Why Most LPs Are Underexposed to Deep Tech Relative to Their Stated Thesis

In December 2012, a company with 3 employees, no product and no revenue was auctioned off in Room 731 of a hotel at Lake Tahoe. The bidders were Google, Microsoft, Baidu, and a London startup called DeepMind. They bid against each other by email, and with no revenue figure to anchor the price, the price of the company doubled within days before closing at 44 million dollars.

The company was DNNresearch, consisting of a University of Toronto professor named Geoffrey Hinton and two of his students, Ilya Sutskever (later OpenAI co-founder) and Alex Krizhevsky. The “DNN” in the name stands for “deep neural networks.” Just months earlier, the researchers had built a program based on this technology that could recognize objects in photographs far more accurately than anything before it, winning a well-known contest by a wide margin.

But how come these companies were bidding for a business with no product or customers? It’s usually the revenue that normally gives buyers a basis for valuing a company, because it helps them estimate future growth. In this case, the bidders were valuing DNNresearch based on the technology itself and the expertise of the researchers developing it. The goal of the acquisition here was to give the buyer access to skills and knowledge in a field that is strategically important for them. And Google’s purchase did pay off. The expertise helped accelerate advances in their image and speech recognition, translation, and search capabilities, eventually contributing to products such as Google Photos, Google Lens, Pixel cameras, and other AI-driven services. 

The downside is that deals like these inevitably carry a higher degree of risk. These technologies – and the companies or research groups behind them – are still at an “early stage” when they begin attracting investors’ attention. Their applications are often not perfect, or maybe the product doesn’t even exist yet. Investors, therefore, have to carry out comprehensive due diligence before being comfortable saying yes to backing them.

Labeling them “early stage” is not exactly correct either. For a company like DNNresearch to exist, decades of difficult, foundational research in neural networks were needed. The work for such new technologies is time-consuming and expensive to develop, with no guarantee that it will ever lead to a viable real-world application. 

The progress heavily depends on the judgment and persistence of the researchers behind it. They must continue advancing the work while repeatedly trying to persuade investors that, with just a little more time and capital, it may eventually succeed. And how can anyone know for sure that it will? Few investors are willing to make that bet when safer opportunities are available. 

But this is the story behind nearly every major technological advancement. Breakthroughs like these only happen when a few people are willing to believe in the vision before the rest of the market does. 

Once the technology proves its potential, everyone suddenly wants a stake in it; then the bidding war begins, and valuations quickly rise. These are the companies we call “deep tech.”

Deep Tech, Emerging Tech, Frontier Tech… Which Is It?

Now, if you ask any investor about deep tech, most would say they actively invest and want more of it. But if you look at what they hold, it often tells a different story.

Part of the problem is that investors frequently use terms such as “deep tech,” “emerging tech,” and “frontier tech” interchangeably. That is understandable, because clear and consistent definitions are surprisingly difficult to find. Instead, these terms are often described with vague phrases such as “groundbreaking technologies at the edge of scientific and engineering research” – a definition broad enough to apply to all of them and explain almost nothing. 

These categories do overlap, but they are not identical. Before diving deeper into this topic, let’s clarify what each term means and how they differ.

1. Deep tech – The Difficulty Descriptor

Deep tech describes organizations, typically startups, built to deliver technology solutions based on substantial scientific or engineering challenges. These require lengthy R&D and large capital investment before commercialization. The primary risk is technical risk, while market risk is comparatively low, because the value of the solution (if it works) is obvious. I.e., there is a higher proportion of technical risk relative to market risk. For example, nobody doubts that people would buy cheap fusion power, but they would doubt whether you can build it.

Deep tech has science-based foundations. The core innovation comes from advances in physics, biology, chemistry, math, or engineering, long development cycles (5-15 years lab-to-market is common), and hard-to-replicate IP that creates structural barriers to entry. McKinsey’s 2025 analysis of European deep tech found that European deep tech unicorns hold 9 times more patents than their regular-tech counterparts, with robotics unicorns alone holding more than 1,500.

Deep tech examples: fusion energy (Commonwealth Fusion Systems), quantum computing (PsiQuantum, IonQ), mRNA therapeutics (Moderna pre-2020), novel neural network architectures, photonic chips, synthetic biology (Ginkgo Bioworks). Think of a company developing a new solid-state battery chemistry that could store more energy and charge faster. 

