This chapter was written by Lorenzo Chiavarini and Felix Ullmer (Dealroom.co), Gaétan de Rassenfosse (EPFL), Kritika Saxena (University of Groningen) and Davide Bonaglia, Christopher Harrison, Hong Kan, Shengyu Qin and Sacha Wunsch-Vincent (all WIPO).
This year’s special GII theme looks to the future of science-powered entrepreneurship and asks:
How do deep science startups and spinouts scale?
What will it take to transform deep science startups into growth and productivity engines?
The question of how to unlock the full potential of scientific breakthroughs has become central to innovation policy worldwide.
Are we advancing fast enough to turn frontier science into scalable industries, or will promising discoveries stall in the long corridor between laboratory and marketplace?
Throughout history, major leaps forward in living standards have come about not from scientific discovery alone, but from its systematic translation into the novel products, services and production processes that reshape industries.
Recent decades have seen substantial investment into frontier science – from synthetic biology and quantum technologies to AI-accelerated materials discovery and clean energy. And public and private R&D spending has reached historic highs (
At the same time, translating scientific advances into products, firms and industries of the future has never been automatic. How this is best achieved remains a central question for innovation policy that has yet to be settled. One view locates the binding constraint within the process of translation, that is, the work of turning discoveries into investable, scalable ventures. Another notes that disruptive discovery has itself become more challenging (
Rather than assume that once an invention has been perfected only its translation remains to be solved, this chapter treats the science-to-industry gap as intrinsic to deep science: that is to say, the very traits which make a discovery valuable, namely, its novelty, complexity and distance from existing markets, are precisely what makes it difficult and costly to commercialize. And this is where deep science entrepreneurship is so important.
This year’s GII 2026 special chapter examines which and how deep science startups and spinouts are able to scale, what obstacles there are in the way, and what needs to be done about them (
This chapter begins by defining what deep science entrepreneurship is and why it matters now. The phenomenon is quantified and its importance gauged using new data. Obstacles to scaling are then analyzed and deep science entrepreneurship in developing economies examined. The importance of intellectual property to the success of deep science ventures is underlined. And the chapter concludes by summarising the policies and practices needed to strengthen deep science entrpreneurship ecosytems based on the evidence pesented.
Essential to this work are the Dealroom–WIPO Deep Science Startup Tracker – the first global mapping of deep science startups;SPARK: Patenting in Deep Science Startups and Spinouts (Global Innovation Index 2026 Special) (
What is deep science entrepreneurship, why does it matter, and what is new?
Deep science entrepreneurship refers to startups and spinouts in fields such as advanced materials, energy science, genomics, health technologies, quantum physics, and synthetic biology – domains requiring extended research, specialized infrastructure and rigorous validation. The term is narrower in scope than what is referred to by investors as “deep tech” and excludes software, e-commerce, fintech, and other digital-native services.
What is more, these ventures differ from software firms in four key respects (Box 1). They rest on scientific and engineering breakthroughs and need multi-stage validation, prototyping and trials before there is any revenue (
Longer development cycles. Validation, prototyping, regulatory approval, clinical testing, certification, and manufacturing scale-up span 5–15 years before revenue.
Higher capital intensity. Wet labs, clean rooms, testing facilities and pilot lines are needed before product-market fit can be tested.
Interdisciplinary, shifting teams. Success requires researchers, regulatory experts, process engineers and commercialization professionals. Moving from research team to scalable venture means the passing of those “critical junctures” that bring in external entrepreneurs and shift control structures (
Clear IP architecture. Deep science startups collaborate with universities, incumbents and other startups for the purposes of research access, manufacturing, distribution and complementary technologies (
Multiple commercialization routes. New knowledge reaches markets via incumbent commercialization, licensing or new venture formation (
Policy implication. Seed funds, pitch contests and accelerators built for fast-scaling digital ventures fit deep science poorly. What is needed is proof-of-concept funding, translational infrastructure (shared labs, pilot facilities, testing), regulatory expertise and staged financing matched to deep science risk profiles.
Several developments make the question of how to power entrepreneurs at the frontier of science more pressing now than ever before.
On the supply side, advanced artificial intelligence (AI) is shortening paths of discovery and expanding the candidate set faster than validation capacity can grow, with deployed systems already producing measurable productivity gains (
Brynjolfsson et al., 2025Brynjolfsson, E., D. Li and L. Raymond (2025). Generative AI at work. Quarterly Journal of Economics, 140(2), 889–942. ). Frontier fields are converging: AI-accelerated protein design is reshaping drug discovery, advanced materials are enabling new battery chemistries, and foundation-model robotics are crossing from the laboratory to the factory floor(see GII 2026 Expert contribution from Han, et al. Han, F., L. Guo and D. Guo (2026). Spirit-AI: Science-driven entrepreneurship in embodied AI. In Global Innovation Index 2026. Hangzhou and Geneva: Spirit AI and WIPO. on science-driven entrepreneurship in embodied AI).On the demand side, biomedical translation is strengthening after years of contraction.
Global equity funding for science-intensive ventures reached USD 122 billion in 2025, up 7.4 percent on 2024, amid a broader rebound in venture capital (VC) after the 2022–2023 correction and the continued concentration of megadeals in AI and science-intensive fields, driven less by deal count than by nine- and 10-figure rounds becoming standard in quantum computing, fusion, AI-native drug discovery, and long-duration storage (
Importantly, intermediaries and policies to foster translation of science have spread worldwide (
Finally, emerging economies are moving beyond information and communication technology (ICT) back-office leapfrogging to backing high-growth, science-intensive sectors directly, financing them through public funds rather than waiting for largely absent VC to back domestic scientific domains directly (see "What role does intellectual property play in deep science entrepreneurship?").
How large and important is deep science entrepreneurship?
This next section discusses the extent and distribution of deep science entrepreneurship, as measured by the Dealroom–WIPO Deep Science Startup Tracker.
To be included in the Dealroom–WIPO Deep Science Startup Tracker, startups and spinouts are selected according to three criteria: (i) they need to have received at least one VC round or more (so-called Pre-Seed, Series A, B or C+); (ii) be in one of nine deep science fields – Life Sciences; Medical Devices and Digital Health; Food and Agritech; Semiconductors; Robotics and Autonomous Systems; Space; Energy; Transportation; Advanced Materials and Manufacturing; as well as (iii) satisfy specific requirements for technology risk and capital intensiveness (see Appendix V). The startup and spinout data produced via the tool for this chapter covers only the period 2000 onward. This timeframe provides a time horizon that is long enough for deep science startups to have developed and grown – some companies taking more than a decade to validate a given technology – but also maintains adequate focus on recent innovation activity.
However, these filters have limitations. For instance, there are some deep science startups that might be operating without any VC funding; while rare, they will be missed by this analysis, given the filters applied. Moreover, removing all failed deep science ventures with certainty from the database, as well as detecting fully and accurately very recent entries, for example, those in 2025, is “mission impossible.” Finally, as with any similar data exercise, the country coverage of deep science firms, including in low- and certain middle-income economies, is solid, but nonetheless imperfect. To upgrade further, as of the GII launch in September 2026, there will be a notification system in place whereby WIPO member states and deep science firms can flag omissions, with a view to improving global coverage over the next few years.
