Four Theses on Deep Tech
Getting back to the future
The last few years have given rise to a renewed interest in deep tech1. Peter Thiel’s punchy “we wanted flying cars, and all we got was 140 characters” has captured technologists’ imaginations. Meanwhile, tech has emerged as an important potential bulwark against climate change, and venture capitalists have responded by raising large funds dedicated to investing in climate tech – in addition to increasingly deploying capital from ordinary funds into the category. The response to COVID and the success of companies like Moderna have cast a spotlight on the incredible progress happening in biotech. Excitement about deep tech has also taken root in the blogosphere, prompting tons of think pieces about American investment in industrial innovation.
On the other hand, the rapid rise and fall of the SPAC market – and its close association with deep tech – might make the whole thing look suspicious. Skeptics might also note that it wasn’t so long ago that we went through a climate tech boom and bust and that a lot of the most visible deep tech promises like autonomous vehicles have been “just a few years away” for… more than just a few years.
As someone who joined a deep tech startup out of college, I’ve been thinking a lot about the sector and how it might develop throughout the rest of the decade. tl;dr: I think a bunch of converging factors suggest entrepreneurs and investors should be pretty bullish on deep tech.
Thesis 1: Hardware Leapfrogs Software
With the tremendous success of software companies over the last ~25 years, it’s easy to forget that the early successes in VC/startups were in what we’d now describe as the eccentric niche of deep tech: semiconductors, biotech, computer hardware, etc. The hardware investments of the 60s and 70s facilitated the rapid growth of the computer market, enabling future entrepreneurs to build great businesses selling software. Now the pendulum seems to be swinging back, and the software developments of the last couple of decades might help unlock innovation in hardware.
The software revolution has made sharing and accessing information much easier. We’ve generated unprecedented amounts of data in uniquely legible forms, and we’ve developed novel tools for using that data effectively. Marc Andreesen famously argued that software was eating the world, but there’s no reason to view these advances as zero-sum: software ate the world, and now deep tech industries have an opportunity to reap the benefits. Manufacturing processes can leverage software to become more standardized and efficient. Biologists have access to increasingly sophisticated software tools that supercharge their research capabilities. Recent breakthroughs in AI might finally make autonomous vehicles viable.
One way to view a prolonged period of software supremacy is as a stretch of hardware regress. Another, perhaps more accurate, perspective, however, is that we have just experienced decades of progress in building the tools and infrastructure to help hardware quickly leap ahead. When you prime a slingshot for launch, it appears to be moving backwards right before it dramatically springs forward.
Thesis 2: Social Convergence on Deep Tech’s Importance
Ezra Klein, Katherine Boyle, Noah Smith, Marc Andreesen, Eli Dourado, and many others have written publicly about a perceived stagnation in non-software industries and have called for concerted efforts to spur innovation, particularly in industrial sectors. Interestingly, as Klein notes, for the last 40 years, “supply-side” concerns were perceived as right-wing, but now the left is adopting the label as well. The trans-partisan support for growing American industrial capacity is definitely bullish for deep tech startups focused on doing just that. Whether for ESG reasons or out of a commitment to American hegemony, there’s a renewed interest in making it easier for innovative companies to operate in the physical world.
This burgeoning ethos is good for deep tech as it offers a narrative through which talent, investors, regulators, and other critical constituencies can think about the importance of deep tech. But, all of the urgent calls to action might obscure what’s actually been happening: Republican and Democratic administrations have already converged on enacting policies supporting investment in deep tech.
Operation Warp Speed is probably the most notable recent government program facilitating the growth of deep tech companies, helping Moderna bring the first product developed with its innovative mRNA technology to market.2 Other elements of the COVID response have similarly unlocked funding for deep tech companies, like the American Rescue Plan’s loan guarantee program for companies working on food supply chain resilience. Before COVID, the 2017 tax bill created the opportunity zone program, incentivizing tangible investments in distressed American communities, and Obama-era Energy Department loan guarantees helped Tesla go on to become one of the most valuable companies in the world. Looking forward, the CHIPS act will support investment in semiconductors and other deep tech areas.
This enthusiasm for deep tech has also spread beyond the United States. Foreign sovereign wealth funds have massive amounts of capital to deploy – and often with mandates to seek more than just pure financial returns. It is not an accident that the first company to commercialize cultivated meat first sold their product in Singapore and received backing from Singapore’s sovereign wealth fund; Singapore, after all, has explicit goals of fostering such technology. All of a sudden, there are many billions of dollars not just open to real technological risk but actively seeking it out – as long as it also aligns with goals that deep tech companies are well-suited to address such as sustainability or supply chain resilience.
That influential figures from New York Times writers to Congress to Singapore’s sovereign wealth fund agree on the importance of investing in deep tech is creating an especially hospitable environment to build new deep tech companies in. These startups will enjoy comparative advantages in government support, investor interest, and talent recruitment and retention.
Speaking of talent…
Thesis 3: Burgeoning Talent Flywheels
Related to Thesis 2, as startups compete for the best talent, deep tech companies will be able to compete on mission. With so many influential figures advocating for and investing in deep tech innovation, exceptional technologists will be increasingly inclined to join exemplary deep tech companies. The ability to win talent will in turn help deep tech companies succeed, and of course more successful companies will in turn attract more talent, creating a virtuous cycle.