2. Emerging tech – The Lifecycle Stage Descriptor

This is the only one of the three with a genuine academic consensus definition. Rotolo, Hicks & Martin (2015) in Research Policy define an emerging technology as “a radically novel and relatively fast-growing technology characterized by a certain degree of coherence persisting over time and with the potential to exert a considerable impact on the socio-economic domain(s)… Its most prominent impact, however, lies in the future and so in the emergence phase is still somewhat uncertain and ambiguous. [1] It has five attributes: radical novelty, relatively fast growth, coherence, prominent impact, and uncertainty/ambiguity.

Now note what’s absent from this definition. There is nothing in there about scientific difficulty or capital intensity. Instead, emerging is a stage descriptor. A technology can be emerging without being remotely deep (consumer social media in 2006, no-code platforms, NFTs in 2021), and a technology can be deep without being emerging anymore (mRNA vaccines post-2021 are proven and scaled, but making a new one is still deep tech work). Emerging is also inherently temporary. Blockchain was emerging in 2016, just like generative AI was emerging in 2023, but both have largely graduated out of the label.

Emerging tech examples: gene editing, brain-computer interfaces, humanoid robotics, solid-state batteries.  Things that are radically novel and growing fast, and whose main impact is still ahead of them. Think of a company building generative AI tools for designing drugs or engineering components.

3. Frontier tech – The Position Relative to the Knowledge Edge

The definitions here get a bit more blurry. The World Economic Forum defines frontier technologies as “cutting-edge innovations poised to have the greatest impact on future economies and societies… often at the intersection of science, technology, and business, such as gene editing, quantum computing, and autonomous systems,” and further framing describes them as groundbreaking technologies existing at the edge of scientific and engineering research, the future of innovation in fields like space exploration and synthetic biology. This in itself is applicable to the previous categories, and not very helpful.

In Arcanum Ventures’ review of sector literature, frontier tech emerges less as a fixed category than as a timing label. Frontier tech describes where a technology sits right now relative to the edge of what’s been demonstrated. Something is “frontier” until it is proven at scale. Although in VC usage it tends to be a sector basket, with frontier tech companies focusing on emerging sectors like VR/AR, cryptocurrency, and quantum computing.

A technology can stay at the frontier for a very long time without ever being “emerging” in the fast-growth sense. For instance, fusion energy has sat at the edge of demonstrated capability for decades, perpetually frontier, but you wouldn’t call it emerging because it hasn’t had the explosive growth curve the emerging label requires. 

But the label is also genuinely contested. McKinsey, PwC and Deloitte include AI, blockchain/crypto and quantum in frontier tech, while some frontier-focused VCs explicitly exclude AI and blockchain, treating them as enabling tools rather than frontier sectors.

Here are some examples to help distinguish between frontier and deep techs:

Deep and frontier: PsiQuantum building a fault-tolerant photonic quantum computer. It sits at the edge of what’s been demonstrated (frontier), and the binding constraint is unsolved physics and engineering, not whether anyone wants it (deep). 

Frontier but not deep: a startup building a slick dashboard that lets enterprises submit jobs to other people’s quantum computers over the cloud. It lives squarely in the frontier sector, markets itself as quantum, but there’s no unsolved science inside it.

Deep but not frontier: a team improving the yield of an established silicon chip fabrication process. Genuinely hard science and engineering with real technical risk, needs PhDs and a fabrication environment (deep). But nobody calls semiconductors frontier anymore. The technology is proven and scaled, so the frontier label fell away decades ago.

What Makes Deep Tech Attractive 

Successful deep tech companies often combine large addressable markets with defensible technical moats. Their value is based on proprietary science and specialized talent that competitors cannot easily replicate. The common argument against deep tech is that companies require longer lifetimes and higher capital commitments, which puts them in the “risky” category. However, deep tech-focused funds actually deliver returns broadly in line with traditional venture funds. Over the previous five years, weighted average IRR was 26% for deep tech-focused funds versus 21% for traditional VC funds. On a non-weighted basis, the figures were almost identical, at 25% and 26% respectively. 

