Does deep-tech actually make money?
The rhetoric amongst founders, investors and accelerators has changed recently. Consumer facing start-ups and dubious implementations of ‘AI’ (i.e. mostly simple off-the -shelf algorithms) has gone off the boil and investors and founders are increasingly looking towards deep-tech.
This isn’t surprising as there has been a proliferation of big exits, major raises and JVs involving deep-tech companies (Rigetti — quantum computers, Galvani —neural interfaces , Cruise — self driving cars, Graphcore — ML chips, Indigo Agri— bacteria to optimise yield, Twist — custom DNA, Bright Agrotech — vertical farming, etc.). This seems to have brought the potential of real, protectable, monopoly forming technology, paired with a smart business model, back into the mind of the wider investment community. At Deep Science Ventures we’re obviously long on deep-tech but I thought it would be useful to explore why exactly this is the case. In this article we explore whether financially it makes sense to start or invest in deep-tech companies.
What is deep-tech?
The phrase ‘deep-tech’ is definitely still solidifying around a norm. From our perspective we are referring to ventures with a unique and highly protectable platform technology at their core. Traditionally this is broken up into Biotech, Cleantech, Medtech, Semiconductors, Materials and Computing (advanced applications such as distributed systems and quantum), however these boundaries are increasingly blurred with areas like synthetic biology and quantum covering several traditional areas.
Does deep-tech actually make money?
Name the UK’s unicorns, go. Transferwise, Funding Circle.. umm… Sadly the fact that many of the UKs largest tech companies are in fact proper deep-tech companies seems to be almost entirely absent from the wider founder and investor latent knowledge. Just a few of the deep-tech unicorns include; ARM (semiconductors £23.4bn) Autonomy (bayesian learning, £7.1bn), CSR (semiconductors, £1.5bn), Cambridge Antibody Technologies (biotech, £702m), Benevolent AI (pharma, £700m), Improbable (distributed computing, £700m) and Oxford Nanopore (genomics, £1.2bn).
So why are founders and investors so cautious about deep-tech? Principally because there is a perception that it is too risky due to long time scales and the unknown nature of technical risk. Neither of these are particularly justified (the average time to exit of a social tech company is now longer than biotech for example). Likewise, just because deep-tech has seen some large exits doesn’t make it good investment territory. So let’s break up deep-tech and look at what, historically has made money (excluding computation which is well covered elsewhere), and why we think deep-tech is becoming an increasingly attractive area to build and invest.
The four deep-tech learnings so far*
* past results aren’t a predictor of future success 😉
key: 😍 = good times, 😰 = bad times
Semiconductors are what kicked off ‘silicon’ valley in the first place (fantastic history lesson from Lux Capital here). Intel grew from hacking in a garage to challenging IBM. PhDs with semiconductor skills were as valuable as machine learning PhDs are now. Then everyone left the party, what happened?
Principally, keeping up with Moore’s law became harder and harder over time, requiring exponentially more investment for incremental gains in speed, which in turn lead to slow adoption rates and reductions in margins that only incumbents could sustain. Then the possibility of designing chips using CAD without owning a facility (fabless) emerged and this spurred Broadcom, Qualcomm and ARM, but again differences became incremental and start-ups had little scope to compete.
The investor group memory that ‘semiconductors are hard’ has persevered since, few would even give a deck the time of day until last year. A couple of years ago this took a surprising change of direction (comical illustrative graph below from TechCrunch). Why? Cruise sold to GM, Nervana just sold to Intel for $400m and NVIDIA stock went through the roof.
The new wave of companies (including UK company Graphcore) are focused on picking off specific parts of application stacks such as computer vision or machine learning and doing it incredibly fast and at low power.
Semiconductor funding over the last 5 years
Do semiconductors make money as a venture capital class? Hard to say. Just like early web it went through a bubble. For example Broadcom’s market capitalisation peaked at $60b on revenue of $1.1b in 2000 before collapsing and settling back out to now around $30b. If you got out at the right time you did well. The major criticism is still cost. At around $1m and 18 months for initial design and $2–3m for prototype production it isn’t cheap. More worryingly there are often only a couple of potential customers and this can be a precarious position (ask any company that used to have Apple as a customer). Overall though if you can find something that gives you a significant advantage and defensible head-start whilst keeping prototyping to ~$10m or $20m, then, at least at the current time, it definitely seems like there’s a market there and semiconductors are still able to make venture returns.
Who doesn’t love Phages?
The stats for biotech are impressive. Whilst it only represents 11% of VC financing, 54% of VC backed IPOs are biotech companies. On a net IRR basis it outperforms the VC asset class as a whole at 27% vs 22% with the rate of major wins (i.e. 5x or greater return) 50% higher than software and median time to exit 8 years vs. 9 for the wider sector. Moreover it’s oddly predictable with a much flatter power law distribution (All of these stats from the excellent LifeScience VC blog).
There are of course still significant unmet medical needs (if anything we’re just at the beginning of a revolution in advanced therapies) and the buyers, Pharma, are increasingly outsourcing research and feeling the pressure of impending patent cliffs. Moreover biotech is now finally beginning to make sense in other areas from materials and chemicals to agritech and food potentially causing enormous shifts across entire industries beyond pharma. As such if you have access to sufficient expertise to understand the nuances of the science and the field biotech can certainly be a rewarding area.
