One of the hottest areas of venture investment right now is deep tech. New and legacy firms alike are positioning themselves to have a deep tech practice and the pace of funding shows no signs of slowing down. Here are some rapid-fire facts supporting the influx:

  • Deep tech investing represents ~20% of venture funding, up from ~10% a decade ago (BCG)
  • Weighted average IRR was 21% for traditional VC investors vs 26% for deep tech funds (BCG)
  • 7-10% of venture (deep tech) has been responsible for 10-30% of the unicorns (Jordan Nel)

I believe the intentions are good and that a readjustment of venture back to the original ethos and purpose of the asset class is the medicine we all need. That said, I don’t believe the vast majority of these new entrants have a coherent definition of what a deep tech company is. The most common definition I’ve heard is that deep tech companies are overweight technical risk instead of market risk. The repeated adage is “If we build it, they will come.” Here’s BCG’s take from their deep tech investor guide:

Deep technologies are the technologies of the future – the solutions to major global challenges, such as climate change, food shortages, and disease.

I don’t mean to pick on BCG, but what does this even mean?! The problem is that this is a loose definition that fails to capture an actual investable category. B2B SaaS, consumer, fintech, etc (whatever your opinions on their prospects) are all concrete terms for verticals. A layman knows what a company in one of these categories may do. But deep tech doesn’t have that.

Is it fair to say that all deep tech companies are science projects? Is an overly ambitious idea, like a space elevator, considered a deep tech company? Are deep tech companies synonymous with high capex? The answer to all of these questions (and many more) is no. Yet without a clear definition, strategies are being mislabeled and capital is flowing where it shouldn’t.

The ambiguity has led to the bastardization of deep tech and frankly has provided investors with an excuse to invest in uninvestable categories. In this piece, I’m going to look at what it means to be a deep tech company, address the biggest traps I see, and challenge many of the fallacies that deep tech tourists are leaning on.

RapBattle

I’ve written previously about what investment categories excite me. I labeled them broadly as The Next First Leg, or the technologies that will enable the next great platform shift – think how the mobile revolution was enabled by cloud infrastructure, performant hardware, and the like. This idea of enabling technology feeds into what deep tech is, but it too isn’t fully representative. In addition to a technical/scientific or engineering innovation, deep tech companies, like any other good business, must also provide a business unlock via a new capability or capacity and supply existing (or future) market demand.

These are all very primitive, surface-level ideas, yet I am still deeply urged to reassess from first-order thinking. All three must be true for a venture capitalist to consider investing in a deep tech company. Each one of these requirements has nuance and should not be viewed as binary (have or have not), but as scales that plug into a fairly subjective investment decision.

Technical innovation

The first requirement for a deep tech company is that there must be a breakthrough along the dimension of technology/science or engineering. The first two can be thought of as fairly synonymous, whereby some discovery in a research lab results in a new understanding of how the world works, creating valuable IP to be commercialized. On the other hand is engineering, where a new way of creating or integrating known technologies results in a vastly superior system. In his recent piece, Brian Potter does a phenomenal job bifurcating these two in the context of nuclear fusion.

The trap here is creating technologies that ultimately become a solution in search of a problem. I see this all too often when reviewing opportunities where a company has a novel solution (a new material, process, etc) that just can’t seem to find an application, so they make one up.

An example we can all relate to here is the new TSA screeners from Analogic. These new devices utilize CT X-ray scanners to produce 3D images for a more thorough analysis of bag contents over older 2D X-rays. TSA recently announced a $1.3B investment in these new scanners for deployment around the US. While it makes sense on paper, these have been a botched job, significantly decreasing throughput while increasing wait times and secondary screenings.

In this case, Analogic doesn’t care as they’re reaping the benefits of massive contracts while the negative results are passed onto travelers (perhaps a reason to look at this as a positive venture outcome). But what we see here is a new and improved technical approach that should theoretically enhance airport screening yet fail to meet expectations.

