Deep Tech Awakening and How to Predict its Success: Part II

In Part I, we discussed the rise of Deep Tech and the factors that propel the category to the forefront of venture investment. We looked at the macro-drivers that create demands for breakthrough technology to solve real-world problems that badly needed solving. We discussed how the many principles used to assess software companies do not necessarily translate into the Deep Tech paradigm. We ended with a question on how to pick successful Deep Tech companies.

Before we explore the answer to our final question in Part I of this series, let’s examine one common premise against Deep Tech. 

Critics of Deep Tech often claim that Deep Tech investment possesses such a high technology risk, the return is often unjustifiable. This premise fails to distinguish between the shade of grey within the maturity curve of each emerging technology. 

By understanding this particular nuance, we stand a significantly greater chance in predicting the success of a Deep Tech company. 

The common misconception of technology risk

report by Dealroom defined Deep Tech as follows:

“For a startup to earn the “deep tech” label, there must be science or engineering risk in getting the idea to actually work and, assuming it does, risk in proving market demand for that product. If there is only one of these risks, but not both, then we’re not talking about a “deep tech” startup.”1


While we agree with the first point that any Deep Tech company would naturally have an associated science or engineering risk, we disagree that there needs to be a product-market risk.

If this was the case, Deep Tech as an investment category would fail completely. It is inconceivable that the risk-adjusted return would make any sense for institutional investors. 

In fact, we believe that the best Deep Tech investment requires a high degree of conceivable product-market fit and a predictable level of engineering risk. 

Exponentially growing adoption force

Market size is almost always one of the top three factors investors look at when making investment decisions. But market size does not answer the question of whether there will be a product-market fit.

Contrary to popular belief, customers do not adopt new technologies simply because the problem appears to be big to them. These big problems are known elements. They have grown numb to their significance, treating them as “business as usual.”

The business’s problem must not only be big; it must also be growing—and at an un-ignorable rate in order to create a sense of urgency.

Secular trends are the threads that link these exponentially growing problems. 

The shift from retail to eCommerce, and the effect of labor shortages driven by demographic shifts, will only increase the need for warehouse automation. Left unaddressed, businesses will have to combat customer complaints of delayed or incorrect deliveries.

In other words, there is no other viable option to cope with increasing demand while having insufficient labor. The other alternative—seeing margins dwindling exponentially—is also not viable for any of the key decision makers’ careers. 

Aligning the stakeholder incentives

Much like the mainframe computer, which was originally used by large organizations for critical applications, most Deep Tech solutions were initially developed with businesses and major corporations as their primary target customer. Unlike consumers, corporations are made up of multiple stakeholders with their own distinct agenda. 

A failure to account for each stakeholder’s role and, more importantly, benefits from adopting novel technologies is a failure to create a product-market fit.

It is not enough to delight users or paying customers with a “10x better” value proposition. Deep Tech companies must align the incentive of all the major stakeholders involved. 

Let’s examine two cases of how this works in real life: 

Construction: A typical construction project involves a few dozen stakeholders. ALICE Technologies understood this well as it searched to find the perfect customer for its AI solution, which automates and optimizes construction scheduling, ensuring that each bridge and skyscraper are built with minimal resources and fastest possible. 

The company initially targeted developers as the customer, but quickly realized that its product is more suitable for general contractors who have the power to shift resources more swiftly and according to changing schedules. 

While the developer benefits from faster completion, the general contractor needed to be the primary user of the technology. Still, the best customer was a transportation general contractor who not only managed the subcontractors, but also owned their own construction equipment. This allowed them to realize cost savings from reduced equipment usage and ownership.

Had ALICE not realized the intricate web of construction stakeholders, the technology might have failed in the hands of developers who were not able to exert direct influence on day-to-day management. 

Healthcare: Perhaps no company did better in incentive alignment than Imvaria, a digital biopsy company that uses AI to analyze MRI images for lung diseases. 

While there are probably several hundred “AI radiology” companies, the brilliance of Imvaria centers around its specific disease target: a progressive and fatal rare disease called Idiopathic Pulmonary Fibrosis (IPF).

The disease severity mandates diagnostics, but current physical biopsy through tissue surgery used to confirm the disease’s existence lends itself to a staggering 15% mortality rate. Neither the physician, the provider, nor the patients want to take the risk of diagnosing for fear of death. Still, insurance companies have every incentive to push for early diagnosis since each IPF patient whose disease has progressed costs the payor six-figures in reimbursement expense per annum. 

Imvaria not only provided 93% diagnostic accuracy, they removed the fear of death for all parties while allowing payors to save on long-term cost of care.

By thoroughly understanding these exponentially growing market forces alongside the stakeholder incentive alignment within the industry ecosystem, we can reasonably comprehend the customer needs and predict whether the solution and business model has a strong chance of attaining product-market fit. 

Then, we can start working backwards to find the most compelling and economically viable technology to address those needs. 

Investing in “Product-Ready” technology, not Moonshot

My partner, James Wang, has already covered this topic extensively in his previous post, Invest in “Product-Ready” technology, not Moonshot.

To summarize his take and elaborate with my own: 

Not all Deep Tech companies are made equal. Some Deep Tech companies, such as biotech, are still in the R&D phase with no way to reasonably predict the viability of their technology. However, there exists a group of Deep Tech companies that are leveraging proven technologies for specific applications under a certain economic structure.

Unlike biotech companies which carry R&D risk, these Deep Tech companies carry what we call engineering risk

Figure 1. Creative’s classification of technology risk

The market often mistaken engineering risk for R&D risk, and savvy investors who are able to see the differences between the two possess the information advantage to generate outsized return.

Let’s bring this idea to life with an example:

Autonomous vehicles are one of the most promising technologies with limitless applications. But we have yet to see a mass roll out of such technology. One of the key challenges is the cost of perception technology, the most expensive of which is LiDAR (a light detecting and ranging sensor that is used to gather depth information which 5 years ago cost anywhere between $2,000 – $20,000 per sensor). 

The technology is not only expensive, but it has not historically been able to perform at an automotive-required range. 

Enter Sense Photonics, a flash-based solid state LiDAR company. The company applies its unique Micro Transfer Printing technology to its design and manufacturing process, which allows for easier and lower-cost assembly of semiconductor devices within the LiDAR unit. Its technology has been proven at scale in the renewable energy sector, and the company is applying to LiDAR. Sense has recently successfully demonstrated a 200-meter detection, typically required for an automotive grade LiDAR at a significantly lower cost. 

The engineering feat Sense accomplished was no small launch, but underneath its success was a technology proven both in its viability and at scale in another field. Such is a distinction of Deep Tech companies that are able to achieve a “10x better and cheaper” outcome while carrying a much lower engineering risk from those with an unproven R&D risk. 

The era of metal unicorn

Contrary to what many might imagine, high potential Deep Tech companies do not typically resemble Star Trek’s transporter that beams people in and out of places. 

High potential Deep Tech companies are attractive because the market demand is both proven and growing exponentially by a known cause. 

Their business model has been designed not only to serve its users and paying customers, but also to benefit other stakeholders. Their technology is proven, involves a calculable risk, and can reasonably be engineered to fit within commercially viable parameters. 

They devour less cash because they do not need endless advertising dollars, can become profitable without sacrificing growth, and are more likely to exit faster through acquisition.

Welcome to the new era of the Metal Unicorn.

Leave a Reply