in strategic domains such as defence, energy and health. These agencies define ambitious capability goals, empower temporary programme managers, and accept high failure rates in exchange for sporadic but transformative successes. They supply the pipeline with radical ideas and prototypes that would not exist under conventional research funding.
The article then turns to the financing spine of deep tech. Here we examine blended grant–equity vehicles like the European Innovation Council, development banks and public venture funds such as Bpifrance or High-Tech Gründerfonds, and co-investment schemes that anchor specialist deep-tech VCs. Together, these institutions form a capital stack capable of handling both technology risk and market risk, while crowding in private investors who would otherwise stay on the sidelines.
On the supply side of ideas and talent, we look at university and national-lab commercialisation pipelines, and at structured founder development programmes such as NSF’s I-Corps and entrepreneurial doctoral schools. These institutions turn public research into a steady flow of spin-offs and train scientists to act as entrepreneurs, not just inventors. Standardised IP frameworks, proof-of-concept funds and entrepreneurial curricula are the building blocks of a scalable dealflow factory.
The middle of the pipeline is occupied by deep-tech incubators and accelerators and by shared clusters and testbeds. National innovation centres, specialist incubators and university-anchored cells provide deep-tech startups with access to labs, pilot lines, test environments and corporate partners. Testbeds and clusters—whether in energy systems, advanced networks or AI hardware—play a dual role: they reduce capital costs for individual firms and create geographic concentrations of expertise and investment.
No deep-tech architecture can function without a coordination and demand layer. National deep-tech strategies and regulatory frameworks define missions, budgets, and institutional roles, ensuring that agencies are aligned rather than working at cross-purposes. Innovation-oriented procurement and regulatory sandboxes turn the state into a sophisticated lead customer, providing early markets and regulatory learning for technologies that would otherwise remain stuck in demonstration mode.
Across the article, we repeatedly move from principle to player: each archetype is linked to concrete institutions—DARPA and ARPA-E for mission agencies, EIC and Bpifrance for blended finance, SGInnovate and Digital Catapult for deep-tech incubators, Oxford and IIT-Madras for spin-off pipelines, as well as Israeli, Indian, European and US examples of strategies, sandboxes and co-investment funds. This allows us to anchor abstract design patterns in observable practice and measurable outcomes, from spin-off counts and follow-on capital to jobs and industrial facilities.
The goal is not to prescribe a single universal model, but to offer a toolkit for strategic design. By the end of the article, readers should be able to map their own country or region onto this architecture, identify missing or weak modules, and derive concrete priorities: where to introduce ARPA-style programmes, how to structure blended finance, how to standardise spin-off rules, or how to use procurement to pull deep tech into real markets. Deep-tech competitiveness is ultimately a question of institutional intelligence; this article is an attempt to make that intelligence legible and reusable.
Summary
1. The core idea: deep tech needs an architecture, not just isolated programs
Deep tech (AI, biotech, quantum, new materials, energy, robotics…) is structurally different from SaaS or consumer apps:
It is more capital-intensive,
It is slower to mature,
It lives in regulated, mission-critical domains,
It depends heavily on public science and national infrastructure.
Because of that, you don’t get a healthy deep-tech ecosystem by randomly sprinkling grants and accelerators. Countries that succeed build an architecture: a set of coordinated institutional “modules”, each solving a specific failure of the market or the state.
The ten archetypes we mapped are those modules. They recur across countries and can be standardised, copied and combined.
2. ARPA-style agencies: the “breakthrough engine”
What they do:
ARPA agencies (DARPA, ARPA-E, ARPA-H, ARIA, others) are small, mission-driven, high-risk R&D funders. They live between basic science and markets and fund aggressive portfolios of projects aimed at specific strategic capabilities (defence, energy, health).
How they operate:
Lean and flat organisations with a few dozen program managers (PMs).
Temporary, empowered PMs (3–5 years) who design entire programmes: goals, metrics, calls, portfolios.
Flexible contracting and high failure tolerance – many projects fail, a few become foundational technologies (internet, GPS, new energy tech, etc.).
Deliberate transition logic – every program has a plan for how technologies leave the lab (procurement, further grants, industry, investors).
Why they matter:
They are the system’s breakthrough engine: they push the technical frontier forward in mission-critical areas and generate high-potential “raw material” for later modules (incubators, development banks, clusters).
3. Blended grant–equity investors: the “valley-of-death bridge”
What they do:
Institutions like the European Innovation Council (EIC) and the French Deeptech Plan (via Bpifrance) behave as deep-tech investors of last resort. They combine:
Non-dilutive grants (to handle technology risk), and
Equity or quasi-equity (to handle market and scale-up risk).
How they operate:
EIC Accelerator offers up to ~€2.5M grants + €0.5–15M equity tickets via the EIC Fund.
Bpifrance runs dedicated deep-tech grants, repayable advances and equity products under France 2030.
Both are highly selective, with strong signalling effects for winners.
Both are explicitly designed to crowd in private capital (typical leverage 3–5×).
Why they matter:
They standardise the transition from “promising prototype” to “fundable company” for deep tech, where timing and capital needs are structurally different from classic startups.
4. Development banks & public VC: the “capital backbone”
What they do:
Development banks (Bpifrance, KfW, etc.) and public VC vehicles (High-Tech Gründerfonds, Israel Innovation Authority funds) act as long-term capital backbones.
They provide a multi-layer capital stack:
Innovation grants and soft loans,
Seed and growth equity,
Fund-of-funds and LP commitments into specialist deep-tech VC funds,
Guarantees and growth loans.
Examples:
Bpifrance: national development bank and investor; central operator of the Deeptech Plan; directly or indirectly present in a large share of French deep-tech deals.
KfW / HTGF: KfW anchors High-Tech Gründerfonds for early-stage high-tech and runs growth-loan and fund-of-funds schemes.
Israel Innovation Authority: combines incubator grants, direct co-investment and a new programme to support deep-tech VC funds.
Why they matter:
They create permanent institutional capital for deep tech, rather than temporary programs, and they systematically amplify private VC by anchoring funds and co-investing.
5. University & national-lab pipelines: the “deal factory”
What they do:
These are the commercialisation pipelines inside universities and public labs: technology transfer offices (TTOs), proof-of-concept funds, spin-off policies, and venture-building partnerships.
Key features:
Strong TTO with strategic mandate (Oxford, MIT, leading US universities).
Internal proof-of-concept funds to de-risk early technology before a company exists.
Standardised IP & equity templates (clear default equity shares, licensing terms, etc.).
Partnerships with venture builders, incubators, and public investors (e.g. PUIs in France, national lab partnerships, Bpifrance, IIA, HTGF).
Why they matter:
They convert public R&D into repeatable streams of spin-offs, not random accidents. For deep tech, this is crucial, because the highest-value IP is often locked in universities and national labs with heavy infrastructure.
6. Deep-tech incubators & accelerators: the “execution arm”
What they do:
These are specialised, often public or quasi-public incubators/accelerators that provide:
Labs, testbeds, pilot lines, not just coworking desks,
Long-horizon, deep-tech-specific mentoring,
Access to corporates, investors and regulators.
Examples:
Digital Catapult and other UK Catapults – deep-tech innovation centres with advanced facilities, supporting thousands of companies and helping them raise hundreds of millions.
SGInnovate (Singapore) – venture-builder + investor focused on PhD-led deep-tech startups.
EIT Digital / EIT Climate-KIC – EU-wide accelerators with thematic focus (digital, climate), tied into EIT education and innovation.
IIT Madras Incubation Cell – a university-anchored deep-tech incubator with hundreds of startups and billions in aggregate valuation.
