Welcome to the 128th edition of Deep Tech Catalyst, the educational channel from The Scenarionist where science meets venture!
This week, I sat down with Sabrina Sasaki, Partner at Monozukuri Ventures. We explored how scientific founders can move from laboratory validation to industrial adoption, why milestones must reflect the physical complexity of the product, and how Japan’s manufacturing ecosystem can support international commercialization through pilots, co-development, and strategic partnerships.
Key takeaways from the episode:
🤖 Physical AI requires more than great hardware
Hardware creates access to the physical world, but long-term value comes from turning data into useful predictions, insights, and operational decisions. The strongest companies understand how the physical system fits into a broader software and data architecture.
🔬 A laboratory prototype is not yet a commercial product
Real industrial environments introduce variability, integration constraints, and reliability requirements that controlled experiments cannot fully reproduce. Founders need to test assumptions in the field and adapt the technology around the customer’s actual problem.
💰 Deep Tech milestones must reflect the physics of the product
Different technologies require different amounts of time, capital, and technical progress before reaching a credible result. A meaningful milestone should demonstrate measurable progress toward a commercially viable product while reflecting the time, capital, and technical resources required at that stage of development.
🇯🇵 Japan offers more than access to a large industrial market
Its specialized manufacturers, suppliers, and engineering capabilities can help startups validate, develop, and manufacture complex technologies. For the right company, these partnerships can provide access to resources and expertise that may complement or offer an alternative to developing certain capabilities internally.
A World Bank for Defence | Deep Tech Briefing 119
Independent intelligence for deep tech allocation and industrial strategy.
BEYOND THE CONVERSATION — STRATEGIC INSIGHTS FROM THE EPISODE
Physical AI Becomes Investable When Hardware Meets Industrial Reality
Physical AI is often discussed as the next frontier of artificial intelligence: a world in which machines can perceive, interpret, and act within physical environments.
But from an investment perspective, the category cannot be reduced to smarter robots or more capable automation systems. Its real significance lies in the convergence of hardware, software, data, manufacturing, and industrial adoption.
Hardware gives an AI system access to the physical world. Sensors allow it to observe. Actuators allow it to respond. Robotic systems allow it to perform tasks. Yet the hardware alone does not necessarily create a defensible company.
A growing part of the value comes from what happens after the physical interaction has been captured.
Software must interpret the information generated by the system, identify which signals matter, and convert those signals into useful operational decisions. In industrial environments, this may mean supporting predictive or preventive maintenance, or helping operators understand what action should be taken next.
This must shape the logic of venture building.
A hardware startup cannot assume that engineering complexity will automatically translate into commercial differentiation. The technology must be connected to a broader system of data, intelligence, and business value.
At the same time, the company must demonstrate that the technology can perform effectively in real customer environments, not only under controlled laboratory conditions.
That requires more than a working prototype. It requires a team capable of learning from customers, adapting to uncontrolled environments, financing capital-intensive milestones, and choosing partners that reduce rather than increase execution risk.
A Prototype Is Not Yet a Commercial Solution
Deep tech companies often begin in laboratories, research institutions, or highly controlled engineering environments.
That is where the underlying science can be tested. It is where researchers can isolate variables, measure performance, and demonstrate that a technical concept is possible.
This work is necessary. But it can also create an incomplete sense of validation.
Researchers naturally speak with other researchers. They encounter similar technical problems, use a common vocabulary, and operate within environments that share many of the same assumptions.
A new component or system may therefore receive strong interest from other universities, laboratories, engineers, or designers building specialized prototypes. Those early users may understand the technical achievement immediately.
The difficulty is that validation inside this community does not necessarily prove that a large industrial customer has a commercially meaningful problem.
A technology can be valuable to another research team and still have a limited path toward broader commercial application.
Leaving the controlled environment
The transition becomes particularly difficult when the technology involves robotics, physics, chemistry, or industrial automation.
Inside a laboratory, the founders can control the environment. They can determine how the system is positioned, which variables are introduced, and how each experiment is conducted.
A customer environment operates differently.
Conditions may change, unexpected variables may affect the system, and the founders may no longer be able to control the experiment in the same way.
Field robotics makes this gap especially visible. A machine that performs perfectly under controlled conditions may behave very differently when the environment around it changes continuously.
The same applies to materials and chemical systems. Physical and chemical processes may behave differently when they are moved from a controlled laboratory into a real customer environment.