Medtech 😰 -> 😍
Unfortunately, the same can’t be said for Medtech. In a study of over 170 medtech ventures founded between 2003 and 2006, just 12 are left today with an average IRR of 1%! Although the situation has improved recently as more incumbents outsource innovation. Why is this so different to biotech? Because most medtech interventions are incremental (slightly better knee, tool, pipe, wound care, time to receive diagnostics results) and because of that slow to be adopted, slow and expensive to get through regulatory requirements, quickly surpassed and expensive to sell into highly distributed, complex markets. To further echo this just 6% of all healthcare funding goes into devices whilst 70% goes into treating chronic diseases. It’s the worst of all worlds.
The exception is those companies that sit on the edge of physical devices and therapies such as Medtronic (behind a large chunk of the pacemaker market amongst many other devices) and BTG which doubled sales last year with clever devices such as a way of killing tumors by freezing them. Likewise neuro-devices including non-drug alternatives to address pain (e.g. Google’s Galvani Bioelectronics) and provide better interfaces in cases of sensory loss, represent an exciting area due to the potential to own almost entirely unaddressed markets at a fraction of the cost of traditional drug development.
Digital health 😍
Digital Health is now the principal focus within medtech, i.e non-therapeutic based healthcare, with low cost scalable solutions that can quickly generate tangible costs savings, easy updates and relatively little regulation. Obviously, not all of digital-health is particularly deep-tech with a range of companies from life-extension and genomics to health insurance and patient records, and it has to be said that the biggest deals have been in the less tech-driven companies. It’s too early to know what returns look like on this sector but VCs certainly seem to have confidence with over $3.5bn invested over 188 companies so far this year.
An increasing number of $100m+ deals as the sector matures but few exits to date. Image from RockHealth.
Cleantech has had a rough ride, $25bn went in and over half was lost. Investment peaked in 2008 but by 2012 it was dying with less and less VC money and less LPs willing to commit to dedicated fund. As an asset class it returned just 1.6% IRR (vs. 19% across the rest of the class over the same period) on early stage deals.
Why so rough? Principally due to long sales cycles, small margins in commodity markets, high capital cost, customer skepticism (i.e. if it’s just clean but doesn’t save money it’s hard sell) and switching customers from an opex to a capex upfront spending model. And of course we shouldn’t forget China’s extensive subsidies on cleantech manufacturing whilst nearly every other government withdrew tariff subsidies. The challenges are even more evident in the split across the company types with renewable power development (e.g. deploying solar farms) doing the best whilst technology manufacture struggled the most. The real problem is that customers are few and far between and highly incentivised not to change (energy companies in most cases), IPO markets are still traumatised and infrastructure developers prefer more typical deals (e.g. roads). This leaves VCs to fund expensive manufacturing and deployment which is something that isn’t really suited to equity based funding.
Invested vs. returns. Ouch. Graph from MIT’s Energy report.
Where does this leave those of us wanting to save the world? In addition to the renewed governmental focus on clean power and the emergence of new financial instruments to cover the VC-infrastructure gap there’s likely a new cleantech wave on the way. This time it will look very different from wind turbines and slightly more efficient solar cells. Instead advancements in distributed computing and automation coupled with more efficient manufacturing and modularisation will open up opportunities for significantly reducing power and resource demand in a way that directly benefits a very wide potential customer base.
Advanced Materials and Chemicals 😰
Materials and Chemicals crosses over with many other areas including cleantech / energy (batteries, solar cells, and greening various chemical processes etc.), semiconductors (transistors, diodes etc.), agritech (fertiliser production uses 1% of the world’s energy!) and of course biotech. However it’s worth addressing separately because as chemistry becomes digitised and synthetic biology allows us to turn life into factories there is increasing entrepreneurial activity in this area.
There are undoubtedly huge opportunities in this area, if you can solve the challenge of producing fertliser with far less energy, make bioproduction cost-effective (nod to portfolio company uFraction8), or produce materials that trap ions in a sufficiently isolated way to scale quantum computers for example. However, where most efforts fall down is the clean-tech trap described above, i.e. in attempting to green processes without fully considering the incredibly entrenched processes, supply chains and performance risk to potential customers. Sadly, it rarely makes sense to switch.
Despite this, regulation is becoming an increasing pressure across many chemical and material driven industries and this represents an opportunity, particularly for those companies taking a more data driven approaches such as that deployed by portfolio company Materialize.X. So what was once an ignored sustainability issue is increasingly business critical. This LEIF report goes into sector by sector drivers if of interest.
However, we remain bullish on materials for reasons better summarised in the quote from Pangea Ventures below.
Breakthroughs in advanced materials are becoming increasingly important for companies to excel in almost any market. Advanced materials are solving fundamental problems necessary to make products more efficient, sustainable, less expensive, and better performing, key attributes necessary for widespread adoption of any product. Keith Gillard, General Partner, Pangaea Ventures. Taken from the London Environmental Investment Forum report.