--- title: "Three Investors Argued for 40 Minutes: How Would You Invest 100 Yuan in AI?" author: "Unique Research" sourcePublication: "Unique Research Substack" originalPublishedAt: "2026-03-29T12:03:11+00:00" canonical: "https://ffcap.cn/en/research/src-20260329-01html" source: "https://uniqueresearch.substack.com/p/src-20260329-01html" language: "en" --- # Three Investors Argued for 40 Minutes: How Would You Invest 100 Yuan in AI? _Original · Unique Research · 2026-03-29 · Shanghai_ _Editor's note: This is a historical panel account, not investment advice. Portfolio, valuation, staffing, performance and product claims below are attributed to the source or named speakers and have not been independently verified. “100 yuan” renders the source's colloquial hypothetical monetary prompt, not US dollars or an actual portfolio. The source spellings “Applide Intuition” and “Open Cloud” are retained rather than silently corrected; the latter differs from “OpenClaw” elsewhere in the article. Laoying Fund, Yingce Capital and Qidian Capital are provisional English renderings, and “Director Kong” is the source's speaker handle. The repeated Liblib paragraph is retained intentionally. The source refers to a roundtable on the 21st; this edition preserves its March 29 publication date. Image-content completeness remains under review._ Unique Research · ClawdChat Hardcore Party Roundtable 3 Investors Argued for 40 Minutes: Is Now Actually a Good Time to Build an AI Startup? “Have 100 yuan? I wouldn't invest in applications. I'd buy existing ByteDance shares.” At the recent ClawdChat Hardcore Party, moderator Wu posed a question to the investors. Is now actually a good time to start an AI business? Three investors gave three different answers. Fan Youxian of Laoying Fund said it definitely was. Stacy Jiao Wanrong of Yingce Capital said she was excited about the moment. Director Kong of Qidian Capital paused for a second, then said: “You can do it, but I'm very pessimistic about the application layer.” But Wu said: “Have 100 yuan? I wouldn't invest in applications. I'd buy existing ByteDance shares.” “Large Models Devour Everything” Director Kong does not like to beat around the bush. He has been active in Shanghai Zhangjiang's technology communities for more than 18 years, lived through the two bubbles of Web 1.0 and Web 2.0, and invested in industrial robots as well as a chip company serving network cameras. Recently, he has been helping a company prepare to go public, while also entering AI marketing. Asked how he sees the application layer now, he said bluntly: “I'm very pessimistic.” His logic is simple. He has lost money on SaaS twice. In his view, charging for software in China is either something the big technology companies can pull off, or the software is free. Ordinary startup teams caught in the middle face a dead end. “Today's agent businesses charge for software. I don't see what's different.” But his pessimism does not mean he is sitting back. He is already positioning around three hardware projects, including a “home box for the lobster”—a control terminal with 6 cameras, millimeter-wave radar and temperature sensors that comprehensively senses the home environment. “For intelligent AI hardware, you have to go to Shenzhen,” he said emphatically. The Night Top MIT Graduates Allegedly Built a Wrapper According to the source, Laoying Fund, where Fan Youxian works, has invested in 300 projects over 13 years. Her investment track record includes YC unicorns and GitLab, as well as drones before they became prominent and autonomous-driving simulation company Applide Intuition—still held, without an exit, at a reported US$15 billion valuation. She is not someone who casually declares one side superior in comparisons between China and the United States. But at the roundtable on the 21st, she made a rather pointed observation: “I saw news that Cursor had built a wrapper around Kimi. That was the work of four top MIT students. They didn't say anything at first; later, they apparently acknowledged it— they had built on our foundation model.” The author's interpretation of that remark was that Chinese large models had reached the point where a hugely popular, top-tier project founded by outstanding MIT graduates and built up jointly by Sam's OpenAI Startup Fund and Andreessen's a16z fund would quietly use one as the basis for a wrapper. This is the source's characterization, not an independently verified account. In this telling, we may be a step behind in cutting-edge technology, but once it emerges, the span from hardware and supply chains to software and markets means applications for 1.4 billion people will surely be broader. Her conclusion: look now at application deployment in China's vertical markets. A glance around the venue suggested that this era's opportunity was right here. The Investor Who Keeps Cutting Jobs at Portfolio Companies Director Kong's examples were more concrete. Recently, companies in their investment portfolio had let many people go, he said. “One fifth.” Who was being laid off? “Marketing staff—the kind who used to manage lots of accounts. Customer service: a company that used to have 40–50 customer-service staff can now basically have one AI cover the work. And another group—ordinary programmers implementing routine functions. This wave is really painful for them.” He