This is why a perfect laboratory prototype is not necessarily close to a commercial product.
One purpose of early customer engagement is not simply to receive confirmation that the technology is impressive. It is to discover which assumptions fail once the system enters the real world.
Building toward the commercial problem
At the earliest stages, the startup may not be able to build the version that a large corporation will ultimately purchase.
That final product may require too much capital, too much time, or too much engineering work. The level of maturity and performance expected by the customer may be impossible for a young company to achieve immediately.
The startup still needs to demonstrate progress.
An effective approach is to identify shorter milestones that move the technology closer to the commercial need.
A second version of the prototype may improve one critical part of the system. Each milestone can test and improve part of the solution while showing customers that the technology is advancing.
These steps do not constitute full commercial traction. They show that the company is learning in the right direction.
Investors need evidence that the founders understand the gap between what currently exists and what a customer will eventually pay for.
They also need to see that the team can break that gap into milestones that are technically credible and appropriate to the scale and resources required by the product.
Learning from customer feedback
One of the hardest transitions for scientific founders is learning how to respond to negative feedback.
Positive feedback is comfortable because it confirms the original hypothesis. It reinforces the value of the science and supports the idea that the founders are building the right solution.
Negative feedback can be especially useful.
It may reveal that the customer does not experience the problem in the way the founders expected. It may also challenge the original assumptions about the application or show that the solution needs to adapt to what the market requires.
The point is to apply the same willingness to test assumptions to the market.
Founders can treat customer interactions as another source of evidence. They can test assumptions, examine contradictions, and adjust the product when reality does not support the original design.
Milestones Must Match the Scale of the Product
It is difficult to discuss comparable roadmaps for physical AI or industrial deep tech. The capital required to develop a sensor is not comparable to the capital needed to build a large robotic infrastructure system.
An actuator can sometimes be prototyped using existing components. A new tunneling platform, energy system, or aerospace product may require substantial resources before a credible minimum viable product can even exist.
The investment stage alone does not determine what progress should look like. The scale and technical nature of the product are equally important.
At pre-seed, the team is the main evidence
The initial market thesis may change. The use case may evolve. The founders may discover that their differentiation is different from what they originally believed. Customer conversations can redirect the product, and early experiments can expose limitations that reshape the entire company.
For that reason, pre-seed investment is often primarily an assessment of the founders and the relationship among them.
The important questions concern how the team works under pressure, how responsibilities are divided, and how the founders respond when experienced people challenge their assumptions.
Accelerators, mentors, and advisors can be useful partly because they create this pressure. They force founders to explain why the company is different, where the value comes from, and what evidence supports their claims.
The purpose is not simply to improve the pitch. It is to observe whether the founders can process criticism without becoming defensive or losing conviction.
Early-stage company building requires both.
A team must be confident enough to pursue an uncertain technical vision and flexible enough to change when the evidence demands it.
Seed stage
By seed stage, investors expect more than a scientific concept, but what counts as meaningful progress varies widely.
For a component company, it may be possible to build multiple prototypes, test them with customers, and improve the design relatively quickly.
Off-the-shelf parts may reduce development time, and the startup may be able to reach a functional demonstration with limited resources.
For a company developing a more complex infrastructure technology, each prototype may be expensive. The company may need specialized equipment, large testing environments, or physical installations before it can prove that the system works.
The first few units can cost disproportionately more than later production.
The relevant question is whether the company has defined milestones that reflect the scale of what it is building.
A credible plan must show what can be learned from each unit of capital. It must explain which technical risks will be retired, which assumptions will be tested, and how each development stage brings the company closer to a product that customers can adopt.
Atomic-Scale Process Control for the Ångström Era of Chips | Rumors
Advanced fabs are entering the Ångström era. An analysis of four early-stage startups building for the atomic edge of chip manufacturing.
Japan’s Industrial Ecosystem and Opportunities for Startups
Japan remains an attractive environment for industrial and manufacturing technologies because much of its competitive strength exists below the level most founders initially see.
International entrepreneurs usually recognize the country through its large consumer brands. They know the automotive companies, electronics groups, and industrial names that became symbols of innovation during previous decades.
But many of those companies built their products through networks of highly specialized suppliers.
Small and medium-sized manufacturers developed deep expertise in specific materials, components, processes, and production techniques. In many regions, clusters formed around a narrow category of industrial capability.