Representation is an underrated force in human decision-making. The financial crisis was probably bad for the perception of financiers and kicked off the decline of the popularity of economics as a major for Harvard students. For most of the 2010s, the most visible figures in tech were Mark Zuckerberg and Jeff Bezos – software entrepreneurs par excellence. The release of a movie about Zuckerberg seems to mark the inflection point in the popularity of Computer Science as a college major and the increase in applications to Y Combinator. Now the most famous person in tech (the world?) is Elon Musk, who is more closely identified with his deep tech companies like Tesla and SpaceX than with his software ventures, Zip2 and PayPal.
The power of these virtuous cycles doesn’t just depend on the most famous founders inspiring future technologists. Tech behemoths beget further behemoths (or if not behemoths, at least unicorns) by training talent directly. The PayPal mafia is especially notorious, but Google/Facebook/Amazon/etc all have spawned similar networks. For a while, there simply wasn’t a critical mass of amazing deep tech companies to train talent to go build other deep tech companies, but SpaceX has become a clear counterexample, spurring the creation of many other space companies. The amount of entrepreneurial talent in deep tech has been bubbling and seems ready to really break out in the 2020s.
Thesis 4: Innovative Technology is Not a Substitute for Product-Market Fit
The 2020-2021 SPAC market offered deep tech startups the ability to go public at multi-billion dollar valuations and enjoy greater access to capital to further grow their (often capital intensive) businesses. Usually, companies need meaningful revenue to go public via IPO, but since SPAC targets are uniquely able to market themselves based on future projections, startups during the SPAC boom often didn’t need to have demonstrated actual demand for their products before going public. This feature of the SPAC market enabled companies that had invested in developing innovative technology but had not yet successfully commercialized a product (and therefore had little to no revenue) to go public. This path to public listing suggested a model for building deep tech companies: use venture dollars for R&D, go public with the promise of selling the technology you’ve developed, and then finance the commercialization of your new technology with capital from the public markets.
Unfortunately, the companies that followed this model have generally not done very well post SPAC merger (broadly even worse than the rest of the tech ecosystem, which has also seen dramatic drawdowns in 2022). SPAC IPOs (ie the formation of new SPACs) have also slowed down, meaning there will be far less capital available for SPACs in the coming years. All of this suggests that SPACs will not be a reliable way for pre-product-market-fit deep tech startups to go public in the future.
The inability to follow this path to exit might seem like a problem for building or investing in early-stage deep tech companies. However, there is an alternative model that does not rely on the possibility of exit via SPAC (as the below chart from Cantos illustrates really well): develop innovative technology, create a product with latent demand, and then commercialize and quickly capture market share.3
SpaceX, arguably the paragon of deep tech companies, won a contract worth $1.6 billion just six years after its founding. Anduril, one of the most exciting deep tech startups right now, achieved ~$100 million in annual revenue 3 years after its founding. The keys here seem to be (1) having a great engineering team that can develop a market-ready deep tech product in just a few years and (2) not taking much market risk: deep tech companies should develop products that don’t currently exist on the market and that a huge portion of the market would obviously use if only they existed. Meeting these two planks would enable a deep tech startup to see the kind of step change growth depicted in the graph above. With those caveats, both deep tech and software companies should be able to use venture capital to find product-market fit and build a business, but, for the foreseeable future, it seems deep tech can no longer take the shortcut to going public without meeting those milestones.
The loss of the SPAC market isn’t bad for deep tech startups in general, but it will change the kinds of deep tech business models that can be reliably invested in.
Conclusion:
I’m excited about deep tech. The SPAC boom doesn’t offer a sustainable model for building these kinds of companies, but the combined effects of software advances, social and political interest, and a growing talent base might be enough of a cocktail of tailwinds for era-defining companies to be built.
If you’re interested in any of the topics mentioned above, I’d love to chat. Follow me on Twitter or shoot me an email!
Deep tech is one of those words that can kind of mean anything and therefore sometimes sounds like it means nothing. Fundamentally, deep tech represents innovation in the world of atoms (ie physical matter). Deep tech is best understood in contradistinction to innovation in the world of “bits” (ie software). Social media and enterprise SaaS are not deep tech. Space, robotics, biotech, energy, food, and manufacturing are all deep tech. Pharma is similar to other deep tech sectors in a lot of ways but sometimes functions like its own thing, and something like software for self-driving cars is maybe both software and deep tech. For some concrete examples, see the portfolios of Lux, Cantos, 50 Years, Lowercarbon, and Eclipse.
Though it has definitely received a lot of attention, it is still probably underrated in terms of importance, and for some interesting analysis on the replicability of the model, I would recommend Alex Tarborak's post on Marginal Revolution.
Pharma doesn’t really work like this – pharma companies often IPO without any products on the market, but of course pharma remains a fundable sector. I think this is just a case of pharma sometimes looking like its own third category distinct from deep tech or software. Note that carving out pharma here does not impact how we should think about biotech in general. Synthetic biology companies developing commodity/industrial products seem to have to commercialize on deep tech timelines.