Why they matter:
They are the bridge from “research project” to “operational company”: translating prototypes into real products with industrial partners, testbeds and investor-ready narratives.
7. Talent fellowships & entrepreneurial PhD pipelines: the “founder factory”
What they do:
These policies treat deep-tech founder development as a first-class objective. The unit of intervention is the individual researcher rather than the project.
Instruments:
Fellowships and stipends for PhDs/postdocs to explore commercialization,
Structured entrepreneur training (bootcamps, I-Corps-type programmes),
Entrepreneurial doctorate tracks and innovation schools,
Founder-in-residence schemes, venture studios.
Examples:
NSF I-Corps (US) – standardised 7-week customer discovery curriculum; thousands of teams trained, hundreds of startups, billions in follow-on funding.
EIT education programmes – master’s and doctoral schools where entrepreneurship is built into technical training.
SGInnovate’s focus on PhD-led ventures and targeted talent programmes.
Why they matter:
Deep-tech often fails not because the tech is bad, but because founders have no market/regulatory intuition. Talent pipelines encode entrepreneurial skills into scientists themselves, increasing the conversion rate of research to viable companies.
8. National strategies & regulatory frameworks: the “coordination layer”
What they do:
These are top-level deep-tech strategies that signal political priority, allocate budgets, and align instruments across ministries and agencies.
Typical components:
A definition of deep tech and sectors of focus,
Quantitative targets (startups, unicorns, IP, R&D % of GDP),
Pillars covering IP, funding, infrastructure, regulation, skills,
Assigned roles for agencies (development banks, ARPA-style bodies, incubators).
Examples:
France’s Deeptech Plan / France 2030 – explicit numeric targets; Bpifrance as operator; PUIs as regional university-industry hubs.
India’s Draft National Deep Tech Startup Policy (NDTSP) – pillars on R&D, IP, funding (fund of funds, impact bonds), shared infrastructure and regulatory reform; backed by large national RDI fund commitments.
Regional strategies (e.g. Karnataka) mirroring these goals at state level.
Why they matter:
Without this layer, everything below is fragmented. Strategies make sure ARPAs, development banks, incubators, universities and regulators are pushing in the same direction, with compatible rules and incentives.
9. Procurement & sandboxes: the “demand engine”
What they do:
They attack the market side of the valley of death. Instead of only subsidising R&D, governments:
Become lead customers, and
Create regulatory sandboxes to safely test novel models.
Instruments:
Pre-Commercial Procurement (PCP) and similar schemes where governments buy R&D services in stages.
SBIR/STTR programmes that act like mini-procurement for early-stage technologies.
Regulatory sandboxes (FCA in the UK, MAS in Singapore, etc.) where firms test innovations with real users under relaxed rules.
Why they matter:
They create first reference customers for deep-tech startups.
They provide real-world validation (technical, economic, regulatory).
They feed empirical evidence back into regulators, enabling smarter rules for new tech.
In deep-tech domains like energy, health, and finance, this can be more decisive than any grant.
10. Co-investment funds & fund-of-funds: the “amplifier”
What they do:
Here the state leverages its money by anchoring specialist VC funds, instead of doing everything directly.
Mechanics:
Public money as LP in deep-tech funds,
Matching/co-investment in individual rounds,
Target ratios for private capital leverage.
Examples:
EIC Fund – cornerstone investor in deep-tech rounds, systematically crowding in private capital.
Israel Innovation Authority deep-tech VC programme – grants to VC funds focused on advanced technologies to help them reach first close and attract global LPs.
National fund-of-funds structures (France, Germany, Nordics, etc.).
Why they matter:
They build sustainable private deep-tech VC capacity: specialist GPs, networks, pattern recognition. The public sector takes part of the risk, but leaves investment discipline and portfolio construction to professional fund managers.
11. Clusters, testbeds & shared infrastructure: the “physical substrate”
What they do:
These are shared physical and data infrastructures needed to develop and validate deep tech:
labs, fabs, clean rooms,
testbeds (microgrids, 5G/6G networks, autonomous vehicle corridors),
AI/compute facilities and programmable “cloud laboratories”.
Examples:
Catapult Network (UK) – sectoral centres (manufacturing, energy, digital, compound semiconductors, etc.) providing high-end facilities and engineering support.
DOE national labs & AI testbeds (US) – AI hardware testbeds, grid and energy testbeds, fusion, HPC.
NSF AI-programmable cloud labs – national-scale infrastructure for automated, AI-driven science.
Why they matter:
They turn infrastructure from a private capital sink into a shared service, so startups don’t each need their own lab or pilot plant. They also anchor geographic clusters: researchers, startups, corporates and investors naturally concentrate around shared testbeds.
12. How the modules connect: a system view
You can think of a functional deep-tech state as assembling these modules into a pipeline:
Strategy (7) sets missions, priorities and budgets.
ARPA agencies (1) attack frontier technological problems aligned with these missions.
University & lab pipelines (4) and talent programmes (6) convert research and people into early spin-offs and entrepreneurial teams.
Incubators/accelerators (5) and testbeds (10) provide infrastructure and mentoring to turn prototypes into investable companies.
Blended investors (2) and development banks/public VC (3) provide stage-appropriate finance, while co-investment funds (9) expand private VC capacity.
Procurement & sandboxes (8) provide real demand and regulatory learning, enabling scale and system integration.
Each module solves a structural bottleneck:
“We don’t have radical ideas” → ARPAs.
“We have ideas but no spin-off machinery” → university/lab pipelines.
“We have prototypes but no capital/infrastructure” → incubators, testbeds, blended investors, development banks.
“We have tech but no customers or regulatory path” → procurement, sandboxes.
“We have some VC but not enough in deep tech” → co-investment and fund-of-funds.
“We have all these but they’re uncoordinated” → national strategies.
The Kinds
1. ARPA-style mission agencies
(DARPA, ARPA-E, ARIA, ARPA-H, etc.)
1.1 What this model actually is
ARPA-style agencies are small, mission-driven public R&D funders that sit between basic science and commercial markets. They are designed to fund high-risk, high-impact projects that normal agencies and private investors avoid, and they do so via:
a lean, flat organisation,
empowered, temporary program managers, and
flexible contracting that lets them shape and pivot portfolios quickly.
A recent overview of “ARPAs” as a family notes that compared with classic research councils, they operate with lean structures, flexible contracting and empowered program managers who can rapidly launch and pivot programs and actively shape outcomes; this model has produced advances such as GPS, mRNA vaccines and cutting-edge AI/biosecurity systems and is now being copied globally. Emerging Technology Policy Careers
The canonical example is DARPA (US). The “DARPA model” is described in US Congressional analysis as a flat organization with tenure-limited program managers who are given autonomy and risk tolerance, backed by flexible acquisition and hiring authorities. Congress.gov
Newer agencies like ARPA-E (energy), ARPA-H (health) and the UK’s ARIA explicitly adopt the same logic for different domains. techuk.org+3IEA+3PMC+3
1.2 Governance and operating model
Core structural features:
Small and flat
DARPA has a staff in the low hundreds and only a couple of management layers (office directors + director/deputy), allowing very fast decisions. National Academies+1Tenure-limited, empowered program managers (PMs)
PMs are recruited from top industry, academic and lab talent for 3–5 years. National Academies+1
They design entire programs: technical goals, metrics, budget, performers, and transition strategy. books.openbookpublishers.com+1
Their job is not to fund safe incremental work; they are explicitly expected to take big bets, knowing that many will fail.