In portfolio terms, deep tech is tied to areas such as energy, defence, advanced manufacturing, robotics, compute and life sciences, where demand is increasingly driven by industrial, sovereign and infrastructure needs rather than consumer adoption alone. This is reflected in the market as well, with capital increasingly recognizing deep tech as a viable venture allocation due to its great demand.

The Deep Tech Market is Growing

For most of venture capital’s history, companies commercializing new science were treated as separate industries. Semiconductor startups, synthetic biology companies, and robotics manufacturers appeared to occupy unrelated markets, despite sharing similar financing and commercialization challenges. The term “deep tech” entered the venture vocabulary in the mid-2010s and gave these companies a common identity, meaning businesses whose central advantage came from original science or substantial engineering advances.

Getting that label brought practical benefits. Investment categories shape how capital is organized and give fund managers a coherent mandate. Once deep tech became legible as a category, specialist funds could raise around a coherent strategy, and more venture capital began flowing into deep-tech companies. 

BCG’s An Investor’s Guide to Deep Tech (November 2023) found that deep tech’s share of global venture funding had doubled over the preceding decade, from around 10% to a consistent 20%. The trend has since accelerated sharply. Analysis presented by Celesta Capital in April 2026, built on BCG, Dealroom, PitchBook and Crunchbase data, put deep tech at 36% of total global VC funding. Tracxn counted $177 billion raised by deep tech companies globally in 2025, an 82% jump year on year, with 336 unicorns now in the category.

Europe provides further evidence of this reallocation. The European Deep Tech Report found that deep tech captured a record 28% of all European VC funding in 2023 and 2024, making it the single largest category in European venture, ahead of every other sector. The 2026 edition of the same report shows the share rising again to 32% in 2025, with $20.3 billion invested and the combined enterprise value of European venture-backed deep tech reaching $690 billion, up from $73 billion a decade earlier. The category has also proved unusually resilient. While the broader European technology market remains 54% below its 2021 funding peak, deep tech sits just 4% below its own high.

Much of the recent growth is being driven by strategically important technologies. PitchBook tracked $27.6 billion into robotics across more than a thousand deals in 2025, and a record $49 billion into defense technology, nearly double the prior year. In Europe, Dealroom and the NATO Innovation Fund counted a record $8.7 billion into defense, security and resilience startups in 2025, up 55% year on year and nearly four times the level of five years ago, against 16% growth for the broader European venture market over the same period.

Deep Tech Gains LP Attention

Institutional investors have noticed deep tech’s rise. In Adams Street Partners’ 2026 Global Investor Survey of 100 LPs, Europe overtook North America as the most attractive private markets destination for the first time in the survey’s history, cited by 61% of respondents, with deep technology named explicitly among the reasons. The firm’s 2025 survey found technology and healthcare to be the most favored sector, identified by 47% of LPs.

Moreover, according to Atomico’s State of European Tech 2024, 57% of European LPs performed due diligence on a venture fund with a deep tech focus, making it the second most popular fund category after generalist funds. 

Importantly, this is happening even as institutional investors remain cautious about venture capital overall. Private Equity International’s LP Perspectives 2026 study of 103 institutional investors found most LPs holding their venture commitments steady for the second consecutive year, even though half say their VC holdings are underperforming benchmarks. 

Institutional Portfolios Have Yet to Catch Up

The growth of deep tech within venture capital does not necessarily mean that institutional portfolios have adjusted at the same rate. The figures in the previous section measure capital reaching deep-tech companies, but not how broadly that exposure is distributed across the pension funds, insurers, endowments and other institutions that provide capital to venture funds.

Available evidence suggests that the underlying LP base remains narrow, particularly in Europe. A 2025 report found that European pension funds manage more than €3 trillion in assets but allocate only around 0.12% to venture and growth capital. Because deep tech represents only part of that allocation, the proportion invested in venture-backed science and engineering companies must be smaller still.

The fundraising data point in the same direction. The European Commission’s report on untapped opportunities for European venture capital found that EU pension funds supplied just 5% of the capital raised by European VC funds in 2023. In the United States, pension funds provide more than half of venture-fund capital. The comparison is not a direct measure of deep-tech exposure, but it demonstrates how limited the participation of European institutional investors remains in the primary channel through which such exposure is obtained.

The United Kingdom shows a similar pattern. Pensions for Purpose reports that UK defined-contribution schemes hold approximately 0.5% of their assets in venture capital. This is especially striking given that the UK is Europe’s largest deep-tech ecosystem by funding. A substantial domestic deep-tech market can therefore coexist with very limited participation from domestic pension capital.