Deep science firms vs. startups and spinouts since 2000
Based on the above definition, it is expected that by end-2025 more than 30,000 deep science firms will have been created since 2000 – 9.5 times more than there were 15 years earlier in 2010, representing a compound annual growth rate of 16.2 percent. From 2024 to 2025 alone, the number of deep science startups grew by around 6.3 percent, from 28,200 to almost 30,000 (Figure 1).
In absolute terms, the number of new deep science startups and spinouts reached a high in 2018, then again in 2022, and sat at 1,800 in 2025 (note that 2024 and 2025 figures are likely underestimated, as information on funding tends to be disclosed after a time lag) (Figure 2).
In contrast, by the end of 2025, slightly more than 250,000 VC-backed startups and spinouts (including deep science ones) had been created since 2000 – about 13 times more (13.1×) than there were in 2010, representing a compound annual growth rate of 18.7 percent. From 2024 to 2025, the number grew by 4.5 percent, from 243,000 to 254,000.
Among the panoply of VC-backed startups, about one in 10 new ventures is of the deep science type as defined here. This share remained relatively stable between 2010 and 2024.
Despite incomplete data, 2025 looks like an outlier, with the 16 percent share a 15-year high.
The highest number of deep science startups and spinouts was in the United States of America (US), with 12,752 firms (1,980 of which were late-stage) followed by China 3,466 (582 late-stage), the United Kingdom (UK) 2,133 (136 late-stage), France 1,234 (59 late-stage), Germany 1,051 (57 late-stage), Canada 1,014 (65 late-stage), Israel 828 (80 late-stage), Switzerland 719 (49 late-stage), Japan 715 (40 late-stage) and the Republic of Korea 665 (34 late-stage) (Table 1).
Other notable economies outside of Northern America and Europe are India 453 (18 late-stage), Australia 387 (19 late-stage) and Singapore 193 (12 late-stage), with Singapore illustrating how a small, open economy can build a deep science startup hub through coordinated finance, infrastructure and talent policies (see GII 2026 Expert contribution from Tan). The top economies within the Latin America and the Caribbean region are Brazil, with 53 firms (2 late-stage) – where the GII 2026 Expert contributions examine both Brazil’s deep science entrepreneurship policy challenge and the profile of its science-based startups – followed by Argentina with 39 firms (1 late-stage) (see GII 2026 Expert contributions from Pereira and Lopes Filho; and from Pimentel and Delgado).
The top economy in Sub-Saharan Africa is South Africa, with 25 firms, including AI Diagnostics, a startup providing an AI-powered digital stethoscope for tuberculosis screening.
In Northern Africa and Western Asia (outside of Israel mentioned earlier), the United Arab Emirates, with 16 firms (2 late-stage), including SARsatX, a space startup developing small satellites; and Egypt, with 14 firms, including Pearl Semiconductor, a fabless semiconductor startup, which leads in high-performance integrated circuits (ICs).
China is showing the strongest acceleration among the top 20 economies in terms of deep science startup and spinout count, with a 138 percent increase between 2020 and 2025, compared to 59 percent for the United Kingdom and 30 percent growth for the United States.
When looking at smaller economies, Estonia shows the strongest growth (+294 percent) (see Figure 3). Estonian Aio.bio, a foodtech startup, leverages biotechnology to create sustainable alternatives to traditional vegetable oils and animal-based fats; and Estonian UpCatalyst, focuses on bringing carbon nanomaterials for various applications to market.
Türkiye (118 percent), India (117 percent), Malaysia (117 percent), Brazil (96 percent) and South Africa (71 percent) are the top global middle-income economies in terms of growth of deep science startups, with Brazil’s trajectory discussed in the GII 2026 Expert contributions from Pereira and Lopes Filho and from Pimentel and Delgado.
By region, the largest concentration of deep science startups is in Northern America, with 13,773 (2,045 late-stage), followed by Europe, with 9,303 (457 late-stage) and South East Asia, East Asia, and Oceania, with 5,602 (699 late-stage).
In turn, Central and Southern Asia is the region showing the strongest growth in the number of VC-backed companies, with a 118 percent increase in deep science startups between 2020 and 2025 – strongly driven by India. Latin America and the Caribbean, and Sub-Saharan Africa have also seen stronger growth in numbers of firms, with 71 percent and 65 percent, respectively, between 2020 and 2025, than have either Northern America or Europe, albeit from lower levels.
In value terms, and based on the last known valuation only, the 30,000 deep science startups tracked here are worth USD 7.6 trillion, up 23 percent from 2024 (Figure 4).
Unsurprisingly, deep science ventures are particularly prone to receiving greater late-stage funding. According to the reference framework for this chapter, 10 percent of deep science ventures since 2000 have been in receipt of late-stage funding versus 4 percent of all VC-backed startups taken together.
Venture capital funding into deep science startups amounted to USD 122 billion in 2025, the third highest year for funding, just behind the zero interest rate-fueled peak of 2021–2022 and up 7.4 percent on 2024. Deep science startups attracted 27.8 percent of total VC funding in tech startups in 2025, a slight decline from previous years due to an unprecedented amount of capital having been funneled into foundational AI model makers, which are excluded from the deep science scope. Excluding the USD 250M+ financings skewed by megadeals for foundational AI model makers, deep science startups attracted 36.5 percent of global VC funding in 2025, which is almost on par with the highest ever 36.6 percent reached in 2024, and nearly double the 19.7 percent share a decade ago in 2015.
As to sectors and applications, the composition of the new deep science startups is shifting away from the Life Sciences and toward Semiconductors and Robotics and Autonomous Systems (see Figure 5, while noting that the shares shown are global aggregates, and mask a significant heterogeneity across economies).
Life Sciences is the largest sector, accounting for over 11,000 of the 30,000 deep science startups initiated between 2015 and 2025, more than 1,660 of which were late-stage. However, its share of new startups has fallen sharply over the period, from 55 percent a decade ago in 2015 to 23 percent in 2025. Leading firms span income levels, ranging from the Republic of Korea’s Celltrion, a biosimilars pioneer, to Viet Nam’s Gene Solutions in molecular diagnostics and liquid biopsy.
Medical Devices and Digital Health is the next largest sector and one that is relatively established, with 5,600 startups (nearly 400 late-stage), led by firms such as Finland’s wearable pioneer ŌURA and Mexico’s AI-radiology startup Eden.
Semiconductors is the fastest-rising major sector. Its combined enterprise value (VE) has grown over 410 percent since 2020 to USD 1.1 trillion, making it the third-largest segment by value, while also becoming the second-largest source of new startups in 2025 (23 percent, up from a historical 10–15 percent). Value growth is driven largely by Chinese firms (Cambricon Technologies, Hygon, and SMIC), though leaders elsewhere include Canada’s quantum-chip company Xanadu.
Advanced Materials and Manufacturing and Space, though smaller, are fast growing sectors (+128 percent and +101 percent since 2020, respectively), with Luxembourg’s graphene-nanotube maker OCSiAl and India’s Skyroot Aerospace included among the top deep science startups.
Robotics and Autonomous Systems is also accelerating, reaching third place both for new startups (17.6 percent) and VC funding in 2025 (see Table 2). Examples range from China’s Unitree, a leading humanoid-robot maker, to Nigeria’s drone startup Terra Industries.
One advantage of the Dealroom–WIPO Deep Science Startup Tracker is that the most promising deep science startups can be detected and tracked at the economy level, including for the purpose of analyzing characteristics and IP use (see "What role does intellectual property play in deep science entrepreneurship?").