Flexible contracting & hiring
Congress has given DARPA special acquisition and personnel authorities so it can work with unconventional partners and move money quickly – unlike standard procurement/grant systems. Congress.gov+1High tolerance for failure, measured at portfolio level
ARPA-style agencies are designed on the assumption that a large share of projects will fail, but a few “home runs” justify the entire portfolio. This is explicitly recognised in ARPA-E’s communication (“high-risk, high-reward” projects “too early for private investment”) and in commentary about ARPA-H and ARIA. techuk.org+3OECD+3Tech Brew+3Autonomy and mission focus
ARPA-H is being structured as an entity with its own culture within the US health system to protect its high-risk mission. PMC+1
ARIA is set up as an independent funding body with broad remit and high tolerance for failure, explicitly separate from UKRI’s standard mechanisms. Research Briefings+1
1.3 Instruments and program lifecycle
Despite the mythology, the mechanics are fairly standardised and highly replicable:
Program conception
PM identifies a mission-critical problem (e.g. resilient grid storage, ultra-efficient power electronics, pandemic-scale diagnostics).
They draft a program concept: what breakthrough is needed, what makes it non-incremental, what performance metrics define success, and what time horizon is realistic. books.openbookpublishers.com+1
Call and selection
ARPA-E and others publish notices of funding opportunities (NOFOs) tied to specific initiatives. arpa-e.energy.gov+1
Competitive peer review is used, but PMs keep strong discretion to build a coherent, complementary portfolio rather than just fund top-scoring proposals.
Flexible project management
Transition / “hand-off”
Programs are explicitly designed with a transition path: to other government agencies, private investors, or procurement. darpa.mil+1
For defence, this often means DoD service branches picking up technologies; for energy, ARPA-E emphasises partnerships with utilities, OEMs and investors. NREL+1
1.4 Impact and success cases
DARPA
The ARPA/DARPA model is widely credited with enabling foundational technologies such as the early internet (ARPANET), GPS, stealth aircraft and autonomous systems. Emerging Technology Policy Careers+1
Its success is not any single project but the institutional capability to consistently generate such breakthroughs.
ARPA-E
Since 2009, ARPA-E has provided about $4.07 billion in funding to more than 1,690 energy innovation projects, focusing on early-stage technologies “too early for private-sector investment.” arpa-e.energy.gov+1
In its first years, ~580 project teams receiving $1.5 billion formed 56 new companies and attracted more than $1.8 billion in follow-on private funding. Bipartisan Policy Center+1
As of early 2025, ARPA-E reports 34 exits with total reported value of $22.2 billion, showing that a subset of projects achieve major commercial outcomes. arpa-e.energy.gov+1
ARIA and ARPA-H (early stage)
ARIA has an £800 million budget and is explicitly exempted from standard procurement regulations to allow “high-risk, high-reward, transformational research” with PM flexibility. techuk.org+1
ARPA-H aims to fund aggressive, high-risk health programs that are “not readily accomplished through traditional federal biomedical research”, again embedding ARPA design into health. PMC+1
1.5 Design principles you can extract
If you want to write about principles, the ARPA model can be boiled down into a handful of transferable design rules:
Mission before mechanisms
Start from “what radical capability does the country need?” and let that drive programs, rather than fitting ideas into existing instruments.Empowered, temporary PMs
Tenure-limited, technically strong PMs with a mandate to shape portfolios – not just administer grants – are the institutional engine. OpenEdition Books+1Flat, autonomous agency
Keep the organisation small, with minimal hierarchy, broad freedom in contracting, and insulation from short-term political swings. Congress.gov+1Portfolio thinking and failure tolerance
Accept that 60–80 % of projects may fail; judge success by the aggregate impact of the few that work (ARPA-E’s exits and follow-on capital are the proof-point). arpa-e.energy.gov+2Bipartisan Policy Center+2Designed transitions
Every program must have a theory of change for how technologies will leave the lab: downstream procurement, regulatory changes, or investors already at the table. darpa.mil+1
2. Blended grant–equity “deep-tech investor of last resort”
(EIC, Bpifrance Deeptech Plan, etc.)
2.1 The problem this model solves
Deep-tech ventures typically need more capital and more time than digital SaaS or consumer startups. A 2023 deep-tech report estimates they take 35 % more time and 48 % more capital to reach modest revenue levels than traditional startups, which makes classical VC less comfortable. Dealroom.co
If you only provide grants, you often lose leverage and alignment at scale-up stages. If you only rely on VC, entire domains (quantum, novel materials, climate hardware) may remain underfunded. The blended grant–equity model is an institutional response:
Public body offers non-dilutive grant for proof-of-concept / validation.
Same body (through a fund) then provides equity for market entry and scaling.
The fund is structured to crowd in private co-investors, rather than crowd them out.
2.2 European Innovation Council (EIC) as archetype
Instrument structure
The EIC Accelerator under Horizon Europe supports high-risk, high-impact SMEs and startups with: European Innovation Council+1
Grants up to €2.5 million (non-dilutive) for technology development (typically TRL 5–8).
Equity or quasi-equity from about €0.5–15 million through the EIC Fund to finance market deployment and scale-up.
A “blended finance” option that combines both, used by the majority of selected deep-tech companies. European Innovation Council+2NCP Brussels+2
The EIC Fund acts as a public VC, taking minority stakes, often with other investors alongside. It aims explicitly to generate 3–5× private follow-on investment per euro invested. AECM+1
Scale and impact
According to the 2023 EIC Impact Report and subsequent summaries:
In 2023, the EIC Fund completed 100+ investments in deep-tech companies, totalling around €1.2 billion. European Innovation Council+2IP Helpdesk+2
Each euro invested leveraged over €3.5 of additional private investment. European Innovation Council+2APRE+2
Across several years, the EIC now manages a portfolio of deep-tech companies whose aggregate valuation is about €70 billion, positioning it as one of Europe’s most active deep-tech investors. European Innovation Council+2ACRID Network+2
Operational logic
Stringent selection & due diligence
Multi-stage process (short application, full proposal, pitch to a jury); very low success rate, which creates a strong signalling effect for winners. adrforum.eu+1
Separation but coordination of grant & equity
Technical evaluation is done under the Horizon Europe rules; investment decisions follow VC-style due diligence via the EIC Fund’s investment committee.
Patient capital
Equity is designed as patient, with longer holding periods than typical VC, recognising deep-tech time horizons. AECM
Leverage and crowding-in
Structuring EIC Fund tickets in a way that invites co-investment (often taking a minority stake, leaving room for private lead investors) maximises leverage. sciencebusiness.net+1
2.3 Bpifrance and the French Deeptech Plan
France gives you a slightly different version of the same archetype, wrapped in a national strategy.
Strategic objective
The “Deeptech Plan”, launched in 2019 and integrated into France 2030, sets clear numeric goals by 2030:
500 new deep-tech startups per year.
About 10 deep-tech unicorns and 100 industrial sites annually to host them as they scale. DataScientest+3STIP Compass+3French Expert in Ireland+3
The plan recognises that France was producing strong scientific results but too few industrial innovations and spin-offs, so the focus is on turning scientific discoveries into companies. EE Times+1
Financial instruments
Bpifrance, the national development bank, is the main operator. Its deep-tech toolbox includes: DataScientest+2Bpifrance.com+2
Grants and subsidies for very early-stage (“Deeptech Emergence”, “French Tech Lab Grant”) – used to finance feasibility, prototyping and pre-company work.
Repayable advances / soft loans for validation stages.
Direct equity via seed and growth funds (often co-investing with private VC) in labelled Deeptech startups.
Fund-of-funds and co-investment in venture funds specialising in deep tech.
