--- title: "A New Direction in AI Investing: Back Those Who Combine Technology with Real Use Cases" author: "Unique Research" sourcePublication: "Unique Research Substack" originalPublishedAt: "2026-03-20T11:07:27+00:00" canonical: "https://ffcap.cn/en/research/src-20260320-03html" source: "https://uniqueresearch.substack.com/p/src-20260320-03html" language: "en" --- # A New Direction in AI Investing: Back Those Who Combine Technology with Real Use Cases _Original · Unique Research · 2026-03-20 · Shanghai_ _English edition note: This full translation preserves the source author's analysis and Lu Hongyu's interview views as of March 2026. Statements about investment timing, fund-return concentration, business models and generational advantages are attributed opinions, not independently verified performance findings or investment recommendations from this English edition. Chinese birth-cohort labels are rendered as the corresponding decades of birth. Person, firm and network names are translated or romanized from the source without establishing official English names or independently confirming roles._ Unique Awards · Guest Interview AI Investing Is Shifting from Looking at Technology to Looking at Results A Conversation with Detong Capital's Lu Hongyu: The Biggest Opportunities Are Not Only in Models, but in Remaking Traditional Industries Interviewee: Lu Hongyu, director and senior partner at Detong Capital Interview date: March 2026 Over the past two years, almost every investment firm has been looking at AI. But as they looked more closely, the market began to diverge. Some still focus on model parameters, securing positions in compute, and the infrastructure battle. Others have turned to a more practical question: where will AI ultimately be applied, and who can genuinely turn it into a business, results, and lasting value? What struck me most after speaking with Lu Hongyu, a director and senior partner at Detong Capital, was not his excitement about a particular hot sector. It was a characteristic investor's judgment: AI matters, of course. But what is worth backing is not technology alone. It is whether combining technology with real use cases can produce new companies, new models, and even new ways of organizing industries. In other words, investing in AI is no longer only about backing a smarter model. It is moving toward backing a system that can deliver better results. And that may be what makes this wave of AI truly interesting. Why Is Detong Going All in AI? Many people assume venture capital's (VC's) large-scale turn toward AI happened only this year. But in Lu Hongyu's account, Detong Capital's sustained attention to AI began much earlier. It has been watching since ChatGPT appeared, and its conviction has grown: this is not a short-lived trend, but a profound generational shift in technology. Detong's current stance can be summarized in one phrase: All in AI, AI for All That phrase deserves unpacking. The first half expresses investment commitment. Rather than treating AI as a subsector within a thematic fund, it means allocating around AI as the most important variable of the coming years. The second half is more important. It does not look only at the most upstream, dazzling layer. It reflects a belief that AI will eventually permeate every industry. Opportunities belong not only to model companies or a small technical elite, but also to entrepreneurs who can bring AI into healthcare, manufacturing, services, content, consumption, and enterprise processes. What Has Changed in AI Investment Logic—and What Has Not? Every technological revolution creates an illusion: have the old methods of investing stopped working? Lu Hongyu's answer was interesting. He did not describe AI as an entirely new rulebook overturning the old world. It was more like this: " Some underlying principles have not changed, but AI has indeed rewritten many key variables. The most obvious change is that AI products differ from many internet products of the past. They come with traffic, stickiness, and a kind of vitality of their own. This reflects a practical development. Previously, taking a startup from 0 to 1 often meant building a team, developing channels, buying traffic, and doing sales—a heavy, slow process. In the AI era, the possibility of a one-person startup has suddenly increased. A sufficiently good product can itself become a distribution entry point and even generate its own growth. Lu Hongyu noted that compared with earlier TMT projects, AI projects need to be more product-driven than channel-driven. The remark sounds understated, but it is crucial. It means the decisive advantage for many startups today is no longer who has the broadest reach or deepest relationships, but who first delivers the experience users genuinely need. Large Models and Vertical Applications Are Not an Either-Or Choice Many AI entrepreneurs get stuck because they feel they are standing at a fork in the road: Should they bet on general-purpose large models or build vertical applications? Some even wonder whether they should wait until the large-model landscape becomes clearer before deciding what to build. But Lu Hongyu's position was clear: There is no need to wait. This sends an important signal to many entrepreneurs. At this