Yunqi Capital's Chen Yu: When Technology Paradigms Shift, Long-Term Opportunities Never Arrive Linearly | Yunqi Insights

2026 is nearly halfway through, and AI's connection to the physical world is reaching deeper.

AI agents have moved from buzzword to real-world deployment, while physical AI tracks like embodied intelligence and autonomous driving continue to heat up. The industry's focus has shifted from "is the technology sufficiently advanced" to "can it enter real scenarios, achieve stable delivery, and close the commercial loop."

When a new technological paradigm emerges, long-term opportunities rarely unfold in a straight line. They require leaps in technical capability, but also the simultaneous maturation of demand, cost structures, supply chains, and commercial environments.

Recently, PEDaily published a video interview with Yunqi Capital managing partner Chen Yu. Chen shared Yunqi's thinking on technological variables, industry inflection points, and startup moats.

This issue of "Yunqi Insights" excerpts the core perspectives, inviting readers to observe the long-term opportunities within this technological paradigm shift.

---

Opportunities Often Arise When Variables Emerge

When we evaluate investment opportunities, we tend to focus on changes in key variables. Macro policy, technological paradigms, and commercial environments can all serve as starting points for industry transformation.

Once a variable appears, you need to capture the signal quickly and position your investments accordingly.

Take robotics and autonomous driving. Yunqi first began tracking these sectors in 2016. At the time, China's demographic structure was shifting — aging and declining birth rates were becoming visible trends, suggesting that labor shortages could become a chronic issue.

If labor supply changes, using technology to supplement the workforce becomes a critical direction. Over the past decade, labor costs have risen steadily on one hand, while robotics, autonomous driving, and related technologies have advanced rapidly and supply chains have matured, bringing down the cost of relevant products and solutions.

When real demand exists and can be met at lower cost, the conditions for industry breakout gradually take shape.

New Technological Paradigms Often Trigger New Industry Breakouts

The large model industry is relatively younger. Its true inflection point came with the release of ChatGPT in late 2022. Since then, the entire industry has caught fire and developed at extraordinary speed.

This validates a judgment we had made earlier: the emergence of a new technological paradigm can trigger the breakout of entirely new industries.

In 2021, Yunqi met MiniMax founder Junjie Yan. At the time, the dominant approach in machine learning was still training separate small models for different tasks. Yan introduced us to the foundation model path: when parameter scale reaches a certain threshold, a single model could potentially solve multiple problem types, eliminating the need to train a dedicated model for each individual task.

At the early stage of an emerging technological paradigm, Yunqi chose to invest in MiniMax's angel round.

Looking back today, large model capabilities have progressed extraordinarily fast. They may not score 100 on every problem, but they have already reached sufficiently usable levels across a wide range of scenarios, with clear advantages over the previous small-model approach.

Large Models Have Changed the Underlying Logic of AI Application Entrepreneurship

In the past, when a team had a good product idea, the first question was whether it could technically be built.

Now, large models possess powerful general capabilities, and many problems can be solved relatively simply through them. This brings an important shift: entrepreneurial teams can devote more energy to product definition and scenario understanding, rather than solving every technical implementation problem from scratch.

This also brings a certain "democratization of technology." Entrepreneurs no longer necessarily need very large engineering teams; they can use large models to generate code, prompt engineering, and context engineering to complete product development that previously required much larger teams.

Under this shift, a 10-person AI team can potentially build what previously required a 100-person team — and do it even better.

The Key Question for AI Application Startups: Where Is the Moat?

The stronger large model capabilities become, the more opportunities there are for AI application entrepreneurship — but also the greater the challenges.

Because some opportunities that originally belonged to startups may be captured by the large model companies themselves. So when we evaluate AI application investment opportunities, we pay particular attention to one question: where is the startup's moat?

A good AI application first needs to genuinely understand user needs and be familiar with the problems and specific scenarios it addresses. Second, it needs to possess certain data barriers. If the relevant data is also available to the large model companies, then this problem will likely be solved directly by them as well.

Additionally, AI application companies have one advantage that differs from large model companies: large model companies tend to primarily use their own models, while application startups can call upon and orchestrate multiple large models, taking the best from each to achieve better results at lower cost. This is something that large model companies themselves may not easily accomplish.

Long-Term Opportunities Are Not Linear

Long-term opportunities in technology investing are often not linear.

Take hard-tech sectors like autonomous driving and robotics. From technology validation to product maturity to commercial breakout in real-world scenarios, there is typically a long journey of engineering refinement, scenario validation, and industry cycle fluctuations.

Many long-term opportunities do not unfold linearly. A wrong critical decision at any point, or a founder who doesn't persist, and the project can fail.

In the coming years, artificial general intelligence will most likely arrive. When it does, the world will change again, bringing new opportunities for entrepreneurship and investment.