Between launch and 2021, the plan already raised the number of new deep-tech startups to around 250 per year, a 26 % increase vs 2020, even though it was still short of the 500/year target – indicating strong acceleration but also how ambitious the goal is. Bpifrance.com+1
Non-financial components
Deeptech label: a national label signalling that a company is based on breakthrough tech and qualifies for specific instruments, recognised throughout the ecosystem. DataScientest
Pôles Universitaires d’Innovation (PUIs): regional university–industry hubs co-financed by Bpifrance to strengthen spin-off pipelines, including proof-of-concept support before company creation. Bpifrance.com+1
2.4 Cross-cutting principles of the blended model
From EIC and Bpifrance you can extract a fairly crisp “design pattern”:
Stage-appropriate money
Grants for technology risk (get TRL up, generate IP, validate basic performance).
Equity for market risk (building a sales force, industrialisation, regulatory approvals).
Each stage has different risk and information profiles; instruments are tuned accordingly.
Single institutional interface, dual instruments
Founders don’t have to navigate a completely different universe for grants vs equity: the EIC Accelerator + EIC Fund bundle them; Bpifrance acts as both grantor and investor. European Innovation Council+2AECM+2
Public investor as anchoring capital, not crowding-out capital
Ticket sizes and terms are structured to invite private co-investment (minority stakes, market-based pricing). EIC’s leverage of >3.5× private capital per euro invested is the cleanest quantitative proof. European Innovation Council+2sciencebusiness.net+2
Link to national / continental missions
EIC’s portfolio is explicitly aligned with EU missions and strategic priorities (green, digital, health). adrforum.eu+1
France’s Deeptech plan is embedded in France 2030’s broader reindustrialisation/decarbonisation goals. STIP Compass+2ActuIA+2
Selective but highly visible
Both schemes are extremely competitive; being selected sends a strong signal to markets and talent. This “certification effect” is an important part of the value.
Standardised yet flexible deal templates
EIC Accelerator has clear ranges for grants and equity; Bpifrance has named instruments with known parameters. This standardisation makes it easier for founders and co-investors to understand what is on offer, while still allowing case-by-case structuring. European Innovation Council+2AECM+2
3. National development banks & public VC as deep-tech backbones
(Bpifrance, KfW/HTGF, Israel Innovation Authority, etc.)
3.1 What this model is
Here the state doesn’t just fund R&D; it builds a capital stack for deep tech by:
running a development bank (or similar) that lends, guarantees and co-invests, and
backing public or public–private VC funds that specialise in high-risk tech.
So instead of one-off programs, you get a permanent capital institution whose mandate includes deep tech.
3.2 Bpifrance: the “all-in-one” French backbone
Position in the system
Bpifrance is France’s national development bank and investment institution, with eight business lines: financing, guarantees, innovation financing, direct investments, fund-of-funds, export, etc. In 2024 it injected about €60 billion into the economy across loans, guarantees and investments. It’s both a bank and one of Europe’s largest LPs/GPs.
Deeptech Plan integration
The French Deeptech Plan (2019) sits inside the broader France 2030 strategy and is operated by Bpifrance.
Launch budget ≈ €2.5 billion for deep-tech instruments (grants + equity + funds).
By 2021:
Deep-tech startups raised €2.3 billion, +91 % vs 2020.
€375 million invested directly by Bpifrance into deep-tech startups.
€401 million into deep-tech investment funds.
Bpifrance was directly or indirectly involved in 70 % of deep-tech fundraising rounds.
Instruments Bpifrance runs for deep tech
Innovation grants & repayable advances – very early TRL work, prototyping, feasibility.
Direct equity – seed and growth tickets via Bpifrance’s own venture and growth funds, often leading or co-leading rounds.
Fund-of-funds – LP commitments into specialist funds (e.g. climate, biotech, industry 4.0).
Sector partnerships – e.g. INRAE/Bpifrance agreement to develop agri-deeptech startups under the Deeptech Plan, explicitly aimed at turning research into “industrial champions”.
What this does structurally
For France, this creates a single financial spine that:
touches startups directly (grants, equity),
amplifies private VC capacity (fund-of-funds, co-investment), and
is explicitly calibrated to deep-tech timescales and capital needs.
3.3 Germany: KfW + High-Tech Gründerfonds
High-Tech Gründerfonds (HTGF)
Public–private seed investor founded 2005 to close the early-stage gap in high-tech.
Focuses on deep tech, industrial tech, climate, digital, life sciences, chemistry.
Seed “sweet spot”: €800k+ initial tickets, with capacity to invest up to €30 million per startup over its lifetime.
Fund volume now >€2 billion across fund generations.
Role of KfW and the state
HTGF investors include the German Federal Government and KfW Banking Group (state-owned development bank), plus dozens of industrial corporates.
KfW has repeatedly invested in HTGF; a recent round indicates a planned fund volume of up to €300 million, with KfW as second-largest investor and an increasing share of private LPs.
Separately, KfW runs Venture Tech Growth Financing (VTGF) to provide growth loans to high-growth tech firms, funded by the German “Zukunftsfonds” to address late-stage financing gaps.
Net effect
Germany uses:
HTGF as a high-volume, early-stage, tech-specialist seed investor, and
KfW as a growth-stage supporter (growth loans, fund-of-funds via KfW Capital).
This covers both seed gap and scale-up gap in deep tech.
3.4 Israel: Innovation Authority + incubators + fund-of-funds
The Israel Innovation Authority (IIA) is the main public innovation financier in Israel. For deep tech, its role is:
direct early-stage grants,
technological incubators that turn academic IP into ventures, and
new fund-of-funds and venture incubator schemes.
Deep-tech incubators
In 2024–25 the IIA launched a tender for three new deep-tech venture incubators, each eligible for up to NIS 40 million (~US$10 M) in grants over 5 years for operating and lab costs.
Incubators act as “innovative investment entities”, bringing local and international investors together and providing shared labs for startups in health, bio-convergence, climate, agri-tech, food-tech, etc..
Deep tech funds & fund-of-funds
A new Deep-Tech Startups Fund co-invests with private investors in very early-stage deep-tech startups.
In 2025 IIA announced a US$70 M fund-of-funds programme to support VC funds specialising in deep tech, with grants up to $10 M per fund, echoing Israel’s historic Yozma programme.
System logic
As one analysis puts it, “through the Israel Innovation Authority, the state lays the groundwork for deep tech development by funding early-stage startups, establishing technology incubators that turn academic research into commercially viable ventures, and helping companies expand trade internationally”.
So the pattern is:
grants + incubators for raw IP and founding teams,
co-investment funds for follow-on capital,
all under one umbrella agency (IIA) with mandate to be catalytic, not crowding-out.
3.5 Principles from the “development bank + public VC” archetype
If you want to abstract this:
Permanent institutional capital
– Deep tech is not a 3-year programme problem; it’s a 30-year capability problem. A development bank / public VC is a standing institution with rolling funds and multiple generations of capital.Multi-layer capital stack
– Same institution (or tightly coupled set) provides:grants / soft loans (innovation financing),
direct equity (seed/growth funds),
fund-of-funds (LP in specialist VC),
guarantees / growth loans.
Bpifrance and KfW are almost textbook examples.
Crowding-in design
– Structures intentionally designed so every public euro leverages several private euros (co-investment, LP positions, matching schemes). Bpifrance being involved in 70 % of French deep-tech fundraisings is exactly that role.Integration with strategy
– Deep-tech is not generic “SME support”. Funding mandates are tied to national missions (climate, health, sovereignty, reindustrialisation).High dealflow + professional investment discipline
– HTGF, Bpifrance and IIA all run like professional investors (VC processes, due diligence, portfolio monitoring), but with public goals and longer horizons.
