Deep Tech, Dark Matter: An opportunity to solve bigger problems
Part 1: Dark matter: A missed opportunity for deep tech innovation
Software continues to dominate the startup landscape1. However, many of the world's most pressing problems cannot be solved with software alone. Deep tech innovation in the hard sciences of biology, chemistry, and physics is essential to address urgent problems from pandemics to climate change.
To meet the moment, deep tech innovation needs to move faster.The pace of software development benefits from common languages, an open-source ethos, and systems like GitHub that facilitate knowledge transfer. In contrast, deep tech innovation is rarely standardized. Every team has its own R&D methods and knowledge management systems. Tacit knowledge stays locked inside the heads of key scientists. Worse yet, because this 'dark matter' is unstructured and non-standard, it is lost at key transitions, especially when a company shuts down.
Advances in AI will enable new approaches to deep tech innovation, in which knowledge, including the dark matter of informal IP, is preserved, transferred, and queried to accelerate the pace of discovery and commercialization. Within this opportunity, we have identified a significant well of untapped innovation momentum: harnessing the dark matter from failed startups.
Dark matter, the lifeblood of deep tech
The value of intangible assets, primarily intellectual property (IP), has grown 13-fold since 19962, and now represent as much as 90% of the value of the S&P 5003. Within deep tech, intellectual property represents most of an early-stage company's technical and enterprise value. However, the share of formal IP (patents, trademarks, copyright) and informal IP (trade secrets, know-how) differs considerably across companies and stages.
In our survey of deep tech founders, we found that many of them focused heavily on developing informal intellectual property (IP), generally as trade secrets, as the best way to quickly establish a defensible technology position. Creating formal IP, such as patents, was often seen as an expensive distraction from early technical progress, but a necessity to create a perception of enterprise value for investors. Indeed, one 2023 analysis of almost 300,000 early-stage startups showed that less than a third of deep tech startups had filed a patent at the seed stage4.
This is despite the observations that:
Formal IP can increase the odds of future financing by up to tenfold,
Startups with patents raise 40-60% more capital, and
Have public exits at 5x higher rates and with 155% higher valuations.5
Formal IP has clear financial value for one very simple reason: it's much easier to write a contract that transfers the ownership of a patent than to write a legally binding agreement to transfer know-how.
However, from an intellectual and innovation standpoint, informal IP assets have greater value. Behind every patent are experiments, data sources, and observations that confirm what works and, more importantly, confirm what does not work. Lessons learned from failure don't make it into the patent. As Thomas Edison famously stated in his work on the light bulb:
"I haven't failed, I've just found 10,000 ways that won't work."These lessons, along with the collective know-how, raw data, operating processes, and trade secrets of a company, are the company's 'dark matter'. Much like the dark matter of the universe, the dark matter of innovation contributes a huge, invisible proportion of the enterprise value of a company and what it can produce. Patents alone do not make an innovation. Many patents are simply unusable without the dark matter behind the development of these inventions.
The data supporting dark matter's value is abundant. Looking at 50 years of data from Stanford University's innovation platform, inventions licensed to the patent's inventor were almost 3 times more likely to generate >$1M in royalties6. Of technology that is licensed, 73% ends up at small companies or startups, often spinouts created by the inventors themselves. Self-licensing, or the involvement of inventors, is repeatedly correlated to positive commercial outcomes7. More than half of spinouts use tacit knowledge instead of codified knowledge8, showcasing how essential dark matter and know-how are in successful commercialization.
Dark matter is incredibly valuable, but in today's startup ecosystem, it is not appropriately valued.
Dark matter leaks out during transitions
While dark matter is critical for successful commercialization, because it is difficult to value, it often disappears during major transitions. This includes employee turnover: 35-70% of knowledge is lost when key employees leave, retire, or are laid off9. There are also losses during tech transfers: with over 50,000 unlicensed, high-quality inventions residing in universities, billions of dollars invested in R&D have not been translated into commercial applications, with less than a quarter of IP being licensed10. One driver is that dark matter is rarely captured alongside the invention disclosure used to create formal IP. Even during a merger or acquisition, where the incentives are aligned to maximize the transfer of dark matter, 70-90% of deals fail or underperform, in part because key employees' tacit knowledge is lost; 47% of former employees leave within one year, and 75% leave within three years of an M&A11.
However, there is one transition that is more prone to the loss of dark matter than any other: when a company shuts down. If we believe the common lore that 90% of startups fail, then 90% of our innovation investment — billions of dollars and years of lessons learned in R&D — is at risk of disappearing.
Dark matter is lost in >60% of shutdowns
If dark matter truly is the lifeblood of deep tech, you would assume it would be preserved, learned from, and built upon. However, during a shutdown, the creditors and shareholders prioritize moving quickly and maximizing financial value. The result is a focus on quick asset sales, primarily centered on formal intellectual property. A 2018 study of 285 venture capital-backed startups revealed that 68% of startups sold their patents to new owners after shutting down12. Of the startups we surveyed, ~60% found second homes for their IP. For formal IP, like patents, there appears to be a relatively healthy ecosystem for distressed asset repurchasing.
What happens to the dark matter in a wind-down?
Companies may or may not offer informal IP in a sales process. The decision depends on how the informal IP was organized, who was involved in the wind-down, and the perceived value of the company's technology. One study of startup failures suggests that 87% of shutdowns are likely to result in a significant or complete loss of informal IP 4.
Our research, including the case studies presented in Appendix 2, also supports up to 60% loss of dark matter when a company fails. In 30-40% of deep tech startup closures, even formal IP is lost. As one survey respondent put it:
"The IP is sitting in my closet, unmonetized."
This is a classic collective action problem. Stakeholders optimize for their near-term personal interests and financial value at the expense of the potential for the lessons learned to further the industry and society.
While there is no counterfactual to measure the lost opportunities and slowdowns resulting from the loss of informal IP, we can point to several success stories where the preservation of dark matter has led to future innovation. From Steve Jobs' repurposing of the graphical user interface developed by Xerox's Palo Alto Research Center13 to the revival of the Oak Ridge National Laboratory's molten-salt reactor experiment14, dark matter resurrections have had profound impacts on the very course of history.
Rapamycin: A case for preserving dark matter
One of the most compelling examples of dark matter rescue is the storied history of the drug rapamycin (aka sirolimus), which has delivered anticancer and immunosuppressant benefits, contributed billions in commercial value, and is currently under exploration as a longevity drug to extend human lifespan 15.
Figure from Gambari et al. 2023. The long scientific journey of Sirolimus (Rapamycin)
The story begins with a 1964 research expedition to Rapa Nui (more commonly known as Easter Island), whose people were mysteriously not afflicted by tetanus. The goal was to look for novel antimicrobial compounds conferring their immunity. The team collected hundreds of soil samples, including from locations known to have incredible healing properties in native traditions. In 1969, Ayerst Pharmaceuticals acquired the soil samples. Within was a bacterium called Streptomyces hygroscopicus, which produced a compound that could kill fungi. This compound was named rapamycin after the island. Suren Sehgal, a microbiologist on the project, sent a sample to the U.S. National Cancer Institute for screening and found that it was a new class of anticancer agent.
Despite these promising findings, Ayerst faced challenges formulating rapamycin for drug trials and closed the rapamycin project in 1982. They ordered all materials to be destroyed. Sehgal, convinced of rapamycin's value, preserved a set of vials in defiance of the company's orders. He stored these samples next to tubs of ice cream in his home freezer for five years, labeled "DON'T EAT!!".
Sehgal in his lab at the National Research Council of Canada, Ottawa, circa 1959. Credit: Courtesy Sehgal Family.
In 1988, Wyeth, another pharmaceutical firm, merged with Ayerst. Faced with a new leadership slate, Sehgal put a memo together outlining the incredible potential of rapamycin and was able to resurrect the exploration of the samples' potential. Today, over 100 patents have resulted from rapamycin work, with multiple analogs and dozens of applications in transplant medicine, oncology, and antiaging.
This story highlights the critical importance of dark matter in enabling the journey for these incredible, life-saving drugs to get to market. If Sehgal had changed firms and had not been involved in the new merger, the samples in his freezer would have not just been illicit, but illegal. Had Sehgal not had the instinct and conviction for rapamycin's potential, he might not have preserved those samples. And most importantly, if the indigenous people of Rapa Nui had not shared their oral history, the specific soil samples containing rapamycin may not even have been collected at all.
The 'real’ dark matter. Dark matter filaments (shown in red) bridge the space between galaxies (shown in white) on this false colour map. Credit: University of Waterloo.
Coming up:
Part 2: Why are we losing over half of our innovation investment?
Part 3: How to accelerate innovation by harnessing dark matter
Public information about startup closures, IP transfers, and informal IP is sparse and unreliable. For this piece, we've spoken with over 150 people in the innovation ecosystem including people working at startups, venture capital firms, government grant programs, university tech transfer offices, non-profits, philanthropies, large corporates and more. We ran a survey that captured detailed and specific views from 25 current founders and 5 investors, who collectively have deployed over $3B in innovation capital. We did detailed interviews with 20 people, including founders, investors and philanthropists. Eight founders generously contributed significant information about their companies for our case studies. We analyzed CrunchBase and Pitchbook for trends in deep tech startups. A.I. search engines were used to surface trends and primary sources. We primarily focused on agtech, biomanufacturing, climate tech, and materials science, and secondarily focused on life sciences, space tech, and artificial intelligence.