Scientists from top labs can often find themselves stuck in these situations whereby they gain attachment to their research and attempt to spin out and start a deep tech company without fully grasping how to commercialize the product or exploit an inefficiency in a market. Shiny new tech can be exciting, but it doesn’t mean it’ll be a good investment.

Now I’m sitting here in my ivory tower sharing what ought to be, and I recognize that. This is not a jab at the builders or founders who dedicate years of their lives trying to sell a product. It’s instead a critique of the way the market operates and overfunds science projects that are not fit to become a for-profit business. This creates a warped perception of venture capital as a limitless pool to draw from to break free from the shackles of academia and continue to operate as a quasi-commercial research organization.

Venture capital isn’t meant to do this – it’s meant to rapidly bring new solutions for giant problems to the market. If we want to avoid the death spiral of 2020/2021 playing out again in deep tech, it’s up to investors to only fund technologies that have a feasible chance of making it to market. The burden rests on the underwriter.

Business unlock

I think we’ve lost the plot a bit on venture in general. I firmly believe that all venture investors should have financial experience (a surprisingly controversial take). The reason is that the goal of an investment is to receive an outsized return on investment. This doesn’t mean the person has to have done their time on Wall Street, but they should at least understand that a business’s present value is the discounted value of future cash flows. They should conceivably believe there’s a path to the sales or EBITDA required to achieve a fund-returning outcome in the public markets. They should understand the levers that can be pulled (both operationally and via financial engineering) to modulate the business in the desired direction. It’s all table stakes, albeit can be learned (though so can many other things).

This preamble tees up my second point that deep tech startups must create business value. It’s silly to write this out, but I think investors can easily get lost in the narrative of a “technological platform shift” without really knowing where value will accrue. David Cahn from Sequoia’s recent follow-up piece AI’s $600B Question frames this problem in the context of AI infrastructure capex (although I do think it’s too early to tell here).

Deep tech businesses should either create multiple expansions, often by enabling new classes of profitable businesses (the rising tide that lifts all boats is the argument for today’s high AI stock multiples) or improve earnings via top-line growth and/or margin expansion.

Multiple expansion can be best categorized as a Next First Leg business – AI is the canonical example of this. However, this is where I find most of the slip-ups in venture today. Too many investors believe that too many new technologies will result in unbridled multiple expansion. If only we could colonize the stars, create AGI, or produce limitless energy, then we’d achieve Science Victory and we all would enjoy a maximalist egalitarian society. Going after this goal is noble, but it’s not realistic. This cycle of throwing money at end goals fails to capture the near-term opportunities that could return a fund. Believing that an end goal exists is great as a thesis’ starting point, but deep tech investors should work backward to determine what inputs are needed when to get there.

Our job as venture investors – as stewards of financial capital with fiduciary duties – is to generate distributions for the people who trust us to manage their money. There should be a path to exiting a position within the 7-10-year time horizon of an investment. This means that it should generate the financial metrics to stand on its own post-IPO (yes, I know companies stay private forever, so include secondaries in this). An investor should always be able to underwrite this outcome, regardless of how early they invest. Investing in The Next First Leg can be on a realistic time horizon – it doesn’t have to be the Last, Next First Leg.

This gets to the next point, that a deep tech company should then improve earnings for other companies. Every sale should answer the customer’s question: “Does this investment give me more money, productivity, etc?” In fact, this alone could be an argument for venture investors having some experience in M&A, as all acquisitions must generate an accretive outcome, but I digress. There are two ways this happens, either by offering a new capability or by offering the capacity to do more with less.

A new capability is something like hypersonics or a novel drug delivery mechanism. Both of these examples provide clear value that a business would pay for. Hypersonics could enable the rapid delivery of medical isotopes while a new drug delivery system could enable a whole new class of therapeutics. In each case, other businesses are given a new capability to provide value to other stakeholders/customers that was impossible before. These are good deep tech companies.