4. University & national-lab commercialisation pipelines
(Oxford, MIT, national labs, PUIs, etc.)
4.1 What this model is
Here the focus is the conversion of public research into companies: repeated, industrialised spin-off creation rather than random, one-off success stories.
Key elements:
strong technology transfer office (TTO),
internal proof-of-concept (PoC) and seed funds,
standardised IP and equity terms for spinouts,
venture-building capacity (in or around the university).
4.2 Oxford as a high-throughput spin-out machine
Oxford University Innovation (OUI) – the TTO
OUI is a UK leader in patent filings and spin-outs.
It facilitates creation of ~15 spinout companies per year on average; since 2015 the annual number of spinouts has increased by 166 %, and capital raised by them has increased by 687 %.
Between Aug 2023 and July 2024 alone, Oxford spinouts raised £872.1 million.
Proof-of-concept and seed funds
Oxford has long operated internal PoC and seed funds (UCSF & OIF) with guidelines for researchers on how to access them. Basic pattern:
small PoC grants to demonstrate feasibility, reduce technical risk;
follow-on seed for IP protection, early team formation, initial commercial validation.
OUI itself welcomes policies that increase PoC and scale-up capital, noting that these bottlenecks are critical for unlocking more investment and faster spinout growth.
Standardisation & policy influence
The UK government’s 2023 Independent Review of University Spin-outs benchmarks Oxford and peers and finds that transparent, standardised deal terms (e.g. standard equity ranges for university share at formation) and streamlined processes can significantly accelerate spin-outs and attract investors. OUI broadly supports the recommendations to simplify and speed up spin-out creation.
4.3 MIT and the “entrepreneurial university” template
MIT’s spin-out ecosystem has been studied to death because it combines:
a strong TTO,
dense entrepreneurial culture,
student and alumni-led activity (clubs, accelerators), and
deep ties to VC and industry.
A classic paper on MIT and similar institutions outlines five structural models for spinning off new companies from universities and government labs (internal ventures, licensing, external venture capital partnerships, etc.). Later work on university entrepreneurial ecosystems (including the MIT–Skoltech report) highlights:
strong leadership and supportive culture,
student-led communities and events,
physical hubs (incubators, “venture garages”),
intense external relationships with VC, corporates and alumni.
MIT’s tech transfer isn’t just about contracts; it’s an ecosystem logic: TTO + culture + capital + community.
4.4 Best-practice building blocks (from global reviews)
Recent research on university technology transfer and the UK spin-out review converge on a set of best practices:
Strategic TTO positioning and senior buy-in
Effectiveness is strongly influenced by strategic choices of university leadership (how much autonomy & budget TTO gets, how risk-tolerant IP policy is).
Successful universities treat commercialisation as a core mission, not a side-office.
Proof-of-concept & translational funds
Internal PoC funds (Oxford, many US universities) provide small, rapid grants to validate ideas and generate IP before any spin-out decision.
This is crucial in deep tech, where technical risk is often higher than business risk at early stages.
Standardised IP & equity frameworks
The UK spin-out review shows that when universities have clear “standard deals” (e.g. default equity ranges, IP license terms), negotiations are faster, uncertainty is lower, and founders/investors are less anxious.
OUI notes that many leading UK universities’ deal terms are now broadly comparable to leading US universities when dilution is normalised.
Integration with external venture builders and funds
Partnerships with venture builders or deep-tech accelerators provide external entrepreneurial capacity: EIRs, CEOs-in-residence, and specialist operators.
France’s PUIs and partnerships like INRAE–Bpifrance are examples where the research side and the development bank co-design spin-out pathways.
Metrics and feedback loops
Mature ecosystems track: number of spin-outs, capital raised, jobs created, time-to-incorporation, IP income vs equity value, etc.
Oxford’s reporting of 872.1 M GBP raised in a single year is exactly this kind of transparent metric.
4.5 National labs and shared public research infrastructure
Beyond universities, national labs (US DOE labs, Fraunhofer, etc.) are critical deep-tech sources:
They control big science infrastructure (synchrotrons, HPC, pilot lines) that startups cannot replicate.
Many run:
internal PoC programs,
lab–industry partnership programs,
embedded incubators or collaborations with external incubators.
This is structurally the same archetype as universities, but often with more applied, mission-driven research and closer alignment with state missions (energy, defence, climate).
4.6 Principles you can pull out from the university/national-lab archetype
Treat commercialisation as a core function of the university/lab, with leadership backing and strategic KPIs.
Give the TTO autonomy, budget and talent; make it a proactive deal-maker, not a passive IP clerk.
Deploy PoC funds as a standard step between research grant and spin-out.
Standardise IP/equity terms to reduce negotiation friction and signalling risk.
Embed entrepreneurial culture (student clubs, founder communities, alumni VC ties) around the TTO so dealflow is constant.
Connect to external capital backbones (development banks, Bpifrance-style, IIA, HTGF) so spin-outs have an obvious funding ladder.
5. Deep-tech incubators & accelerators as public infrastructure
(Digital Catapult, SGInnovate, EIT accelerators, IIT Madras, etc.)
5.1 What this model is
This archetype treats incubators and accelerators as a piece of national infrastructure, not just “startup programs”:
They are specialised in deep tech, not generic SaaS.
They provide infrastructure (labs, testbeds, pilot lines), not just mentoring.
They are deeply connected to public money and missions (climate, AI, health, industry 4.0).
In other words, they are execution arms that turn upstream capital (ARPA, development banks, university IP) into investment-ready companies.
5.2 Digital Catapult (UK): national deep-tech accelerator hub
Digital Catapult is part of the UK’s Catapult Network, which itself is designed as a set of sectoral innovation centres. Digital Catapult’s remit is deep tech: AI/ML, immersive, quantum, advanced networks, IoT.
Scale and impact
Recent figures from its impact reporting:
Since 2018, Catapult-supported startups have raised ~£550–575 million in investment.
The organisation has supported around 3,000 companies since 2018.
It operates 20+ advanced technology facilities across the UK.
External assessments (e.g. Scaleup Institute, parliamentary evidence) highlight that Digital Catapult has helped startups and scaleups in AI, XR, quantum and advanced networks raise >£550M since 2018.
Operating model
Sector focus: calls and programmes are themed (e.g. AI, quantum, 5G/6G testbeds).
Facilities: access to experimental labs, networks and test environments (e.g. 5G testbeds, immersive labs).
Programmes: multi-month cohorts with structured mentoring, access to corporate partners, and investor showcases.
Positioning: they explicitly describe themselves as a “deep tech innovation organisation driving business value”.
Principles you can extract
National but distributed: physical facilities across multiple regions, aligned with regional development.
Metrics on downstream capital (investment raised) and industrial adoption, not just number of workshops.
Strong corporate and public partner network so pilots and procurement can follow quickly.
5.3 SGInnovate (Singapore): the “PhD-holders’ startup club”
Singapore’s SGInnovate is a government-backed company with a very clear thesis: deep tech is primarily a scientist-founded game.
In their own positioning:
They describe deep tech as a “PhD holders’ startup club”, where scientists and engineers with advanced degrees tackle big problems (cancer, climate, congestion), in sharp contrast to “general tech” apps.
Deep tech is framed as lab-originated, long-horizon, capital-intensive – hence the need for specialised support.
What SGInnovate actually does
Venture building: works with universities, research institutes and hospitals to build spin-offs from IP (particularly in AI, med-tech, agrifood, quantum).
Investment: makes early-stage investments into deep-tech startups (often co-investing with private VCs).