The United States is further along, but the available figures still describe venture exposure rather than deep-tech exposure specifically. S&P Global reported that US pension funds awarded $9.25 billion in venture-capital mandates in 2025, an increase of 39% and the highest annual total since 2022. The increase indicates renewed institutional activity, but the data do not reveal how much was directed to deep-tech specialists or ultimately reached deep-tech companies through generalist funds.

Nonetheless, this produces an important asymmetry. Deep tech can account for 36% of global venture funding, and 32% of European venture funding, without representing anything close to the same share of a typical institutional investor’s venture portfolio. Market-level funding figures are shaped by the investors that are active at the margin. However, they do not imply proportionate participation across the full LP universe.

Precise measurement of LP exposure to deep tech is difficult because deep tech is not consistently treated as a standalone reporting category. Publicly available LP portfolio data is typically organized by asset class, strategy, geography, vintage, and fund or manager. Although sector-level data is sometimes available, the classifications are often too broad or inconsistent to isolate deep-tech exposure reliably.

Yet despite deep tech’s growing share of venture-capital investment, many institutions expressing interest in the category have only limited exposure to venture capital overall and often lack the data needed to measure their specific exposure to deep tech.

Why Deep Tech Exposure Remains Limited

The gap between what LPs say they want and what they actually hold appears to be driven primarily by implementation challenges. 

Most LPs access deep tech through venture funds, not direct company investments. That means exposure depends a lot on manager selection and existing fund relationships. LPs tend to back established funds with familiar reputations and long-standing relationships, which makes capital allocation conservative by design. That is a problem when the opportunity set shifts faster than the manager roster. Emerging managers can help address this because they are often closer to overlooked technical talent, new research ecosystems, and less visible company formation, giving LPs access to opportunities before they become widely recognized.

Deep tech also requires a different diligence model. A deep-tech fund must underwrite highly technical concepts that require expert knowledge. As high as 81% of deep-tech entrepreneurs believe investors lack the scientific or engineering expertise to assess the category properly. Among investors who do participate, 79% use external experts, 42% hire PhDs and 37% hire science-degree holders or engineers.

While LPs may agree with the deep-tech thesis, their reporting systems do not show whether a portfolio is exposed to true deep tech companies. Without that look-through, an LP can believe it is aligned with the deep-tech shift while owning only diluted exposure through generalist venture funds.

How LPs Can Keep Up

To assess the state of deep tech exposure, the first step is thorough measurement of existing portfolios. LPs should run a look-through audit of their venture exposure and tag each underlying company by technical depth, sector, stage and primary risk. A useful taxonomy should separate true deep tech from the rest.

The second step is assessing the management. LPs should ask which funds have genuine technical underwriting capacity by reviewing the team’s scientific expertise, adviser network, technical diligence process, reserve strategy, commercialization support and record of helping companies move from lab to market.

It is also important to set expectations correctly. Deep tech should not be benchmarked like enterprise software, for example. Its early milestones are different, and investors often need expert input in the field to make informed decisions.

Backing the Foundations of Growth 

The upside of deep tech is measured in more than financial returns. The strongest companies in the category aim at problems that define economic capacity itself, trying to find ways towards cheaper energy, better healthcare, more resilient infrastructure, stronger security, faster compute and more productive industry.

That is what makes the investment case durable. When technical progress unlocks something better, the value created does not stay confined to one company but can expand the market around it.

For LPs, this is the deeper reason to take the category seriously. Deep tech offers exposure to companies that can generate venture-scale returns while also becoming part of the productive base of the economy. Patient capital, allocated well, can help turn difficult science into useful infrastructure. That is a payoff worth underwriting.

Continue the Conversation

Deep tech is difficult to assess from the outside. 

Distinguishing a company with unsolved science inside it from one that describes itself in those terms takes technical reading, and holding a position through the years it takes for that science to resolve takes patience.

Arcanum Ventures spends time with founders working on challenging obstacles. We publish what we learn through our podcasts, videos and original writing. Join us as a speaker, tune in, or keep pace with what’s moving across tech.

If you are building something difficult and want a second set of experienced eyes, we would love to hear from you.