[Field Patterns] If the Physics doesn't bend - the Business Model must.
The Problem Is Not That Science Fails to Scale. It's The Startup Model That Fails to Understand Science.
A Category Error Hiding in Plain Sight
There is a quiet category error at the center of European innovation policy, persistent like a bad smell. We continue to call science ventures startups, as if the term were merely descriptive. As if it did not encode an entire economic logic. As if it did not quietly prescribe how these ventures are expected to behave. And then we blame founders or even science (like the latter would listen to wishful thinking)…
The Gabler Wirtschaftslexikon defines a startup as a young company implementing an innovative business idea, typically financed through external capital such as venture capital.
The German Startup Monitor and large parts of the entrepreneurship literature follow a similar structure. A startup is young, innovative, growth-oriented, scalable. In Steve Blank’s widely cited formulation, it is a temporary organization designed to search for a repeatable and scalable business model.
The implicit sequence is consistent:
There is a market hypothesis.
There is business model discovery.
There is rapid scaling.
Venture capital accelerates the process.
The entire architecture assumes that the core uncertainty is market-based.
Science ventures do not begin there.
They begin with technological uncertainty.
Before customer validation, there is feasibility validation. Before product–market fit, there is experimental proof. Before growth metrics, there is the simple question: does the underlying science hold under real-world conditions?
That difference is not semantic. It is structural.
Two Risk Architectures, One Label
Traditional startups predominantly carry implementation risk. They test whether customers adopt a product, whether a business model scales, whether execution can keep pace with demand. The technology usually works. The uncertainty lies in adoption and scaling.
Science ventures carry technological risk first. The underlying mechanism may fail. The scaling pathway may not yet exist. The regulatory environment may not be compatible. The validation cycles are longer, more expensive, and less compressible.
Empirical research shows that this structural difference is not trivial. A 2023 study in the Strategic Entrepreneurship Journal documents a 38% decline in startup formation among science and engineering PhDs in the United States since 1997.
This decline is broad and persistent. It is not limited to a single discipline or demographic group.
The authors link this trend to increasing knowledge complexity. Medical knowledge that once doubled over decades now doubles within only a few years. Innovation increasingly requires larger teams, deeper specialization, and more hierarchical coordination.
Large firms can absorb this complexity. They can add layers, redistribute tasks, build knowledge hierarchies. Science founders cannot do so as easily. The same study shows increasing task complexity and administrative burden for founders, alongside relatively declining earnings compared to established firms.
Entrepreneurship has not become simpler in an era of deep tech. It has become cognitively heavier.
Yet we continue to treat science ventures as if their core challenge were business model optimization.
Capital Is Not Neutral
The misalignment becomes sharper when capital enters the picture. Research on venture capital timing shows that science-based startups systematically secure VC funding later than other ventures.
This is not because scientific founders lack ambition. It is because they often prioritize technological validation before business model validation.
Venture capital, however, is structurally optimized for post–product-market-fit scaling. Its time horizons and return expectations assume that the central uncertainty is already reduced to execution.
This creates friction that is rarely articulated explicitly:
The technology clock runs longer than the funding clock.
Validation cycles do not compress to quarterly reporting rhythms.
Scientific uncertainty does not disappear because capital expects it to.
When founders are told they “scale too slowly,” the diagnosis may simply ignore that they are still operating in a different phase of uncertainty.
This is not founder incompetence. It is architectural mismatch.
The Value Capture Paradox
Even when science ventures succeed technically, structural disadvantages remain visible. Empirical evidence shows that science-based startups capture only 46 cents per dollar of acquisition-induced surplus, compared to 61 cents for non-science startups.
They generate significant joint surplus. They capture less of it.
Thinner acquisition markets, higher technological uncertainty, and fewer potential buyers weaken bargaining power. Scientific depth does not automatically translate into negotiation leverage. The narrative of “deep tech premium” often masks a harsher economic reality.
This is not a performance failure. It is a structural distribution issue.
Measuring the Wrong Thing at the Wrong Stage
The distortion becomes most visible in how we measure success. A comprehensive review of technology-based startup performance metrics shows overwhelming emphasis on growth, revenue, funding rounds, and liquidity events.
These metrics make sense in a business-model search framework. They assume that the technology is sufficiently stable and that scaling is the central objective.
But early-stage science ventures often operate in an entirely different regime. Even investors acknowledge that early stages lack robust quantitative KPIs and rely heavily on qualitative judgment.
Hard financial metrics are often unavailable or not yet meaningful.
Yet public funding programs, incubators, and policy evaluation frameworks continue to reward traction, speed, and scale. The implicit message is clear: if you are not growing, you are failing.
The possibility that you are still validating the laws of nature does not fit into a quarterly growth chart.
A Structural Commercialization Problem
This is not a growth problem. It is not primarily a founder mindset problem. It is not a cultural deficit in European entrepreneurship.
It is a structural commercialization problem.
Scientific complexity is increasing. Knowledge is fragmenting. Validation cycles are lengthening. Large firms adapt by building internal coordination capacity. Venture capital adapts by concentrating on ventures with clearer scalability signals. Science founders are left navigating a system calibrated to a different type of uncertainty.
If Europe speaks about technological sovereignty while evaluating science ventures with startup growth metrics, then the system is inherently inconsistent. It expects rapid scalability from ventures whose primary task is still technological stabilization.
The uncomfortable question, therefore, is not whether science founders should become better startup CEOs. It is whether the institutionalized startup model is fundamentally unfit as the default template for science-based ventures.
Precision in language shapes precision in policy. If we continue to mislabel science ventures, we will continue to misdesign the commercialization architecture around them.
And then we will continue to wonder why deep tech stalls in TRL 4–6…
Some people will look at the moon and still try to tell you it’s sun. But data and science don’t like or follow opinions, so here are:
Selected References: