What Is Deep Tech Investment? A Researcher's Guide

Ask what is deep tech investment and the answers come back from one side of the table. Consultancies, banks and venture funds define the asset class for the people allocating money into it. That is useful if you run a fund. It is less useful if you run a laboratory, have a result that works, and want to know what an investor would actually be buying.

This guide answers the question from the research side. It defines deep tech investment by its mechanism rather than its vocabulary, sets out the data that made it a separate asset class, explains why the money arrives on a different schedule from software, and ends with what turns a published result into something a fund can underwrite.

Deep tech investment, defined by its mechanism

The source of the advantage, not the sophistication

Deep tech is a classification of startup company, not of technology. The encyclopaedic definition is a company whose objective is to provide solutions based on substantial scientific or engineering challenges, requiring lengthy research and development and large capital before commercialisation. The term has been around for decades, when it described the research divisions of defence and telecommunications corporations, and it now covers venture-backed companies and awardees of the United States SBIR programme, which provides 2.5 billion dollars a year to small companies commercialising disruptive technology.

The investor's working version is sharper. Celesta Capital describes deep tech companies as those that turn breakthrough science or engineering into commercially scalable products, where the breakthrough sits at the core of why the product performs better, costs less, or enables something that previously was not possible. The test is where the value is created. If value depends on application-layer software, distribution or workflow packaging, it may be a fine business, but it is not deep tech.

The contrast class has a name too. McKinsey calls it traditional or regular tech: technologies based on existing, well-understood and widely implemented scientific principles. A smartphone app built on settled engineering is regular tech however polished it is. A photonic chip is deep tech however rough the prototype looks.

Technical risk first, market risk second

Deep tech investment, then, is the funding of companies whose primary risk is technical. The same encyclopaedia entry makes the point directly: market risk is often significantly lower, because the potential value of the solution to society is clear, while the open question is whether the thing can be made to work reliably and at cost. Celesta puts it as three early differences. Risks begin as technical, with the first questions being whether it will work and whether it will keep working. Time is spent proving rather than pitching, measured in tests, pilots and repeatability. And capital is sequenced differently, with higher up-front research spend before product-market fit.

For a researcher this is the single most important sentence in the article. An investor in deep tech is not buying your market analysis. They are buying a sequence of experiments that retire technical uncertainty, and they price each round on how much uncertainty the last one removed.

Why investors treat it as a separate asset class

Returns: the IRR data

The reason a researcher can raise money at all is that the asset class has a track record. McKinsey, citing Preqin's self-reported fund data, reports that deep tech investments have delivered a net average internal rate of return of 16 percent after fees since 2003, against 10 percent for traditional technology funds. BCG's own analysis, which weights by fund size, finds almost no difference in internal rate of return between deep tech and traditional venture funds, and little difference in how they exit: 51 percent of deep tech investments cashed out through corporate acquisition against 53 percent for traditional venture, 31 percent through IPO against 36 percent, and 3 percent through private equity buyout against 7 percent.

The two reports disagree on the size of the premium and agree on the thing that matters: the long timeline does not cost investors return. Deeptech.build adds the portfolio argument, that deep tech has a low correlation with public software markets, so the decorrelation itself is a source of value for a diversified fund. Its description of the 2022 downturn is the plainest evidence: European deep tech enterprise value sits 4 percent below its 2021 peak while the broader European technology market remains 54 percent below it.

A stable share of venture funding

BCG tracked deep tech at a stable 20 percent of global venture funding since 2019, up from about 10 percent a decade earlier, with billion-dollar commitments no longer uncommon. In Europe the shift is steeper. Dealroom's 2026 report puts deep tech at 32 percent of all European venture funding in 2025, an all-time high and more than double its 15 percent share in 2015. The same report records United States deep tech funding at 144 billion dollars in 2025, up 92 percent, with AI model makers accounting for over half of that total, against 20 billion dollars in Europe.

The practical reading for a research team: a fifth to a third of the venture market now calls itself deep tech. The label no longer filters anything. What filters is the evidence behind the label, which is the subject of the second half of this article.

How the money behaves differently from software VC

Longer to Series A, stronger after it

HSBC Innovation Banking, drawing on Dealroom data, reports that European deep tech startups take two to three times longer to reach Series A than their software-as-a-service peers, but once they do they are more likely to secure follow-on funding. Deeptech.build gives the full horizon as seven to ten years before returns materialise, with heavier capital expenditure on labs, prototype fabrication and regulatory certification along the way.

That shape, slow then durable, is the opposite of a software curve, and it explains most of the friction a researcher meets with generalist investors. A fund with a five-year horizon cannot hold a seven-year asset whatever it thinks of the science.

Two valleys of death

The Global Deep Tech Report 2026, published by High-Tech Gründerfonds, names the two places where deep tech companies run out of money. The first valley runs from discovery to proof of concept, where public grants are too limited and slow and venture investors demand technical proof points not yet available at the pre-seed and seed stage. The second runs from pilot to commercial deployment, around Series B and C. The report's conclusion is that no single capital source bridges both, and that hybrid models combining equity, debt, grants and revenue-based financing are the structures that do.

The same report quantifies how steep the funnel is. Of angel-backed deep tech companies, 93 percent go on to raise a seed round, half of those raise a Series A, and only 7 percent of Series B companies reach Series C. In Europe the Series A to Series B conversion rate is 10 percent, against 24 percent in North America.

Average versus median: where the capital really sits

One pair of numbers from that report should recalibrate any researcher's expectations. The average funding raised per deep tech startup is 39.9 million dollars. The median is 2.1 million. Fifty-three percent of deep tech startups have raised under 5 million dollars in total. Deep tech is a winner-takes-most market in which a small number of companies absorb most of the capital, and the typical company is small, early and still proving.

Which means the question for a lab team is rarely how to raise 40 million. It is how to raise the first two, and from whom.

What makes a research result investable

The proof-to-scale arc

Celesta lays out the arc every deep tech company walks. Proof demonstrates the core effect in controlled conditions. Prototype shows repeatability and defines failure modes. Pilot proves it works in a real environment with real constraints for real customers. Scale-up makes it manufacturable, serviceable and reliable. Expansion turns early wins into a repeatable deployment model with durable unit economics. A published paper usually sits at proof. Investors fund the transition to the next stage, not the stage you are on.

The questions an investor will ask

The same playbook gives the diligence questions, and they are worth preparing for in writing before any meeting. What is the single biggest technical unknown right now. What evidence would eliminate it. What does it cost, in time and money, to produce that evidence. What new risks appear once this one is solved. Which customers believe in this product today. Celesta's phrase for the discipline is that good deep tech investors fund milestones and not narratives.

McKinsey's profile of who tends to answer those questions well is relevant to a university team. Deep tech founders average five to seven years of higher education against two to five for traditional tech founders, because the knowledge barriers are high and the pool of qualified founders is small. The scientific depth that feels like a disadvantage in a pitch competition is the asset the category is built on.

IP as a portfolio, not a certificate

Patents matter, but Celesta is explicit that deep tech defensibility is a portfolio: patents plus process know-how, engineering discipline and systems integration, validation evidence earned in operating conditions, and the manufacturing, qualification and ecosystem position that competitors cannot shortcut. For hardware-anchored companies it treats manufacturability as a first-class risk, meaning yield, supply chain dependencies and reliability under operating conditions get pressure-tested in diligence, not after it.

The steps between a result and that portfolio, including the patent bill a university will hand back to a spinout and the grants that fund the gap, are mapped in our guide to research to commercialization. How a venture fund frames the same questions from its side is covered in our piece on the deeptech investment thesis.