On the other hand, we have companies that create new capacity (or make something a better way). Some examples here would be enterprise AI platforms (no technical innovation) or metal refineries. AI creates the intangible capacity of productivity while a new refinery may allow for the production of critical metals faster and cheaper. In any case, there’s a familiar input and output with a differentiated process in between. The caveat is that the value generated by the process improvements must make up for the capital spent on improving said process. Incremental changes in the right direction are not enough to justify costs. Similarly, non-financial improvements aren’t sustainable in free markets. A new process that cuts carbon emissions won’t be used if those benefits can’t be traded in for credits, or if regulation dictates that not using said new technology will increase the tax burden. It’s always about the money.

In either case, high capex doesn’t necessarily need to be part of the equation, but it also shouldn’t be a deterrent. A clear return on invested capital should warrant massive (or none) upfront capital requirements. In practice, it’s no different than blitzscaling a consumer app with high marketing spend. If dollars-out is greater than dollars-in above an internal hurdle rate, then the company makes sense to fund.

Whether a technology creates a new capability or improves the capacity to do something familiar, both can lead to multiple expansions or an improvement in earnings. And we shouldn’t stray from the ambitious, techno-optimist goals that create all of the headlines. There must be a disciplined, thesis-driven approach to achieving those endpoints though. The Next First Leg is just this – a pragmatic view of the cash flow generating businesses, built using science/technology, that will help us achieve our grand visions of the future.

Market demand

Finally, there has to be demand from the market for the business to exist, which parlays well with the above. I love this incredibly distilled investment thesis from Citrini:

There’s going to be more demand than supply and the only companies that can increase supply are these ones.

That’s it, that’s how you should think about markets when investing in public or private markets. I think there’s a perversion of markets in venture capital whereby too much time is spent conducting top-down or bottoms-up analyses which anchors oneself to misleading statistics and ultimately fails to capture the essence of where dollars are coming from and who’s getting displaced.

The great cliche is to cite 2025’s NDAA allocating $850B for defense spending. Dear defense tech founders and aspiring investors, this is neither your market size nor your opportunity set. You will be lucky to receive a sliver of that. Even the messiah of venture-scalable defense tech himself has reservations about the prospects of defense tech moving forward.

This point on market demand tends to correlate with creating a business unlock via new capability or capacity. A new capability offers the highest chance of a new market being created altogether. SpaceX’s reusable Falcon 9 dramatically cut the price to put something in orbit. Before that, commercial satellite constellations were reserved only for the most well-funded telecom companies. A graveyard of failed satellite Internet companies precedes the explosion of new entrants built atop the Falcon 9 platform. SpaceX offered launch as a new capability that birthed a completely new category and market.

Companies that enable greater capacity often serve a gigantic, existing market. CoreWeave is an example of a company that’s increasing data center capacity so other companies can use compute – they’re slotting new supply into a highly in-demand, commodity product. By increasing the capacity of the bottlenecked resource, CoreWeave also enables the market to grow as customers now have the access required to build their business. Both examples show that additional capability or capacity must service (or create) a considerable market demand.

It’s worth noting that the market doesn’t have to be obvious at inception. Many deep tech companies are the enablers of new markets – the Next First Legs. Therefore there’s further support for the fact that routine market sizing exercises are fraught with traps because they don’t account for new categories emerging altogether. Deep tech investments at seed often suggest capped markets instead of the massive opportunity expansion that they can generate. Market sizing is tricky in deep tech.

Diss

Deep tech isn’t just a buzzword – it’s a critical category of venture investment that drives real innovation and value creation. But we need to be clear-eyed about what actually qualifies as deep tech, and ruthless in our assessment of potential investments. True deep tech companies leverage genuine technical breakthroughs to unlock new business capabilities or dramatically improve capacity, all while serving a large, existing/emergent market. As investors, we can’t get lost in the hype or grand visions of technological utopia. Our job is to find the pragmatic stepping stones – whether the Next First Legs or not – that will generate real returns while pushing us toward those ambitious long-term goals.