Their specialized knowledge and manufacturing capabilities may be valuable resources for startups to access.
Industrial customers rarely behave like early adopters
A corporation will not normally accept technical risk simply because the startup is ambitious. Industrial companies are generally unwilling to take risks on behalf of startups, particularly when evaluating solutions for business-critical environments.
The required quality level is therefore much higher.
This is particularly important when working with large manufacturers, whose reputation has been built around rigorous quality control.
A startup cannot approach these companies with the expectation that they will tolerate an unreliable product while it develops.
Before pursuing a serious industrial relationship, founders must understand the quality standards against which their technology will be evaluated.
Corporate interest does not necessarily indicate commercial progress
Many large companies are still learning how to work with startups. Their internal processes may move slowly. Decision-making authority may be unclear.
The innovation team may be interested in the technology without controlling the budget or business unit required for adoption.
On the other hand, an early-stage company has limited time and limited capital.
A large corporate conversation can appear strategically important because the brand is recognizable and the potential contract seems significant. The startup may invest months in meetings, technical discussions, custom requests, and internal presentations.
Yet the relationship may not move the company toward its real milestones.
Therefore, before achieving product-market fit, identifying meaningful use cases, securing sufficient financing, or developing a stable product, extensive corporate engagement could become a distraction.
This does not mean founders should ignore large companies until they are fully mature. They can begin building the relationship earlier.
They can educate potential partners, share product developments, communicate new use cases, and keep relevant stakeholders informed about fundraising and technical progress.
The distinction is between nurturing a relationship and depending on it.
Conversely, once the startup has reached a greater level of maturity, the conversation with a corporate partner can become more productive.
A company with a stronger product, credible customer use cases, and a clearer market position is better prepared to engage with an industrial corporation.
Before reaching that point, founders can keep potential partners informed about the company’s progress, applications, customers, and fundraising activity without relying on them to drive the startup’s immediate milestones.
This is especially important because some corporations may not fully understand how to work with companies that have not yet reached product-market fit.
The objective is to pursue industrial partnerships at a stage when they can genuinely support the company’s development.
The Right Industrial Partnership Depends on What the Company Needs
There is no single model for working in industrial B2B markets. The right structure depends on the technology, the resources required to build it, and the strategic importance of the market.
For advanced materials and semiconductor companies, collaboration with an established manufacturer can be highly valuable. These startups may need equipment, laboratories, process knowledge, and production infrastructure that would be unrealistic to recreate independently.
A larger industrial partner may already possess the machinery and technical teams required to test the technology at a different scale.
In those situations, co-development can create interesting opportunities.
The startup contributes a new technical approach. The manufacturer contributes facilities, production knowledge, and the ability to translate the concept into a repeatable process.
The challenge is negotiating partnerships that allow both sides to test and develop the relationship effectively.
One option is to begin with a limited agreement that allows both parties to test the relationship. The collaboration may run for a defined period to understand whether the technical work and working dynamics are effective.
This can reduce uncertainty.
Both sides can use this period to understand whether the collaboration works and whether it can support further technical development.
The purpose of the initial relationship can be to create evidence.
It should demonstrate that the companies can build something together and provide a basis for deciding whether a deeper partnership makes sense.
International commercialization considerations
Opening an operation in another country consumes management attention and financial resources. It also requires the company to commit significant time and organizational effort to that market.
A local presence becomes rational when the company’s strategy is genuinely tied to a specific ecosystem.
This may be the case when partners are essential to manufacturing, when the country is a core customer market, or when local grants and industrial partnerships can materially accelerate the business. In other situations, a local representative or commercial partner may be sufficient.
In other words, the operating model should follow the company’s real dependency.
For instance, an established industrial partner can provide the infrastructure and capabilities needed to manufacture the product while helping control manufacturing risk according to established standards.
This can unlock resources that would be difficult for the startup to obtain independently. Moreover, public funding may become more accessible when the company has a credible plan to manufacture locally with a respected partner.
Furthermore, a manufacturer that already serves customers in other Asian industrial economies may become a route into those ecosystems.
But the value of these options depends on how well the partnership model aligns with the company’s strategy and technological needs.
A semiconductor component, an advanced material, or a complete robotic system require different resources and different forms of collaboration.
In conclusion, the company must begin with a clear understanding of what it needs from the partner, what it can develop independently, and how the relationship will help it reach the next stage of commercial development.