stage, general-purpose models will continue evolving, but that does not eliminate opportunities for vertical applications. On the contrary, the stronger model capabilities become, the more room vertical applications have to operate. The key is not whether you are doing something large, but whether you have identified a real, frequent, enduring problem. So what is a vertical AI application's moat? Lu Hongyu's answer was direct: Data. This is not just a familiar platitude. By now, people have gradually recognized that merely wrapping a model does not create a barrier. What enables a vertical application to go deep is the underlying combination of use-case data, user feedback, task chains, industry know-how, and continuous iteration. To B First, Then To C, May Be a More Practical Path for Chinese AI Startups There has always been considerable debate about To B and To C. Some believe the consumer market offers more possibilities and an easier path to a hit. Others see the business market as more solid and more likely to generate revenue. Lu Hongyu's judgment was: B2B in China, B2C overseas The reason is simple, summed up in two Chinese characters: cost. This fits the realities of Chinese AI entrepreneurship. The consumer market certainly has opportunities, but China has a longstanding problem: willingness to use a product does not necessarily mean willingness to pay. When a product continuously incurs model, inference, and service costs, consumer commercialization becomes very difficult without especially strong content, entertainment, or network effects. In his view, Chinese AI companies facing low consumer willingness to pay may find an initial opening in cultural entertainment and games. Is Embodied Intelligence Investable Now? Yes—but Do Not Assume It Is Already Mature Over the past year, enthusiasm for robotics and embodied intelligence has continued to rise. In the Greater Bay Area in particular, projects have proliferated. Almost every conference includes discussion of whether humanoid robots are the next major opportunity. Lu Hongyu's assessment was neither aggressive nor pessimistic. He believes the sector is still early, but the time to invest has arrived—at least for the current stage. That is a carefully calibrated statement. It acknowledges that embodied intelligence is far from mature. At the same time, it suggests that if investment firms wait until everything is clear, many windows of opportunity may already have closed. But he emphasized one point: The use case matters enormously. Many people discussing robots imagine an all-purpose, Transformers-like presence: a general-purpose robot that can move seamlessly into any environment and perform every task. Reality is clearly different. In actual industrial deployment, the scenario determines the need, the need determines the capability boundaries, and those boundaries in turn determine the speed of commercialization. An AI Boom Does Not Make Entrepreneurship Less Brutal When discussing AI startups, it is easy to be swept along by optimism. Funding announcements multiply, valuations rise, and everyone calls it a once-in-a-decade opportunity. Yet ask about investment returns, exit paths, and startup survival rates, and the room quickly falls silent. Lu Hongyu was candid about this. He noted that fund returns already come from 20% of investments. AI will probably be much the same; that pattern does not disappear simply because the sector is AI. It is a sobering remark. It reminds entrepreneurs that AI does not automatically reduce the chance of failure. It raises the ceiling for those who succeed. Why Is He More Optimistic About Founders Born in the 2000s? This was one of the most interesting parts of the interview for me. We keep using the term AI Native. But who is more AI Native? Veterans with extensive industry experience who have been through several waves of internet and enterprise services? Or young founders who naturally treat AI as a working language, a way of thinking, and a creative partner? Lu Hongyu leaned toward the latter. He said: Compared with experienced veterans, I actually feel founders born in the 2000s have more of an advantage, because AI requires creativity. There is an intriguing generational judgment behind this. In traditional entrepreneurship, experience mattered greatly. Understanding an industry, relationships, processes, and how to move things forward step by step could all create barriers. But in the AI era, much is being rewritten. Past experience is valuable, but it can also create path dependence. People not yet fully shaped by the old order may find it easier to accept new ways of working, new product logic, and new organizational structures. Lu Hongyu said the youngest AI founder he had met was 22 and the oldest was 48. He captured the difference neatly in four Chinese characters: Ideals and reality. Younger people are more likely to believe the rules can be redefined; older people better understand the real world's friction and boundaries. Perhaps the interesting question is therefore not which group is stronger alone, but who can put both forces into the same system. Between AI and Traditional Industries, Delivering Results Matters More Than Explaining Technology Many AI founders fall into a trap when communicating with traditional industries: They want to prove how advanced their technology is. But traditional business owners usually care less about the strength of the model, the intelligence of the Agent, or the elegance of the architecture. They care about something else: Will it ultimately deliver results? Lu Hongyu's phrase was simple: Be results-oriented. Those four Chinese characters explain why so many AI companies can discuss frontier concepts at length yet still fail to enter traditional industries' core use cases. For traditional industries, technology is not the end. Business results are. Can it lower costs? Improve efficiency? Reduce errors? Shorten processes? Generate revenue? Unblock a previously stuck part of the organization? If the answer is yes, the conversation becomes much easier. If the answer remains a stream of technical language, most collaborations will stop at sounding promising. This Wave of AI Investing Is Betting on New Relationships of Production, Not on Trends By the end of the conversation, I increasingly felt that the interview's value lay not in how many standard answers it supplied about what to invest in. It lay in the deeper shift it suggested: Investors today are no longer backing only a model, a technology, or a specific sector. They are betting that AI is rewriting relationships of production. Will one-person companies become more common? Will product-driven growth become stronger? Will traditional industries be reorganized? Will young founders gain more opportunities in the new cycle because they think more freely and are more willing to experiment? Will robots and embodied intelligence first succeed in specific scenarios and then help define the next generation of infrastructure? None of these questions has a fully settled answer yet. But one thing is becoming clearer: AI opportunities belong not only to people who build models, but also to those who can recombine models, data, use cases, organizations, and results. The truly large companies may emerge from that process of recombination. Selected Q&A Q1: Please introduce yourself and Detong Capital to our readers. Lu Hongyu: At Detong Capital, I focus mainly on AI, TMT, and overseas investments. Our current attitude toward AI can be summarized in one phrase: All in AI, AI for All. We have actually been following this direction continuously since ChatGPT appeared. Q2: What is Detong Capital's approach to investing in AI? Lu Hongyu: We look at infrastructure, models, and applications, with different teams covering different areas. But in terms of our differentiated approach, I think we lean more toward applications and pay closer attention to combining AI with traditional industries. Q3: What is the biggest change in investment logic from the mobile-internet era to the AI era? Lu Hongyu: AI comes with stickiness and traffic and has greater vitality. Another major change is the increasing possibility of one-person startups. Compared with many earlier TMT projects, AI projects need to be more product-driven, rather than relying purely on channels. Q4: How do you choose between large models and vertical applications? Lu Hongyu: There is no need to wait. Vertical applications are entirely viable. One very important part of their moat is data. Q5: Which direction do you favor for AI startups: To B or To C? Lu Hongyu: I lean toward B2B in China and B2C overseas, mainly for cost reasons. As for low consumer willingness to pay in China, I think AI companies can begin with cultural entertainment and games. Q6: How do you view embodied intelligence and robotics? Lu Hongyu: The sector is still early, but the time to invest has arrived, at least for the current stage. Use cases are very important, however. I have not yet seen that kind of all-purpose, Transformers-like form. Q7: How do you assess early-stage AI projects that have only a demo and no revenue? Lu Hongyu: Not every publicly listed company already has very high revenue either. You still have to make an overall assessment rather than focusing only on current revenue. Q8: Will the 80/20 rule become more extreme in AI? Lu Hongyu: I think it will be similar. Fund returns already come from 20% of investments. Q9: How do founders born in the 1980s, 1990s, and 2000s differ in AI? Lu Hongyu: The youngest AI founder I have met was 22 and the oldest was 48. To some extent, the difference can be summarized as ideals and reality. Q10: Who has more of an advantage: experienced industry veterans or AI Native founders born in the 2000s? Lu Hongyu: I feel those born in the 2000s have a slight advantage, because AI requires creativity. Q11: How can AI founders communicate effectively with traditional business owners? Lu Hongyu: Be results-oriented. Q12: As an experienced investor, how do you understand AI founders born in the 2000s? Lu Hongyu: Find ways to make yourself younger. Q13: How does Detong help portfolio companies build collaboration across generations? Lu Hongyu: Strengthen communication and exchange, use the ecosystem of our network and mechanisms such as the Deyou Club, and help people complement one another's strengths. --- Original publication: https://uniqueresearch.substack.com/p/src-20260320-03html On-site reading page: https://ffcap.cn/en/research/src-20260320-03html