One can also use the Deep Science StartupTracker to construct particular league tables. In respect to deep science startups since 2000, Tesla and SpaceX lead, followed by CATL (Table 3). The United States and China account for 18 of the top 20 startups, with one entry each from the Kingdom of the Netherlands (argenx) and the Republic of Korea (Celltrion).
Beyond the United States, China and high-income economies more broadly, India claims the most valuable deep science company among middle- and low-income economies, with Laurus Labs being valued at USD 9.3 billion (Table 4). (India’s broader science-based entrepreneurship ecosystem is discussed in the GII 2026 Expert contribution (
Among other middle-income economies, Brazil (New Wave, USD 480 million), whose deep science entrepreneurship promise and constraints are discussed in the GII 2026 Expert contribution from Pereira and Lopes Filho, Argentina (Bioceres Crop Solutions, USD 257 million), Türkiye (MicroAlgaex, USD 250 million), Serbia (Seven Bridges, USD 250 million) and Malaysia (ALPS Global Holding Berhad, USD 141 million) all feature companies that have achieved significant scale in sectors ranging from energy to Life Sciences and Food and Agritech.
In Sub-Saharan Africa, Nigeria (Terra Industries, USD 100 million) and South Africa (Aerobotics, USD 68 million) emerge as early, but notable deep science hubs, with strengths in robotics and agritech, respectively.
University spinouts
A final way of looking at the data is to use it to distinguish between deep science spinouts – firms originating from research carried out at a university or research center – and other deep science ventures.
Evidently, spinouts of academic research are more relevant for some deep science sectors than others (see Figure 6). In Energy, 42 percent of deep science startups since 2010 have been spun out of academic research, followed by Life Sciences, with 41 percent.
In Advanced Materials and Manufacturing, and Medical Devices and Digital Health, slightly over one-third of companies are spinouts. Space, Transportation, and Robotics and Autonomous Systems are the three sectors with the lowest shares of companies that are academic spinouts, meaning that most of the deep science startups in these fields have founders with an industry background.
Some sub-segments stand out as being very academically driven. They include Quantum and Photonics in Semiconductors, with 57 percent and 49 percent of startups being spinouts, respectively, compared to an average of 28 percent for Semiconductors.
US institutes have been behind 15 of the top 20 deep science spinouts since 2000, with a wide range of institutes involved, the most recurring being Harvard (especially Harvard Medical School), MIT, and Stanford University (Table 5). China, meanwhile, claims the most valuable spinout, with semiconductor developer Cambricon Technologies having spun out of research conducted at the Chinese Academy of Sciences within the scope of the Cambricon project to develop a brain-inspired processor chip specialized for deep learning.
German research notably underpins two of the top 20 spinouts, both of which are in the Life Sciences: Alnylam, a US-based company spun out of MIT, which also licensed foundational IP from the Max Planck Institute for Biophysical Chemistry, and BioNTech from Johannes Gutenberg University Mainz, whose mRNA technology underpinned the first approved COVID-19 vaccine developed in partnership with Pfizer.
Worth highlighting too is PsiQuantum, a US-based quantum spinout created from UK research. The company was started in the United States in 2015 by four academics from the University of Bristol in the United Kingdom and has grown to be the most heavily funded private quantum startup globally.
Notable deep science spinouts have emerged from research across a very wide range of economies and sectors.
India’s SEDEMAC Mechatronics, spun out of IIT Bombay, is a notable example of deep engineering research translating into significant commercial impact within the transportation sector.
Finland and Japan demonstrate the strength of their space research through Finland’s ICEYE, spun out of Aalto University, and Japan’s Astroscale, from the University of Tokyo, among the most significant spinouts globally in the category.
Iyris (formerly known as RedSea) is a spinout from King Abdullah University of Science and Technology (KAUST) in Saudi Arabia, which provides proprietary nanomaterial additives to optimize how surfaces interact with sunlight, with applications for farmers and to reduce agricultural water usage.
Spain’s Splice Bio, co-founded by the University of Barcelona and Princeton University, exemplifies the value of cross-border academic collaboration in generating commercially significant IP, in this case in the field of RNA therapeutics.
Brazil’s Biotimize, spun out of São Paulo State University, reflects the growing maturity of Latin America’s life sciences research base – a theme developed in the GII 2026 Expert contributions from Pereira and Lopes Filho, and from Pimentel and Delgado – with the company developing computational approaches to biological data.
Similarly, Argentina’s Puna Bio, emerging from CONICET – the country’s national research council – illustrates a broader trend of Latin American public research institutions increasingly serving as the foundation for commercially-oriented agritech ventures, in this case focused on sustainable agricultural inputs.
South Africa’s HearX Group, from the University of Pretoria, stands out for applying digital health innovation to hearing care – a condition disproportionately affecting lower-income populations.
Readers may explore and discover more in the Dealroom–WIPO Deep Science Startup Tracker that complements this chapter.
Obstacles to scaling deep science ventures
This section discusses the challenges faced by deep science ventures. Even with record financing flows, deep science ventures stall at predictable transitions: research to investable proof-of-concept, pilot to commercial deployment, working technology to profit-making venture. Long development cycles, capital intensity, specialized infrastructure, regulatory complexity and uncertain demand interact across these stages, and binding constraints shift as ventures progress. Capital alone does not remove these obstacles; where complementary inputs are missing, larger rounds serve to inflate expectations more than they do output.
Financing cliffs recur at predictable transitions. Science-based ventures need capital across several years, often a decade or more – for R&D, prototyping, trials, reliability proofs and regulatory clearance – before revenue arrives. The two best-documented cliffs sit between research and investable proof-of-concept, and between pilot and commercial deployment, where Series B and C rounds in the USD 20–100 million range typically remain thin (
Industrialization, not invention, is now the binding constraint in most fields. Frontier technologies depend on specialized labs, testbeds, metrology, pilot plants, clean rooms, bioprocessing capacity, and regulatory sandboxes – lumpy investments that single startups are rarely able to finance alone. Public and consortium facilities lower fixed costs and accelerate iteration. Unlike digital products, deep science ventures must build manufacturing processes, quality systems and supply chains, often according to strict standards (
Institutional and regulatory friction blocks ventures before they reach market. At formation, contracting delays, unclear publication rights, restrictive equity and royalty clauses, and opaque licensing terms can stall spinouts before they incorporate, as illustrated by the GII 2026 Expert contributions from Kretzschmar and Dols on CERN Venture Connect, from Sivakumar and Cabrera on university commercialization reform, and from Pereira and Lopes Filho on Brazil’s deep science entrepreneurship ecosystem. The intermediaries that broker these transitions are themselves capability-constrained: deal-structuring skills are scarce, TTO strategies vary widely, and the performance of research parks and incubators depends heavily on the quality of governance and integration into the surrounding system (
Talent and network constraints cap execution. Deep science ventures need researchers, engineers and technicians who possess advanced expertise in AI, biotechnology, materials science and quantum computing skill sets are in short supply almost everywhere, and particularly thin in emerging fields not yet integrated into university curricula. Industry surveys consistently report substantial unfilled vacancies. Junior supply is one part of the picture; senior researcher mobility is the other. Evidence from biotechnology shows that “star” scientists and locally concentrated human capital are strongly associated with firm formation, putting researcher mobility and academic engagement policies alongside IP frameworks in determining translation outcomes (
Demand uncertainty compounds technical risk. Even when a technology works, early customers can be hard to secure, regulated adoption pathways delay revenues and nascent markets require education and proof. Without credible demand pull, pilots remain one-off demonstrations rather than repeatable orders, as shown in the GII 2026 Expert contributions on CATL’s IP and R&D practices (
Addressing these obstacles requires acting across the chain: patient financing, talent pipelines, network facilitation, infrastructure, regulatory agility, and institutional designs fitted to high-risk, long-horizon ventures. The following section turns to the issue of deep science entrepreneurship, particularly in developing economies.
Deep science entrepreneurship in developing economies: building translation capacity where it matters most
Deep science entrepreneurship is increasingly visible outside high-income hubs, anchored in universities or in problem-led needs across health, agriculture, climate adaptation and energy access. Whether productivity gains diffuse beyond existing innovation ecosystems will depend on translation and adoption capacity – facilities, standards, validation and early buyers.
Three shifts have pushed the topic onto policy agendas in lower- and middle-income economies: (i) mission pressure from pandemics, climate shocks and food insecurity; (ii) rising research and university entrepreneurship across parts of Africa, Asia and Latin America, albeit unevenly (
Examples are industrial-scale vaccine manufacturing at Serum Institute of India (India) and Biovac (South Africa); BioNTech’s USD 150 million mRNA facility in Rwanda; pay-as-you-go off-grid solar reaching millions in East and West Africa; precision fermentation in Brazil (
Four commercialization pathways
Deep science ventures cluster around four archetypes, each having distinct bottlenecks in developing economies.
University-anchored spinouts start from research groups and need proof-of-concept funding, credible technology-transfer terms and founder-team support. South Africa accounts for nearly half of identified university spinouts in Africa, backed by statutory IP frameworks; the WIPO Scale Up Your IP program convened four South African universities and twenty of their spinouts in March 2025 (
WIPO, 2025aWIPO (2025a). Empowering Deep-Tech Ventures with IP: Insights from the Scale Up Your IP Program in South Africa. Geneva: WIPO. ;Saxena, 2026Saxena, K. (2026). From research to scale: Evidence on deep-tech startup development in developing economies. Background study prepared for the Global Innovation Index 2026. Geneva: WIPO. ). Capacity gaps persist – investors still favor software and fintech over biotech, and Uganda expects inventors to cover IP protection costs (Saxena, 2026Saxena, K. (2026). From research to scale: Evidence on deep-tech startup development in developing economies. Background study prepared for the Global Innovation Index 2026. Geneva: WIPO. ).Problem-led ventures with demand pull start from urgent local problems and depend on procurement, regulated market access and often public or donor-backed early buyers. Brazil’s Public Contract for Innovative Solutions allows contracts of up to USD 1.5 million without prescribing the direction of innovation, as discussed in the GII 2026 Expert contribution from Pereira and Lopes Filho; Nigeria’s Startup Act (2022) mandates a 10 billion naira annual floor for a startup seed fund (
Saxena, 2026Saxena, K. (2026). From research to scale: Evidence on deep-tech startup development in developing economies. Background study prepared for the Global Innovation Index 2026. Geneva: WIPO. ).Consortium platform ventures emerge from shared facilities (health data networks, climate services, testing labs) where governance, data rights and IP sharing are central. The WHO mRNA Technology Transfer Programme’s South Africa hub, covering 15 partner economies, shows how shared platforms can build manufacturing capability across economies. The African Union’s Partnership for African Vaccine Manufacturing (PAVM) targets 60 percent local production of vaccine doses by 2040, up from roughly 1 percent today, with at least USD 3.5 billion having been committed (
Clinton Health Access Initiative, 2024Clinton Health Access Initiative (2024). Analysis of Vaccine Manufacturing Commitments in Africa. New York: CHAI. ). Realizing such targets depends on procurement commitments, not manufacturing capacity alone.Export-to-scale models develop locally, but scale via regional or global markets, depending on standards, certification, and manufacturing and distribution partnerships. Indian biosimilar manufacturers built global scale by combining domestic capacity with international regulatory expertise and through ties with multinational pharmaceutical companies.
Structural constraints
The commercialization pathways described face binding constraints that differ in degree, not in kind, from those confronting high-income economies. They are as follows:
Weak upstream R&D intensity. R&D intensity sits below 0.5 percent of gross domestic product (GDP) across many low- and middle-income economies, and roughly seven in 10 economies worldwide maintain R&D-to-GDP ratios under 1 percent (
Capital fragmentation and horizon mismatch. In emerging economies, grants often fail to connect to venture rounds, and in turn venture rounds to scale finance, thus raising the cost of capital. Conventional fund horizons have a 10 or 15-year timeline, which is often too short (
Translation infrastructure and capability gaps. Shared labs, testbeds, pilot manufacturing, and certification pathways are scarce, and such gaps interact: for example, without testbeds, investors price risk higher; without finance, facilities are underused; without demand, pilots do not become repeat orders. Some economies are building institutions explicitly to address the translation of scientific discovery into products, services and production processes (Uganda, for instance, has commercialization guidelines and coordination frameworks), though implementation capacity remains thin (
Demand architecture and the double valley of death. Deep science ventures face two transition cliffs: research to concept, and pilot to diffusion. Both are steeper within a developing context: demand signals are weaker, industrial partners fewer and validation infrastructure thinner (
AI cuts across these constraints. As noted in the section "How large and important is deep science entrepreneurship?", AI accelerates discovery, but increases pressure on those institutions that validate it – namely, standards bodies, regulators, patent offices. Dependency on compute, platforms and governance arrangements can redirect scarce resources away from deep science priorities, unless actively managed (
Policy: from copying to fit
A recent review describes a shift in lower- and middle-income economies away from broad-based small and medium-sized enterprise (SME) support toward Startup Acts, legal recognition frameworks and targeted incentives designed to address institutional voids (Saxena, 2026). Such legislation signals a state-level commitment to IP and incentives. But instruments must match capability; in low-capability settings, highly sophisticated instruments underperform, and copying high-income economy models without diagnosing what are the missing complements risks disappointment. Technology transfer performance varies between knowledge transfer offices (KTOs), and building entrepreneurial capacity in universities requires external links, incentive alignment, resource access and integration into innovation systems (
The science-first instruments outlined in this chapter (India’s Deep Tech Alliance and Brazil's Public Contract for Innovative Solutions) are a reflection of local strengths rather than imitation (
Policy recommendations themselves need to be differentiated by context: building the absorptive capacity of firms and providing incentives for industrial engagement matter more for middle-income economies than does reforming university IP ownership, and wholesale convergence on standard IP frameworks without the complementary conditions being in place has produced disappointing commercialization results (
What role does intellectual property play in deep science entrepreneurship?
For deep science ventures, IP is not optional. The 5–15 year gap between invention and revenue means patents are often the only verifiable asset on the balance sheet for years at a time with which to secure the financing, partnerships and licensing arrangements that ventures depend on before a product generates revenue. This temporal function distinguishes deep science from other tech-based ventures such as software, where execution speed and network effects typically substitute for formal protection.
IP architecture also defines the boundaries of collaboration, a point illustrated in the GII 2026 Expert contribution from Luo on CATL’s early-stage IP practices and R&D empowerment. Deep science ventures rarely scale through stand-alone growth; they license to incumbents, form joint ventures, are acquired for their platforms or operate within consortia. For example, in mRNA vaccine development, platform technology from one entity must interface not only with antigen design from universities, but also manufacturing from contractors and distribution agreements with governments – each boundary requires explicit IP terms or else the structure stalls. Patent pools in genome editing and agricultural biotechnology illustrate how shared access to foundational technologies can be combined with freedom to operate on specific applications.
A third dimension is finance. IP is increasingly serving as security for lending. The global value of intangible assets reached USD 80 trillion in 2024, a 13-fold increase over a period of 25 years, yet most of it remains unused by mainstream lenders (
Yet empirical evidence on how patenting actually varies across deep science ventures has been thin. To close that gap, WIPO has matched all startups in the Dealroom–WIPO Deep Science Startup Tracker to global patent data, allowing the first comprehensive global comparison to be made of patenting activity across startup types, geographies, sectors and growth stages. The full results appear in SPARK: Patenting in Deep Science Startups and Spinouts (Global Innovation Index 2026 Special) (
Five patterns stand out: patent adoption is rising, and the gap between deep science and other ventures widening; geographic variation reflects institutional capacity; an income gradient is visible, but deep science startups outpace others at every income level; sectoral differences track regulatory and collaboration models; and adoption rates rise sharply with venture maturity.
Patent adoption is rising faster among deep science startups, but substantial gaps remain
The difference in patent adoption patterns between deep and non-deep science startups and spinouts confirms that IP protection is structurally more critical for deep science ventures (see Figure 7). Overall, half of deep science startups hold patents (49.7 percent), compared to substantially lower rate (15.4 percent) among non-deep science startups.
More significantly, the propensity gap between deep and non-deep science startups has widened. Over the past decade, patent adoption rates have increased for both categories, but the rate of increase has been faster among deep science startups. This divergence suggests that, as VC has gained experience in evaluating deep science opportunities, the expectation of formal IP protection as an entry condition has intensified. It may also reflect learning among deep science founders. As exits more commonly occur through acquisition or licensing rather than stand-alone IPOs, a well-structured patent portfolio is likely to be seen as an essential negotiating asset.
However, the 50 percent baseline adoption rate also implies that half of deep science ventures either lack patents entirely or have not yet filed during the startup phase captured in this analysis. Several factors may contribute to this pattern. First, early-stage startups may file patents after initial venture rounds, meaning that some of the non-patent-holding firms will file subsequently. Second, sector heterogeneity means that certain deep science domains may depend more on know-how, trade secrets or regulatory data exclusivity than on formal patents.
Geographic variation in patent adoption reflects institutional capacity and innovation system maturity
Global comparisons reveal significant differences in patent adoption (Figure 8). The Republic of Korea (69 percent), Finland (67 percent), and Japan (66 percent) recorded the highest national rates of patent adoption among deep science startups between 2010 and 2025. These rates reflect well-known institutional pathways linking university research, corporate spinouts and venture formation that systematically prioritize patent filing early on in the commercialization cycle. Several other high-income economies with mature innovation systems – including the France, Germany, Israel and the United States – cluster within the 45–55 percent range.
Patent adoption rates rise systematically with economic income level, confirming that institutional capacity matters
When economies are grouped by income classification, a clear gradient emerges showing that IP adoption among deep science startups rises with economy income level. Patent adoption among deep science startups rises from 37.6 percent in lower-middle-income economies to 50.6 percent in high-income economies. Notably, deep science startups consistently outpace their non-deep science counterparts across all income groups, confirming that the structural need for IP protection in deep science holds globally.
These findings underscore that expanding deep science entrepreneurship in developing economies, as discussed extensively in the section "Deep science entrepreneurship in developing economies," requires not only proof-of-concept funding and infrastructure, but also institutional development around IP. Affordable patent filing pathways, competent technology transfer support, and credible enforcement mechanisms together make IP protection a workable strategy rather than a prohibitive cost.
Sectoral patterns reveal where IP architecture defines commercialization pathways
Patent intensity varies markedly across deep science sectors, reflecting differences in regulatory requirements, collaboration models, technical appropriability, and target market expectations. Figure 9 shows that Life Sciences, Semiconductors, and Medical Devices and Digital Health lead international patent family growth among deep science startups.
Medical Devices and Digital Health emerges as the most patent-intensive startup sector globally, with 57.8 percent of deep science startups holding patents (not shown in Figure 9).
Space and Robotics and Autonomous Systems startups claim the lowest shares of academic spinouts (see Figure 6) and it is not surprising that they also have a correspondingly lower patent intensity. These two sectors draw more heavily on founders with an industry background, and commercialization often depends on system integration, manufacturing execution, operational know-how and customer relationships, as much as it does on discrete patentable inventions. Trade secrets, first-mover advantages and the tacit knowledge embedded in engineering teams may provide appropriability without requiring extensive formal patent portfolios.
Food and Agritech ventures also display a lower patent intensity, despite substantial technical innovation within that sector.
Patent adoption increases sharply with venture maturity, but the transition cliff is steeper than is often realised
Stage-based analysis shows that patent portfolios evolve throughout deep science startup development (Figure 10). Among deep science startups, patent holding rises from 42.7 percent at seed stage to 69.2 percent at late stage. This progression is due to several factors. As ventures move from concept validation through prototyping and clinical or field trials, it becomes clearer which inventions are worth protecting, and additional inventions requiring protection are likely to emerge. Second, ventures that are able to successfully raise later-stage funding have typically demonstrated technical feasibility and market potential, and in this regard patent portfolios often contribute to investor confidence and valuation support. Finally, as ventures approach commercialization, the need to establish freedom-to-operate, prepare for partnership negotiations or acquisition discussions, as well as defend against competitor entry, is likely to increase the need for patent protection.
East Asia and Northern America dominate deep science patent output, but regional innovation ecosystems are at very different stages of maturity
Asia demonstrates remarkable dynamism in deep science innovation, with China standing out as the most prolific patent filer. In the ranking for South-eastern Asia, Eastern Asia, and Oceania deep science startups founded since 2010, every top 20 company is Chinese, with the number one ranked company, ChangXin Memory Technologies, holding 8,311 patent families – almost five times that of Northern America’s top-ranked CTRL-labs. China’s deep science startups are predominantly concentrated in semiconductors, robotics, and transportation. Beyond China, Japan and the Republic of Korea are also significant innovation forces within East Asia.
Northern America remains a leading center of global deep science innovation. US companies dominate the Northern American top 20 (with Canada’s Synaptive Medical making a strong showing).
Europe represents a multi-country, multi-sector innovation landscape. UK companies claim the most entries on the list (including FlexEnable, CMR Surgical, and Autolus), while Finland ranks well above its size, with two high-ranking companies – ŌURA (1st) and Varjo Technologies Oy (4th). France, Germany, Sweden and Switzerland each contribute notable representatives to the rankings.
Latin America and the Caribbean is emerging as an increasingly diverse and encouraging innovation landscape. Companies from Argentina, Brazil, Chile, Mexico, Peru and beyond have all secured places in the top 20 patent rankings, with Food and Agritech and Life Sciences standing out as the region’s most distinctive areas of strength.
Israel holds an unrivalled position of leadership in deep science innovation across Northern Africa and Western Asia. All top 20 deep science startups by simple patent family count in the Northern Africa and Western Asia rankings are Israeli. Türkiye is an emerging player in the region, while Saudi Arabia and the United Arab Emirates are both beginning to demonstrate promising deep science startup potential.
India is becoming the driving force behind Central and Southern Asia’s deep science rise. Every top 20 company within this region is Indian, with health care, robotics, and energy emerging as the country’s primary areas of focus.
Sub-Saharan Africa is at an early stage of deep science innovation. While WIPO’s research and analysis identified only five patent-holding deep science startups within the region (four from South Africa and one from Kenya), the very emergence of these pioneers is itself an encouraging sign pointing to the nascent potential of the region’s deep science startup ecosystem.
Taken together, global deep science innovation is increasingly defined by a “multi-polar” dynamic.
These results should be considered in light of certain data coverage constraints, particularly in some emerging and less-documented innovation ecosystems. As the analysis draws on startup information available through the Dealroom database, some deep science startups and their related patenting activity may not yet be fully captured in every region. Additionally, please note that, while the analysis above reflects the top 20 deep science startups per region, the accompanying infographic displays only the top five as a teaser (Figure 11). For the complete tables and details, please consult SPARK: Patenting in Deep Science Startups and Spinouts (Global Innovation Index 2026 Special) (
Finally, deeper analysis of the patent filing strategies of deep science ventures reveals that 84.4 percent choose to file via the Patent Cooperation Treaty (PCT) route, whereas only 56.9 percent of non-deep science startups chose to do so. Of all of the published patent families filed by deep science startups, 31.1 percent were filed through the PCT route, compared to only 22.1 percent by non-deep science startups. Deep science startups are more likely than non-deep science ventures to use the PCT route. This is because deep science technologies are typically developed for global markets and commercialized through international partnerships, licensing or acquisition.
What policies and practices can strengthen deep science entrepreneurship ecosystems?
Deep science commercialization differs structurally from that of other tech-based startups such as software. It has a longer maturation cycle, technical and regulatory uncertainty and infrastructure dependence, and relies on complementary assets. Policy frameworks therefore need to be systemic, adaptive to sector and stage, and accountable to outcomes beyond a firm count.
The most reliable strategies strengthen the full translation stack, from proof-of-concept to pilots and on through to scaling, and align supply-side support with demand-side pathways.
Fund the missing layers in the translation stack. Between research grants and scalable private capital, ecosystems typically need proof-of-concept funding, prototyping and field validation, access to pilot and demonstration facilities, and later-stage patient capital. Workable structures include public–private matching funds that share early risk; first-loss tranches that protect institutional investors during pilot deployments; guarantees that enable debt financing for capital-intensive scaling; and reimbursable finance where commercial returns are plausible. Proof-of-concept programs succeed when structured as enabling packages – mentorship, networks and business development alongside capital; the same nominal funding yields different results depending on whether it includes technical validation, regulatory guidance and connections to follow-on resources (
Activate demand. Some deep science ventures stall less because the technology does not work than because no first scalable market emerges. Procurement pathways, standards development and offtake commitments lower investor risk and accelerate learning-by-doing. Where governments or large buyers commit to multi-year purchases, breakthroughs become financeable projects rather than stranded pilots. Mission-oriented programs around societal challenges can create explicit demand pull. Procurement frameworks that emphasize functional requirements rather than prescribed technical solutions allow competing approaches to emerge; performance-based contracts and pooled purchasing turn pilots into pipelines, building repeatable orders rather than one-off demonstrations, as discussed in the GII 2026 Expert contributions from Pereira and Lopes Filho on Brazil and from Katragadda on India.
Standardize and professionalize technology transfer. Variation and opacity in respect to terms – for example, equity stakes, royalties, board seats and milestone clauses – inflate negotiation costs and make startups in general less investable. Standardized templates, service-level time targets and transparent benchmarking reduce friction. Measuring tech transfer performance by downstream outcomes (follow-on investment, partnerships, product milestones) rather than disclosures and filings alone shifts the emphasis toward translation (
Invest in shared infrastructure and verification capacity. Shared labs, testbeds, regulatory sandboxes, pilot manufacturing, and certification pathways shape whether a working technology can reach industrial scale, as illustrated by the GII 2026 Expert contributions from Tan on Singapore’s deep tech hub and from Kretzschmar and Dols on CERN Venture Connect. Soft infrastructure matters equally, requiring clear and accessible certification pathways, standardized contracts that lower negotiation costs, and regulatory-science capacity that allows agencies to assess novel technologies.
Create enabling regulatory and institutional frameworks. Regulatory pathways that provide clarity while managing risk enable rather than block innovation. Sandboxes (experimental environments under maintained oversight) allow novel technologies to be tested before full frameworks exist; extending them beyond financial technology to biotech, clean energy, autonomous systems, and advanced manufacturing creates learning opportunities for both innovators and regulators. Legal recognition frameworks that distinguish deep science ventures from generic SMEs reduce uncertainty and enable targeted support, including statutory access to preferential tax treatment, procurement priority and expedited procedures (
Build the IP architecture that allows translation. Patent protection is structurally more important for deep science ventures than it is for other VC-backed startups; adoption rises sharply with venture stage; an income gradient in adoption persists across ecosystems. The policy emphasis is on the surrounding architecture: clear and affordable filing pathways for early-stage ventures and universities; well-functioning patent pools and collaborative licensing for foundational technologies (as in genome editing and agricultural biotechnology); standardization of university IP assignment terms; and improved visibility of intangible assets in accounting and disclosure. IP-backed finance is moving toward the mainstream where valuation expertise, disclosure frameworks, and guarantee or loss-sharing arrangements are in place (
Build human capital for translation. Ecosystems need regulatory, manufacturing, quality and commercialization expertise to be in place early. Updating curricula for emerging technology domains, building interdisciplinary programs and offering specialized training in commercialization-relevant skills expands the talent base. Where academic career structures penalize entrepreneurial engagement, startup formation rates suffer. Startup-leave provisions (time-limited entrepreneurial leaves with a guaranteed right to return) allow researchers to test venture formation without abandoning academic careers. Tunisia’s Startup Act (2018) provides a concrete model: within the provisions of the Act, public- and private-sector employees can take up to two years leave to build a venture, with a guaranteed right to return to their previous post if it fails (
Enable collaboration through workable governance. Complex innovations increasingly require collaboration among universities, firms, hospitals, utilities, manufacturers and governments. Consortia reduce duplication and pool infrastructure, but only with the help of workable governance, requiring IP clarity, data rights, shared standards and benefit-sharing credible to partners, while protecting public value. Pre-competitive research consortia can accelerate progress, while preserving competitive dynamics downstream. Cross-border partnerships matter particularly for smaller markets, provided that ownership, data use and dispute resolution are clearly defined.
Maintain competitive dynamics in strategic technology markets. Competition policy shapes business models and investment decisions in deep science from the outset. The task is to enable the collaboration and consolidation often necessary in capital-intensive ventures, while preventing premature lock-in and the consequent blocking of follow-on innovation. Clear guidance on permissible partnership structures and a merger review that operates at market speed help founders structure deals with predictability.
Design for learning, not for replication. No single model fits every context. Policy should begin with a diagnosis of binding constraints that is in context, not from a replication of high-profile models developed elsewhere. A recurring pitfall is the proliferation of visible programs (accelerators, competitions, incubators) that raise activity without changing translation outcomes. Spinout counts and patent disclosures are only one side of success metrics. Effective measurement tracks downstream outcomes – follow-on investment, partnership formation, sustained operations, and eventual market adoption (
A final caveat: not all deep science ventures are high-growth candidates. Research-based spinouts differ in terms of resources, commercial orientation and dependence on parent institutions, and these differences map into different growth trajectories. Support for firms directed at building capabilities for startups differs from the introduction of instruments targeted at scale-up candidates; policy design should account for such heterogeneity (
Employment outcomes are also back-loaded and highly skewed: many ventures stay small for extended periods when in the process of proving technical feasibility, navigating regulation and building manufacturable products; and a small minority of scale-ups accounts for a disproportionate share of jobs and value, meaning averages can be misleading when it comes to program evaluation (
Diagnose first. Map where ventures stall (proof-of-concept, prototype, pilot, certification or scale) and track time-to-milestone before designing interventions.