The capital stack is hybrid, and that is the point

Deep tech companies, the encyclopaedia entry notes, are forced to abandon the established funding progression from friends and family through angel, seed and Series A to trade sale or IPO, and many are funded by governments and other non-dilutive grants. The Global Deep Tech Report 2026 turns that from a weakness into the defining feature of the category. Deep tech requires more absolute capital than software, but materially greater access to grants, soft loans and public co-investment reduces founder and investor dilution, reshaping the capital path rather than making it uniformly less efficient.

The public side of that stack is large and budgeted. Deeptech.build lists the European Innovation Council's 1.4 billion euro commitment to deep tech for 2026 and Germany's DeepTech and Climate Fonds at up to 1 billion euros, among others. Which instruments fit which stage, and how grants, equity and debt sit together on one cap table, is set out in EX Venture's guide to EU funding for deep tech startups.

A worked example of the non-dilutive layer at its smallest scale comes from a partner of this summit. The EnerGaia Institute secured 11 approvals from the EUIPO SME Fund to expand its intellectual property portfolio, working with EX-IX. Protected IP is the first thing a deep tech investor looks for, and the fund that paid for it took no equity.

This is also why the EX Future Summit runs a dedicated Global Financial Raise track that maps research and development funding across venture capital, private equity, national grants and Horizon Europe, alongside eight hosted meet-and-greet sessions that match universities with private partners across continents. The matchmaking exists because the first valley of death is a matching problem as much as a money problem: the grant and the investor each want the other to go first.

For the current year's numbers on where that capital is flowing, by sector and by region, see our review of deep tech investment trends.

Where deep tech deals get done

BCG lists the funder types that now treat deep tech as mainstream: corporate venture, venture capital, sovereign wealth and private equity. The investors in deeptech and climate that this summit is built for face a different constraint: deal flow concentrates in the same few cities and the same few rooms, and technical claims are hard to judge without the researchers present.

The summit is designed around that constraint. It runs 18 to 20 November 2026 across three rooms at once, online, Las Palmas in the Canary Islands, and Bali, as a single continuous thirty-hour broadcast between two islands twelve hours apart, produced by EX Venture Inc. with Seraph SL and Equation Labs SL. Online attendance is free for verified researchers and students. The in-person rooms are ticketed. It is a working room where research, capital and policy meet, which for a team at the proof stage is the only kind of room worth the trip.

FAQ

Is deep tech a good investment?

The fund-level data says the long timeline does not cost return. Preqin data cited by McKinsey shows deep tech funds at a 16 percent net IRR against 10 percent for traditional tech, and BCG finds near parity in both IRR and exit routes. The honest caveat is that the distribution of outcomes is wider and the floor is lower, so the asset class rewards specialist diligence rather than index exposure.

What is the difference between deep tech and hard tech?

Hard tech is a subset of deep tech. It refers to companies whose products are tangible and require sophisticated engineering and manufacturing. A novel AI architecture or a computational drug discovery platform can be deep tech without being hard tech, because the breakthrough is scientific but the product is not physical.

What are examples of deep tech sectors?

Celesta's list is semiconductors and compute systems, AI infrastructure, robotics and autonomy, advanced materials and manufacturing, biotech, and energy and industrial decarbonisation. The Global Deep Tech Report 2026 counts novel AI as the largest segment, with 3,900 startups averaging 44 million dollars in funding, and quantum technology as the most capital intensive, with around 200 startups averaging 56 million dollars each.

Can a university research team raise deep tech investment without a company?

Rarely as equity. The first money for a result still inside a university is usually a grant, the technology transfer office's own proof-of-concept fund, or a non-dilutive IP scheme such as the EUIPO SME Fund. Equity investors underwrite a de-risking roadmap, and a roadmap needs a company to own the IP and spend the money. The usual sequence is spinout, non-dilutive funding to reach a prototype, then a pre-seed investor, which is the sequence the summit's university matchmaking sessions are built to start.