Talent programmes: fellowships and apprenticeship-type programmes placing PhDs and engineers into startups.
Community: deep-tech events, technical meetups, and thought leadership, reinforcing the “deep-tech club” identity.
Principles
Talent-anchored: the unit of analysis is the PhD / research team, not the business plan.
Hybrid role: both venture builder and investor, with public mandate.
Dense integration with research system: direct pipelines from national labs and universities into SGInnovate’s portfolio.
5.4 EIT accelerators (EU): scale-up platforms for climate & digital deep tech
Under the EU’s EIT (European Institute of Innovation and Technology), several “KICs” (Knowledge and Innovation Communities) run deep-tech-intensive accelerators:
EIT Digital
Its accelerator has supported 450+ companies; those companies have raised over €1.5 billion in private investment, with more than €100 million directly facilitated by the accelerator.
EIT Climate-KIC
Climate-KIC runs what it calls “the world’s most extensive climate tech accelerator”:
Supported >1,800 climate-positive businesses.
Those businesses have raised >€1.5 billion in follow-on investment and created >10,000 jobs.
These are essentially thematic European deep-tech accelerators, embedded in a larger ecosystem of EIT education, innovation projects and EU funding.
Principles
Pan-European selection and networks – they source teams from many countries and connect them across borders.
Theme-driven (climate, digital) with strong alignment to EU missions.
Integrated with education – master/PhD programmes feed into accelerator pipelines.
5.5 IIT Madras Incubation Cell (India): deep-tech at university scale
A very recent example from the Global South is IIT Madras Incubation Cell:
As of 2025, it has incubated over 500 deep-tech startups in about 12 years.
Those startups:
have a combined valuation of about ₹53,000 crore (~US$6–7B),
have raised ₹17,310 crore (~US$2B) in venture funding, and
filed 700+ patents.
In 2023–24 alone, 190 startups generated ₹4,000 crore in revenue.
This is basically a university-anchored deep-tech incubator operating at national scale.
Principles
Long-term continuity: 12-year track record; not a short pilot.
Domain diversity: multiple deep-tech verticals under one umbrella (AI, hardware, clean tech, etc.).
Strong IP and patenting support, reflected in patent counts.
5.6 Cross-cutting design rules for deep-tech accelerators/incubators
From these cases, your “principles” list for this archetype can be:
Deep specialisation
– Focus on deep-tech verticals (AI, quantum, climate tech, med-tech, semiconductors) with mentors and infrastructure matched to those domains.Infrastructure as a service
– Provide labs, pilot lines, testbeds and regulatory sandboxes – not just coworking desks.Tight connection to upstream institutions
– Direct dealflow from universities, national labs, ARPA programmes, development banks.National or regional mandate with clear metrics
– Companies supported, follow-on capital, jobs, CO₂ reduction, etc. (Digital Catapult, EIT and IIT Madras all publish these numbers).Blend of grant/fee and equity models
– Some programs take small equity stakes; some rely on public grants and corporate sponsorships; many combine the two.
6. Talent fellowships, stipends & entrepreneurial PhD pipelines
(NSF I-Corps, EIT education, SGInnovate, etc.)
6.1 What this model is
This archetype treats deep-tech founder development as a policy object:
Target population = PhD students, postdocs, early-career researchers.
Instruments = stipends, fellowships, cohorts, bootcamps.
Goal = produce people who can move between lab and market, i.e., scientists who understand customers, regulation and capital.
This is where public money directly funds the human capital pipeline, not just projects or companies.
6.2 NSF I-Corps (US): industrial-strength entrepreneurial training for scientists
The NSF Innovation Corps (I-Corps) is the clearest, well-documented example.
Impact facts
Since 2012, more than 2,500 teams have participated.
Nearly 1,400 of those teams have launched startups.
Those startups have raised $3.16 billion in subsequent funding and created >11,000 jobs.
This makes I-Corps one of the world’s largest and most successful entrepreneurial training programmes for scientists.
Operating model
Target group: researchers with NSF-funded projects (and more recently, broader STEM teams).
Programme structure: typically a 7-week curriculum where each team:
conducts ~100 customer interviews,
iterates their value proposition and business model,
learns to distinguish technology risk from market risk.
Mentor role: each team has an entrepreneurial lead, technical lead and an industry mentor.
NSF explicitly frames I-Corps as a tool to train an entrepreneurial workforce and accelerate the translation of basic research into practical applications.
Why it matters for deep tech
Many deep-tech failures are not technical but market/fit/regulation failures.
I-Corps is essentially a standardised de-risking protocol for the people running deep-tech projects, before they ever raise VC.
6.3 EIT & Climate-KIC education: embedding entrepreneurship into advanced training
The EIT KICs combine accelerators with formal education programmes:
EIT Climate-KIC reports having strengthened the climate-leadership potential of more than 44,000 participants via its education programmes, with a strong STEM and diversity focus.
EIT Digital runs Master, Doctoral and Summer Schools designed to combine technical training with innovation and entrepreneurship modules.
This is not always branded as “entrepreneurial PhD”, but effectively:
PhD and master’s students are exposed to startup cases, challenge-based learning and innovation projects.
Many then feed into EIT accelerators (Digital, Climate, Health, etc.) as founders or early employees.
6.4 Singapore & the “PhD-led founder” narrative
Returning to SGInnovate:
Their public narrative explicitly positions deep tech as led by PhD-holders and advanced-degree scientists.
They support that narrative with:
Fellowships and talent programmes that place PhDs into startups or into SGInnovate-backed companies.
Community and mentoring activities where senior scientists and entrepreneurs advise younger researchers.
More broadly, documents on deep-tech investments in Singapore describe the ecosystem of specialised accelerators, early-stage investors and government capability-building programmes (e.g. Clearbridge Accelerator, Zircom MedTech, cleantech accelerators) as part of a coordinated attempt to commercialise research-intensive ventures.
This is less of a single programme (like I-Corps) and more of a cultural and institutional bet: if you build enough support structures around PhDs, a meaningful fraction of them will become founders.
6.5 Typical instruments in a “talent pipeline” policy
Across these examples, you can see a menu of instruments:
Fellowships & stipends
Fund PhD or postdoc time specifically earmarked for exploration of commercialisation.
Sometimes combined with secondments in industry or venture studios.
Bootcamps / cohorts
I-Corps style: short, intense programmes with mandatory customer discovery, interviews and pitches.
Often run in regional “nodes” and then scaled nationally.
Entrepreneurial doctorates / dual tracks
Formal PhD tracks that integrate business modules, internships in startups, or venture creation as part of the thesis work (EIT Doctoral Schools are close to this model).
Founders-in-residence & venture studios
Venture studios (public or public-private) match experienced entrepreneurs with lab teams, with stipends covering early work.
This is happening implicitly around SGInnovate and in some EU deep-tech venture builders.
Micro-grants for student projects
Small, fast grants for student teams to test ideas (hackathons, early prototypes); not all are deep-tech, but they seed entrepreneurial culture.
6.6 Design principles for a deep-tech talent pipeline
If you want to turn this into principles:
Start inside the lab, not after
– Support exploration while people are still PhDs/postdocs (stipends, PoC, I-Corps-type programmes).Standardise entrepreneurial training
– Use repeatable curricula (customer discovery, regulatory mapping, business basics). I-Corps is effectively a protocol that can be ported to other countries.Make founder transitions low-risk
– Offer fellowships, options to return to academia, or shared positions so that leaving the “safe path” is not all-or-nothing.Integrate with accelerators and capital
– Graduates from talent programmes should have clear next steps: EIT accelerators, SGInnovate venture building, university incubators.Track outcomes at the person level
– Not just startups formed, but:careers of alumni,
number who later become repeat founders, investors or CTOs,
their contribution to broader deep-tech ecosystems.