Finance the translation stack. Fund proof-of-concept, pilot validation and later-stage patient capital. Deploy blended instruments (matching funds, guarantees, first-loss tranches, reimbursable finance) matched to deep science risk profiles.
Activate demand. Use procurement with functional requirements, offtake commitments, and anchor customer programs. Consider mission-oriented programs, with government as first customer where appropriate.
Professionalize tech transfer. Standardize templates, set time targets, measure downstream outcomes. Avoid excessive equity claims and reach-through royalties.
Build infrastructure. Provide accessible labs, testbeds, pilot manufacturing and certification pathways, alongside soft infrastructure (standardized contracts and clear procedures) that lowers transaction costs.
Enable regulatory pathways. Extend sandboxes beyond fintech to include biotech, clean energy, autonomous systems and advanced manufacturing; establish legal recognition frameworks; and build regulator capacity to assess novel technologies.
Strengthen the IP architecture. IP serves three roles for deep science ventures: it secures the long gap between invention and revenue; frames the boundaries of collaboration between universities, incumbents and consortia; and underpins non-dilutive finance. Support clear and affordable filing pathways; back patent pools and collaborative licensing for foundational technologies; standardize university IP assignment terms; improve intangible-asset visibility in accounting and disclosure; and develop the valuation, disclosure and guarantee mechanisms that allow IP-backed finance to scale.
Develop translation talent. Invest in regulatory, manufacturing, quality and commercialization expertise. Update curricula for emerging technologies; reduce founder risk through startup-leave provisions and career safety nets.
Support collaboration. Facilitate consortia through clear IP arrangements, data governance and benefit-sharing; provide guidance on permissible partnership structures.
Maintain competition. Build merger review capacity that operates at market speed. Balance enabling collaboration against preventing premature lock-in.
Measure and learn. Track follow-on investment, milestone achievement, infrastructure operation and capability persistence. Refine policy based on evidence; maintain transparency in order to enable cross-jurisdiction learning.
Conclusion
Whether the deep science wave will deliver a productivity payoff is not assured, and the evidence assembled here points to no single lever that guarantees this will happen. Discovery and translation are both demanding, and the binding constraints shift according to field, stage and economy. What the record suggests, however, is that policy has a role to play through providing finance across stages, shared infrastructure, regulatory pathways matched to technology timelines, workable IP frameworks, and the talent to transition work from laboratory to market. None of which assures commercial success, but their absence is a common reason for ventures to stall. Powering entrepreneurs at the frontier of science, and turning more pilots into pipelines, are hard and uneven tasks yet to have a ready solution.
References
AlShareef, M.R., K. Alhussaini and A.A. Ababtain (2026). Saudi Arabia: Building a national ecosystem for frontier innovation and deep tech commercialization. In Global Innovation Index 2026. Riyadh and Geneva: Research, Development and Innovation Authority, King Saud University, Ministry of Investment, and WIPO. Available at: https://www.wipo.int/en/web/global-innovation-index/2026#contributors
Arundel, A., S. Athreye and S. Wunsch-Vincent (2021). Policies and practices for supporting successful knowledge transfer from public research to firms. In Harnessing Public Research for Innovation in the 21st Century: An International Assessment of Knowledge Transfer Policies. Cambridge: Cambridge University Press, 359–463. DOI: 10.1017/9781108904230.
Athreye, S. and F. Rossi (2021). Policy recommendations: Aiming for effective knowledge transfer policies in high- and middle-income countries. In Arundel, A., S. Athreye and S. Wunsch-Vincent (eds), Harnessing Public Research for Innovation in the 21st Century: An International Assessment of Knowledge Transfer Policies. Cambridge: Cambridge University Press, 393–417. DOI: 10.1017/9781108904230.022.
Auerswald, P.E. and L.M. Branscomb (2003). Valleys of death and Darwinian seas: Financing the invention to innovation transition in the United States. Journal of Technology Transfer, 28(3–4), 227–239.
Battaglia, D., E. Paolucci and E. Ughetto (2021). The role of Proof-of-Concept programs in facilitating the commercialization of research-based inventions. Research Policy, 50(6), 104268
Bloom, N., C.I. Jones, J. Van Reenen and M. Webb (2020). Are ideas getting harder to find? American Economic Review, 110(4), 1104–1144.
Bonaglia, D., L. Rivera León, M. Canales, O. Gisbert, M. Shcherbakova, J. Slee and S. Wunsch-Vincent (2025). End of Year Edition – Despite the Odds, Global R&D Spending Grew Again in 2024, Inching Closer to the USD 3 Trillion Mark. GII Innovation Insights Series, December. Geneva: WIPO. Available at: www.wipo.int/en/web/global-innovation-index/w/blogs/2025/end-of-year-edition
Brown, R. and C. Mason (2017). Looking inside the spiky bits: A critical review and conceptualisation of entrepreneurial ecosystems. Small Business Economics, 49(1), 11–30.
Brynjolfsson, E., D. Li and L. Raymond (2025). Generative AI at work. Quarterly Journal of Economics, 140(2), 889–942.
Cabinet Office of Japan (2025). Intellectual Property Strategic Program 2025. Tokyo: Intellectual Property Strategy Headquarters, Cabinet Office of Japan, June 3.
Clinton Health Access Initiative (2024). Analysis of Vaccine Manufacturing Commitments in Africa. New York: CHAI.
de Rassenfosse, G. (2026). Literature review of deep science ventures. Background study prepared for the Global Innovation Index 2026. Lausanne and Geneva: École polytechnique fédérale de Lausanne and WIPO. Available at: https://www.wipo.int/en/web/global-innovation-index/2026#contributors
de Vries, K. and S. Wunsch-Vincent (2022). What is the future of innovation-driven growth: Productivity stagnation or revival? In Global Innovation Index 2022: What is the Future of Innovation-driven Growth? Geneva: WIPO. Available at: www.wipo.int/edocs/pubdocs/en/wipo-pub-2000-2022-section5-en-special-theme-global-innovation-index-2022-15th-edition.pdf.
EUIPO and EPO (2024). Patents, Trade Marks and Startup Finance. Alicante and Munich: European Union Intellectual Property Office and European Patent Office.
Frølund, L., M. Murmann and F. Murray (2024). What is Deep Tech and Why Should Corporate Innovators Care? MIT Regional Entrepreneurship Acceleration Program (REAP) Working Paper.
Gisbert, O. and V. Behrens (2024). 2024 Venture Capital Outlook: From Freefall to the First Signs of Stabilization? GII Innovation Insights Series, October. Geneva: WIPO. Available at: www.wipo.int/en/web/global-innovation-index/w/blogs/2024/2024-venture-capital.
Gisbert, O., D. Bonaglia and S. Wunsch-Vincent (2025). AI Megadeals Fuel Venture Capital Rebound, but Hide Deepening Geographic and Sectoral Divides. GII Innovation Insights Series, November. Geneva: WIPO. Available at: www.wipo.int/en/web/global-innovation-index/w/blogs/2025/ai-venture-capital.
