Embracing the Uncertainty in Hard Tech Investing
This post has been stuck in my drafts for some time. With deep tech's newfound prominence in venture investing, investment blueprints from the last decade are becoming obsolete. A generation of SaaS focused VC investors have had to pivot swiftly from software purists to hardware specialists. To back truly idiosyncratic businesses that will define the next generation of progress, one must become comfortable with a different level of risk-taking and uncertainty. This post is my attempt to clarify my approach to this challenge, this new risk equation.
—
TL;DR:
Traditional approaches often fail to account for the complexity and interconnectedness of risks in hard tech.
By focusing on incremental progress, avoiding binary risks, and embracing antifragility, investors can mitigate uncertainty and improve their chances of success.
The key is understanding how each risk layer builds on the next and working closely with founders to address challenges through iterative testing and execution.
—
People are notoriously bad at evaluating multiple risks at once.
We often stack probabilities and calculations without fully linking different risks or understanding how they interact. Risk, at its core, is a measure of the likelihood of an adverse outcome. Yet, we regularly overestimate failure rates and misunderstand how individual outcomes are connected.
Our decision-making is clouded by biases, such as negativity bias, where negative outcomes stick in our minds more vividly, or loss aversion, which makes us fear losses more than we value gains. Experience also plays a significant role. When we lack expertise, we inflate perceived risks. And let's be honest—99% of VCs in deep tech aren't true experts in the space, nor have they been operators in those specific verticals (as a matter of fact, they shouldn’t be). The ramp up in a particular vertical can be nothing more than an afternoon of digging and a few expert conversations when deals are “hot” and moving fast.
And things become even more complex once we look at how those risks are interconnected within deep tech. There’s the linear way of approaching risk: a 30% chance of solving a technical problem, multiplied by a 20% chance of achieving Y commercial outcome yields a certain overall probability for Z outcome. But this fails to account for reflexivity, or how achieving one outcome can alter the probabilities of X, Y, or Z. Each milestone isn’t a simple pass/fail test, but an opportunity integrate new knowledge into our estimation of the likelihood of a certain outcome to happen.
Mike from Also Capital once shared with me that a venture investment should feel like a collection of problems that are “more likely than not” to be solvable. I’ve taken this advice at heart, and it is why I avoid situations with fundamentally binary risks. I prefer to focus instead on scenarios where there isn’t a single point of failure, or risk bottleneck of some sort.
This approach allows us to consider how a nascent company’s path can be antifragile. An antifragile company thrives on incremental progress, growing more resilient as each uncertainty is resolved and new information is integrated. Instead of relying on a single, high-stakes outcome, the company benefits from the randomness and variability of small victories that compound over time.
Choosing the right risks to take
This doesn’t mean that every risk is worth taking in hard tech.
The two types of risks I try to avoid when evaluating deep tech companies are:
Binary risks with low chances of success at their most fundamental level*.
Risks that affect the trial-and-error process itself, reducing the impact of high-throughput testing.
*One mistake is to confuse a low chance of success at a high level (e.g., SpaceX sending satellites to space 100x cheaper than the current status quo) with the series of essential problems that can be solved along the way, such as designing cheaper Merlin engines, enabling rockets to survive re-entry, and achieving autonomous landings.
A risk I often come across is the execution risk - In other words, whether a team knows how to get sh*t done in the form of finding rapid solutions to the problem they face. That’s a fundamental risk in its essence—if founders don’t have a proven track record, or if you can't sense a high-throughput culture (small consideration: How fast do they get back to you?), there’s likely little point in pushing forward with the investment. This will likely lead the team to stagnation, moving them away from the small iterations needed to reach escape velocity. This is a risk I’m not willing to take.
Picking the right risks ultimately hinges on what remains once a specific risk is resolved. When technical risk is high, the go-to-market strategy should feel straightforward once that hurdle is cleared because the customer is ready, awaiting a solution. Demonstrating that solving this risk makes the company's path to market easier—through customer relationships, co-development, or investments from prospective participants and advisors—can help startup founders clarify their vision and make the investment more appealing and understandable for VCs. If this isn’t the case, founders may need to revisit their approach and ensure the right resources are aligned to address those risks.
Example 1: High technical risk → Check if the founding team has the support of respected technical leaders in the field who are interested and willing to back them.
Example 2: Commercial risk → Verify if key leaders at target companies show interest, dedicating time and resources to the startup even when it’s not yet scaled.
Uncertainty vs. Risk
Evaluating these fundamental risks can sometimes be confused with evaluating uncertainty. If there’s no clear track record of execution (e.g., the founder is fresh out of school), how can we properly assess them through proxies? If a founder might be great at recruiting but lacks past examples to show it, should I avoid that risk altogether when it falls into the fundamental category?
So, is conservatism the answer? No. I’ve seen many angel investors mistakenly confuse their own personal uncertainty about a company or sector with the actual risks involved. Non-linear outcomes demand a higher tolerance for risk.
Maybe a fresh deep tech founder hasn’t yet been able to prove their recruiting abilities, but you can assess other indicators: their ability to articulate a vision, their track record of rallying investors, or their ability to get great advisors on board. Passing on a deal due to uncertainty, especially in an industry driven by outliers, can be an extremely costly mistake. There are often precursors.
Throughput as a mitigant
Investors who understand this dynamic can back companies that embrace a trial-and-error process, even in deep tech companies, knowing that even in failure, the knowledge gained can open alternative avenues for growth. Speed of iteration becomes a powerful lever for success—the ability to quickly test hypotheses, fail or succeed, learn, and adjust. This is what clarifies the path over time.
In the context of deep tech, this puts an emphasis on technical innovations that enable a high throughput culture - Avoiding the pitfall of having to invest a considerable amount of time and resource for each new piece of information (ie: long path to market).
Small iterations build resilience. With each round of testing, a company faces stressors and variability in a controlled manner, becoming more adaptable. If a risk cannot be broken down into fundamental “units of risk” where the desired outcome is more likely than not, it should be avoided.
My process:
The process I use to evaluate and choose risks for deep tech companies can be characterized as:
Identify the most important challenges a company will face on its way up (3-5) for a specific industry.
Break down each of these challenges into their most essential nature.
Separate those challenges into “fundamental risks” and “personal uncertainties”.
For fundamental risks, think about whether they are one dimensional or pluri-dimensional, and evaluate if the outcome if they are solved matches the degree of difficulty.
Review “uncertainties” (risks where a clear answer cannot come from existing information) and see if they can be resolved through iterative testing.
Identify precursors that are likely indicators of the team’s ability to achieve success.
Work with the founder to peel back each layer of the onion. It should feel like we have the ability to address risks in successive layers, with the right resources allocated against the first layer (human, capital, access…) to build an antifragile path as each layer is peeled back.
Our job then becomes allocating investments against risk. This brings up two important considerations:
The founder’s understanding of risks is paramount. A red flag arises when, as an investor, you have a better grasp of the challenges than the founder. They live and breathe the business and industry and are likely dedicating a decade of their life to it—akin to a fund betting on a single investment with the expectation of a 10x return. Another one is when a founder is ‘blind’ to a risk, and completely refuses to see it.
Timing is crucial. This involves phasing challenges appropriately. A company will never tackle all risks upfront, and managing the timing of those risks is key. Each layer of risk removed should exponentially improve the odds of success.
Thank you for reading. The ideas in this post are very much a work in progress, and I would love to get your thoughts on the matter. Please reach out at julien@bleucap.com
—Julien