7. National deep-tech strategies & regulatory frameworks
(France 2030 / Deeptech Plan, India NDTSP, regional strategies like Karnataka, etc.)
7.1 What this model is
Here the “instrument” is not a specific fund or program, but a top-level strategy that:
defines what counts as deep tech,
sets quantitative targets (startups, unicorns, jobs, IP),
aligns multiple ministries and agencies, and
hard-codes enabling conditions (IP, regulation, infrastructure, skills).
It is essentially the coordination layer for all other archetypes.
7.2 France’s Deeptech Plan within France 2030
France is one of the clearest examples of a national-level deep-tech strategy.
Targets and framing
The “Deeptech Plan” under France 2030 aims to create 500 new deep-tech startups per year and around 10 deep-tech unicorns by 2030. STIP Compass+2French Expert in Ireland+2
It is explicitly framed as a corrective: France produces excellent scientific output, but too few research-based startups and industrial champions. EE Times
Policy axes
OECD and French summaries describe three main axes: STIP Compass+2French Expert in Ireland+2
Accelerate formation of startups – via Bpifrance instruments, PoC funds, university innovation hubs (PUIs).
Provide tailored support – grants, loans, equity, mentoring, and the Deeptech label as a system-wide quality mark.
Structure territories – fund regional innovation hubs (PUIs) integrating universities, labs and firms.
Why this matters
This strategy effectively:
gives Bpifrance and other agencies a clear numerical and thematic mandate;
ensures that higher-level budgets (France 2030) explicitly allocate money to deep tech;
sets a signal to universities, corporates and investors that deep tech is a priority, not a side topic.
7.3 India’s Draft National Deep Tech Startup Policy (NDTSP)
India’s NDTSP 2023 is another textbook example of a comprehensive deep-tech strategy.
Overall aim
The policy explicitly defines deep tech as research-driven innovation with potential to solve India’s biggest societal challenges, and aims to stimulate innovation, economic growth and societal development through deep-tech startups. psa.gov.in+1
Key pillars (from the official draft)
The draft and official press note group actions into several pillars: psa.gov.in+2Khaitan & Co+2
Nurturing R&D & innovation – boosting domestic R&D, strengthening linkages between labs and startups.
Strengthening IP regime – in-house capabilities for patent landscaping & FTO analysis, bolstering global IP protection, updating IP rules for frontier tech. Khaitan & Co
Facilitating access to funding – a thematically focused Fund of Funds, better coordination of grants across ministries, Technology Impact Bonds for broader investment. psa.gov.in
Enabling shared infrastructure – shared testbeds, labs and pilot plants. Press Information Bureau
Regulation & standards – regulatory reforms and sandboxes in key sectors.
In parallel, India is now backing this with large national RDI funds, e.g. a ₹1 lakh crore Research, Development, and Innovation (RDI) Fund aimed partly at deep tech (quantum, AI, robotics, drones), intended to crowd in private/VC capital and raise deep-tech share of funding from ~10 % to 30–50 % in five years. The Times of India
7.4 Regional deep-tech strategies (example: Karnataka)
Sub-national regions also adopt deep-tech strategies. Example:
Indian state Karnataka (Bengaluru ecosystem) has a Startup Policy 2025–2030 with a ₹518 crore budget targeting 25,000 new startups, explicitly emphasising emerging deep-tech domains like AI, blockchain and quantum. It includes pillars on funding, incubation, mentorship, market access, international outreach and regulatory support, with a focus on building clusters beyond Bengaluru. The Times of India
This demonstrates how national deep-tech strategies can be mirrored at state or regional level, with their own budgets and cluster logic.
7.5 Principles from the “national strategy” archetype
Explicit deep-tech definition and scope – to avoid dilution into generic “tech”.
Numeric targets (startups/year, unicorns, R&D % of GDP) and time horizon.
Pillar logic – IP, funding, infrastructure, regulation, talent; each with concrete actions. Press Information Bureau+1
Agency mandates aligned with strategy – Bpifrance, IIA, development banks receive clear marching orders. STIP Compass+1
Vertical + horizontal – sectoral missions (climate, semiconductors) sit atop horizontal enablers (regulation, IP, procurement).
Regular updates and independent reviews – policy refreshed and stress-tested (France 2030 refreshes, India’s evolving implementation, etc.).
8. Strategic public procurement & regulatory sandboxes
(EU pre-commercial procurement, US SBIR, UK FCA sandbox, MAS sandbox, GovTech, etc.)
8.1 What this model is
Here the state supports deep tech by being:
a first demanding customer (innovation-friendly procurement), and
a risk-managing regulator (sandboxes).
This is about demand-side policy: pulling innovation into markets rather than only pushing money into labs.
8.2 Innovation-friendly procurement: EU, US SBIR, GovTech
EU Pre-Commercial Procurement (PCP)
The European Commission’s PCP instrument lets public procurers buy R&D services in stages (design–prototype–test) without committing to full-scale deployment: Research and innovation+1
It targets breakthrough innovative solutions for big challenges (health, security, clean energy, climate).
PCP is explicitly designed to:
give first customer references to innovative companies,
allow risk & benefit sharing in prototype development, and
make it easier for startups/SMEs to enter public procurement. Research and innovation
US SBIR/STTR
The US Small Business Innovation Research (SBIR) and STTR programs are classic examples of procurement-like grant funding:
They provide early-stage, non-dilutive funding to small businesses to develop and commercialise technology, coordinated by the Small Business Administration. sbir.gov
Agencies such as DoD, DOE, DHS and DARPA run their own topic calls. DARPA structures its SBIR/STTR efforts in three phases (feasibility, R&D prototype, commercialisation) with escalating funding. darpa.mil+1
Evidence summarised in EU/OECD reports highlights strong impacts on growth and strong signalling effects to venture capitalists, although employment effects are more modest. ricg.org+1
SBIR is, in practice, an innovation procurement pipeline: agencies use it to source solutions to mission needs, while startups get capital + reference customers.
GovTech innovation practices
Recent OECD work on digital innovation in government emphasises experimental procurement: early market engagement, pilots, outcome-based contracts and agile procurement to enable digital GovTech solutions. OECD+1
Countries like Singapore (GovTech) are pointed to as examples using common platforms and stacks plus small expert teams and agile procurement to support innovation across agencies. KordaMentha
8.3 Regulatory sandboxes: UK FCA, MAS, global diffusion
UK – FCA sandbox
The FCA regulatory sandbox launched in 2016 is widely recognised as the first modern regulatory sandbox. MDPI+1
It allows firms to test innovative products and business models in a live market environment under relaxed rules and supervision, with safeguards. FCA
It became the template for many later sandboxes worldwide and is often cited as a benchmark for regulatory innovation. MDPI+1
Singapore – MAS FinTech Regulatory Sandbox
The Monetary Authority of Singapore’s FinTech Regulatory Sandbox lets fintech players test innovative financial products and services using real customers and money under caps and guardrails (customer types, transaction limits, etc.), while certain regulatory requirements are temporarily relaxed. mas.gov.sg+2edb.gov.sg+2
It is explicitly framed as a way to accelerate time-to-market while maintaining safety.