Han, F., L. Guo and D. Guo (2026). Spirit-AI: Science-driven entrepreneurship in embodied AI. In Global Innovation Index 2026. Hangzhou and Geneva: Spirit AI and WIPO. Available at: https://www.wipo.int/en/web/global-innovation-index/2026#contributors
Hayter, C.S., R. Fini and R. Lubynsky (2024). The creation of academic spinouts: University–business collaboration matters. Journal of Technology Transfer, 50(4), 1567–1601. DOI: 10.1007/s10961-024-10153-y.
Howell, S.T., J. Rathje, J. Van Reenen and J. Wong (2023). Innovating to net zero: Can venture capital and startups play a meaningful role? Entrepreneurship and Innovation Policy and the Economy, 2, 107–145.
Jefferies (2025). China Biotechs Reshaping U.S. Biopharma: Outlicensing Deals Rise. Jefferies Equity Research Report, July 14.
Katragadda, G. (2026). The industrial alchemy of science-based entrepreneurship: A comprehensive analysis of the Indian innovation ecosystem. In Global Innovation Index 2026. Bangalore and Geneva: Myelin Foundry and WIPO. Available at: https://www.wipo.int/en/web/global-innovation-index/2026#contributors
Kretzschmar, L. and H. Dols (2026). Venture-based technology transfer at large research organisations: The case of CERN Venture Connect. In Global Innovation Index 2026. Geneva: European Organization for Nuclear Research (CERN) and WIPO. Available at: https://www.wipo.int/en/web/global-innovation-index/2026#contributors
LADP (2025). Financing Deep Tech in Latin America: Bridging the Valley of Death. Latin American Dynamism Project Report. Mexico City: LADP.
Lerner, J. (1999). The Government as venture capitalist: The long-run impact of the SBIR Program. The Journal of Business, 72(3), 285–318.
Lockett, A., D. Siegel, M. Wright and M.D. Ensley (2005). The creation of spinout firms at public research institutions: Managerial and policy implications. Research Policy, 34(7), 981–993.
Luo, X. (2026). Early-stage intellectual property practices and R&D empowerment: The CATL case. In Global Innovation Index 2026. Ningde and Geneva: Contemporary Amperex Technology Co., Limited (CATL) and WIPO. Available at: https://www.wipo.int/en/web/global-innovation-index/2026#contributors
Modina, M., F. Capalbo, M. Sorrentino, G. Ianiro and M.F. Khan (2024). Innovation ecosystems: A comparison between university spinout firms and innovative startups – Evidence from Italy. International Entrepreneurship and Management Journal, 20(2), 575–605.
Mosey, S. and M. Wright (2007). From human capital to social capital: A longitudinal study of technology-based academic entrepreneurs. Entrepreneurship Theory and Practice, 31(6), 909–935.
Mustar, P., M. Renault, M.G. Colombo, E. Piva, M. Fontes, A. Lockett, M. Wright, B. Clarysse and N. Moray (2006). Conceptualising the heterogeneity of research-based spinouts: A multi-dimensional taxonomy. Research Policy, 35(2), 289–308.
NASSCOM (2024). Indian Tech Startup Ecosystem Report 2024. New Delhi: National Association of Software and Service Companies.
OECD (2019). University–Industry Collaboration: New Evidence and Policy Options. Paris: Organisation for Economic Co-operation and Development (OECD) Publishing.
Park, M., E. Leahey and R.J. Funk (2023). Papers and patents are becoming less disruptive over time. Nature, 613, 138–144.
Pereira, L.V. and J.A. Lopes Filho (2026). Brazil's deep tech promise: From scientific creation to scalable ventures. In Global Innovation Index 2026. São Paulo and Geneva: gen-t, Banco Fator and World Intellectual Property Organization.
Pimentel, D. and L. Delgado (2026). Profiling science-based startups in Brazil. In Global Innovation Index 2026. São Paulo and Geneva: Emerge and WIPO. Available at: https://www.wipo.int/en/web/global-innovation-index/2026#contributors
Raff, S., F. Murray and M. Murmann (2024). Why you should tap innovation at deep-tech startups. MIT Sloan Management Review, 65(2), 61–66.
Rothaermel, F.T. and M. Thursby (2005). Incubator firm failure or graduation? The role of university linkages. Research Policy, 34(7), 1076–1090.
Royal Academy of Engineering and Beauhurst (2024). Spotlight on Spinouts: UK Deep Tech Report. London: Royal Academy of Engineering.
Saxena, K. (2026). From research to scale: Evidence on deep-tech startup development in developing economies. Background study prepared for the Global Innovation Index 2026. Geneva: WIPO.
Siegel, D.S. and M. Wright (2015). Academic entrepreneurship: Time for a rethink? British Journal of Management, 26(4), 582–595.
Siegel, D.S., D. Waldman and A. Link (2003). Assessing the impact of organizational practices on the relative productivity of university technology transfer offices: An exploratory study. Research Policy, 32(1), 27–48.
Sivakumar, R. and Á. Cabrera (2026). Disrupting the disruptors: Rebooting academia's commercialization playbook. In Global Innovation Index 2026. Atlanta and Geneva: Georgia Institute of Technology and WIPO. Available at: https://www.wipo.int/en/web/global-innovation-index/2026#contributors
Solberg, A. and A. Brem (2016). Frugal innovation: The connection between innovation and sustainability. In 2016 Portland International Conference on Management of Engineering and Technology (PICMET).
Tan, C.C. (2026). Nurturing a global hub for deep tech startups: Singapore. In Global Innovation Index 2026. Singapore and Geneva: National University of Singapore and WIPO. Available at: https://www.wipo.int/en/web/global-innovation-index/2026#contributors
Tracey, I. and A. Williamson (2023). Independent Review of University Spin-Out Companies. London: Department for Science, Innovation and Technology and HM Treasury.
Vohora, A., M. Wright and A. Lockett (2004). Critical junctures in the development of university high-tech spinout companies. Research Policy, 33(1), 147–175.
Wang, J. and Y. Qian (2023). The Impact of University Patent Ownership on Commercialization. NBER Working Paper No. 31021. Cambridge, MA: National Bureau of Economic Research.
Wennberg, K., J. Wiklund and M. Wright (2011). The effectiveness of university knowledge spillovers: Performance Differences between university spinouts and corporate spinouts. Research Policy, 40(8), 1128–1143.
WIPO (2025a). Empowering Deep-Tech Ventures with IP: Insights from the Scale Up Your IP Program in South Africa. Geneva: WIPO.
WIPO (2025b). Moving IP Finance from the Margins to the Mainstream. Geneva: WIPO. DOI: 10.34667/tind.58498.
WIPO (2026). Technology SPARK Report and GII Special: Patenting in Deep Science Startups and Spinouts. Geneva: WIPO, November.
Wright, M., B. Clarysse, A. Lockett and M. Knockaert (2008). Mid-range universities' linkages with industry: Knowledge types and the role of intermediaries. Research Policy, 37(8), 1205–1223.
Zucker, L.G. and M.R. Darby (1996). Star scientists and institutional transformation: Patterns of invention and innovation in the formation of the biotechnology industry. Proceedings of the National Academy of Sciences, 93(23), 12709–12716.
Zucker, L.G., M.R. Darby and M.B. Brewer (1998). Intellectual human capital and the birth of U.S. biotechnology enterprises. American Economic Review, 88(1), 290–306.