Global sandbox landscape
A recent study of 199 sandboxes across 92 countries notes that sandboxes have become a widely used tool to promote innovation in finance, energy, telecom and health, with the US, Singapore and the UK accounting for a quarter of all sandboxes. SSRN
Another OECD report on AI sandboxes confirms that the FCA’s 2015–16 fintech sandbox triggered a global diffusion of sandbox models, now being extended to AI and data innovation. OECD+1
8.4 Principles from the procurement/sandbox archetype
State as lead customer – use PCP, SBIR-type grants, and pilot contracts to buy solutions to real problems and thereby validate deep tech. Research and innovation+2sbir.gov+2
Risk-sharing contracts – staged contracts that share R&D risk and don’t require startups to fully comply with all legacy requirements from day one.
Sandboxing “edge cases” – where regulation is a barrier (fintech, health, AI), create supervised environments where deep-tech firms can play with real data and clients under guardrails. mas.gov.sg+2MDPI+2
Strong signalling effect – being selected into a sandbox or winning SBIR/PCP contracts is a powerful quality signal to investors and partners. ricg.org+2sbir.gov+2
Feedback into regulation – sandboxes are not just for firms; they generate empirical input that regulators use to update rules.
9. Co-investment vehicles & fund-of-funds
(EIC Fund, Israel IIA deep-tech VC programme, etc.)
9.1 What this model is
Here the state doesn’t necessarily invest directly into startups, but instead:
anchors specialist funds as a limited partner (LP), and/or
co-invests alongside private investors into specific deals.
Goal: expand the amount of private capital available for deep tech and shape its direction, without nationalising the entire VC stack.
9.2 EIC Fund as leveraged co-investor
We already used EIC as a blended grant-equity archetype. In the co-investment lens, the key feature is leverage:
The EIC Fund, the investment arm of the European Innovation Council, has invested over €1 billion into startups and SMEs selected via EIC Accelerator, and these investments have crowded in over €2.6 billion of additional private and strategic capital (≈3:1 leverage). APRE+1
The EIC Impact Report 2023 and follow-ups report that:
In 2023, the EIC Fund made 100+ investments totalling €1.2 billion,
leveraging over €3.5 of private investment for every euro invested. sciencebusiness.net+2UNDP+2
The total value of the EIC-supported portfolio is almost €70 billion, up €20B in two years. European Innovation Council
Structurally, the Fund acts as a public cornerstone investor in rounds, with private VCs taking lead roles, but the public money de-risks deals and enables bigger tickets.
9.3 Israel Innovation Authority: “Yozma-style” deep-tech funds
In 2025, the Israel Innovation Authority (IIA) launched a new programme to support deep-tech venture capital funds:
A ₪250 million (~US$70M) fund that provides grants to deep-tech-focused funds, helping them reach first close, expand capital base and start investing in advanced deep-tech technologies (semiconductors, energy, climate, quantum, health, etc.). רשות החדשנות+2ctech+2
Grants of up to $10M per fund are available, with attractive funding terms and risk-sharing. The IIA’s participation acts as a “quality seal” that facilitates additional investor engagement and international collaboration. רשות החדשנות+1
This is explicitly described as a new incentive programme for deep-tech venture capital funds, echoing the historical Yozma fund-of-funds model that kick-started Israeli VC.
9.4 General pattern
If you abstract these:
Fund-of-funds role – public capital as LP in specialist deep-tech funds (EIC, Bpifrance, KfW Capital, IIA Yoazma-Deep Tech). UNDP+2APRE+2
Co-investment into rounds – public fund joins syndicates, often with anti-crowding design (minority stakes, market pricing).
Leverage targets – explicit goals for private capital multiples (3–5×). EIC hits ~3.5×; IIA’s programme is structured to help funds get to first close and attract global LPs. sciencebusiness.net+2swisscore.org+2
Selection of specialist GPs – public capital chooses managers with deep domain competence rather than generalist funds.
Embedded strategic themes – climate, semiconductors, resilience, defence, etc., encoded in investment mandates.
This archetype complements the development-bank model: instead of doing everything in-house, the state backs private deep-tech GPs with aligned mandates.
10. Clusters, testbeds & shared infrastructure
(Catapult Network, DOE national labs & AI testbeds, NSF AI programmable labs, etc.)
10.1 What this model is
Deep tech is often bottlenecked not by capital but by infrastructure:
hardware (labs, fabs, pilot lines),
data (secure real-world datasets),
environments (test cities, grids, networks).
This archetype is about building shared testbeds and clusters that multiple firms can use to build, test and demonstrate technology.
10.2 Catapult Network (UK): cluster-style applied R&D centres
The Catapult Network comprises independent, not-for-profit technology and innovation centres that bridge research and industry. Parliamentary reviews describe Catapults as unique institutions that “bridge the gap between research and industry to turn great ideas into new products and services” and emphasise their role in the UK’s R&D system. UK Parliament Committees+1
Impact evidence from Catapult and government sources:
Catapults help businesses access growth markets, anchor high-value jobs and attract inward investment. The Catapult Network+1
Specific Catapults show significant economic impact; for example, the CSA Catapult impact study estimates that its activities support around 3,393 safeguarded FTE jobs and over £600M of UK GVA, with about half of projects reaching TRL 6 or above (i.e. demonstration stage).
Facilities include:
advanced manufacturing lines,
power electronics labs,
5G/6G testbeds,
robotics environments, etc.
These are testbeds and cluster anchors: startups and corporates co-locate around them.
10.3 DOE National Labs & dedicated AI / energy testbeds (US)
US Department of Energy national labs host a variety of testbeds that deep-tech firms and researchers can access:
DOE lists dedicated AI testbeds at seven national labs, which support:
AI hardware development and testing,
reliability testing, and
application development for DOE’s larger-scale production computing facilities. The Department of Energy’s Energy.gov
These testbeds range from single processors to systems with hundreds of nodes and explore diverse computing architectures and accelerators. The Department of Energy’s Energy.gov
In parallel, DOE and its labs run accelerators and entrepreneurial programmes that increase the number of technologies turning into commercial ventures, using labs as platforms for entrepreneurs both inside and outside the lab system. globalventuring.com
More broadly, US energy policy discussions emphasise testbeds for modernising energy systems (microgrids, grid storage, hydrogen), recognising that shared demonstration sites are essential for commercialising energy deep tech. Amazon Web Services, Inc.
10.4 NSF AI-programmable cloud laboratories & automated science
The US National Science Foundation recently announced investment in a national network of AI-programmable cloud laboratories:
Funding opportunity to test, scale and demonstrate new methods and tools for automated science and engineering, with a view to strengthening US leadership in science and technology. NSF - U.S. National Science Foundation
This is a testbed at the methodological level: infrastructure for AI-driven experimentation (robotic labs, automated workflows) open to multiple researchers and eventually startups.
10.5 General cluster/testbed logic
From these examples:
Shared, neutral infrastructure
Facilities are often operated by neutral entities (Catapults, labs, consortia) so many firms can use them without competitive conflicts. UK Parliament Committees+1
Direct link to higher TRLs
Testbeds are explicitly designed to push technologies into TRL 5–7 (demo/prototype in relevant environment) – exactly where many deep-tech projects stall. csa.catapult.org.uk+2The Department of Energy’s Energy.gov+2
Cluster formation
Physical co-location around these facilities leads to regional clusters (e.g., power electronics around CSA Catapult, AI/HPC around DOE labs).
Open access with clear rules
Access is usually defined via calls or service agreements; IP on results, cost structures and safety/regulatory policies are standardised, enabling repeatable use by startups.
Mission alignment
Facilities are tied to national missions: net-zero, energy security, AI leadership, etc. (Catapults for net-zero/industry, DOE for energy, NSF for AI-driven science). Digital Catapult+2The Department of Energy’s Energy.gov+2
 | A guest post by
|