paused, then said: “I've met them, and they're a little anxious. But I ran an experiment: a project that used to take five people was given to two people using AI comprehensively this time. It was genuinely fast. Even I didn't expect it.” His investment criteria have been recalibrated accordingly: “What used to take 1000 people now takes only 10. That's the kind of company worth investing in. If they cut what can be cut, I'll invest.” Stacy Tried a Whole Range of AI Tools—and Still Paid for Gamma's Annual Plan The source says Stacy of Yingce Capital invested in LiblibAI—whose overseas product is called Lovart AI, a design agent, and which recently introduced a LibTV video-generation Agent—as well as a European travel-planning agent that takes users from a Douyin-style travel-video feed to automatically planning routes and booking flights and activities, all in one click. She said OpenClaw had opened up a new paradigm, which excited her. “It lets me remotely control computer hardware to do work for me without carrying a computer.” But she was also candid: “Asking my grandparents to deploy OpenClaw now would be practically a fantasy.” That leaves room for a true mass-market consumer breakout that has yet to arrive. She believes the opportunity lies in “extremely specialized, atomic skills.” She offered her own example: “I've paid for Manus, Genspark, Perplexity, MiniMax and Doubao, and tried a whole pile of PPT tools. In the end, I still paid Gamma. The annual fee is over US$200, but it really is the best at making PPT presentations.” Choose a skill and become the best in the world at it— users will pay. There Is No Standard Answer to Where You Should Start a Company The moderator asked: “If I want to start an AI agent business now, where should I begin?” That question produced the most interesting disagreement. Stacy's answer: it depends on your own profile. If you know Chinese users' needs and understand how things work here, compete in China. If you have lived overseas, understand Western culture and can win American business customers, go there: investment is more active in the United States and there are more acquisition opportunities. “And AI is developing so quickly now that the starting point matters less—Liblib began in Beijing and became an international product in less than a year. They have an office in Silicon Valley now, too.” “And AI is developing so quickly now that the starting point matters less—liblib began in Beijing and became an international product in less than a year. They have an office in Silicon Valley now, too.” Fan Youxian's answer was more direct: “If you're a world-class entrepreneur, go overseas from day one and save yourself a lot of trouble. When you compare valuations, it's just too hard for founders in China.” “But most people should still stay rooted here and solve problems in this market. Put down roots and focus first. Once there are promising signs, find ABC partners and start expanding overseas, like ByteDance—Zhang Yiming's approach is to use local people in whichever country he enters. That's a top strategist.” Director Kong was the most pragmatic: “For hardware startups, the first stop is Shenzhen. For overseas marketing, the first stop is Hangzhou. Shanghai, well… I live here, ha ha. The money is here.” How Would You Spend 100 Yuan? As the session was ending, moderator Wu Wei asked one final question: Suppose you had a 100-yuan software budget. How would you divide it among models, Infra and agent applications? Stacy said this was not a question of how to spend money, but of investment sequence: infrastructure must be intelligent enough before good applications can grow out of it. Model valuations are already very expensive—she cited Anthropic at upwards of US$100 billion—but models are a prerequisite. Director Kong was more pessimistic: “The application layer? I don't think it exists. Large models devour everything.” Then he added: “Let me add a few more words today—as thin as a blade.” Fan Youxian's answer was straightforward: “Forget the first two. Put all 100 yuan into applications deeply deployed in vertical markets. Find a genius founder.” And the moderator himself? “I'd use the money to buy existing ByteDance shares.” The room laughed. The author presents this as the most candid answer among a roomful of investors: if you cannot tell who will break through, buy the platform first. A Final Word Director Kong said: “The person at the top needs to think this way. Hesitate, and before you know it, the payroll is pouring out.” He was originally talking about a company's attitude toward embracing AI—either go all in, or don't move. But I think it is also a miniature portrait of entrepreneurship in this era. Of course you can wait. Wait for large models to become more stable, for the paradigm to become clearer, for the market to mature. All of that is reasonable. But waiting has a price: the first people to break out are already looking back and counting heads. Which side are you on? Selected Q&A Q1: Who has the advantage—AI entrepreneurs in China or in the United States? Fan Youxian (Laoying Fund): In my investment track record, American founders make money for me, and I use it to subsidize Chinese founders—that's a joke. But at this stage, I think we should refocus on China. Our large models are getting better and better, we have a complete hardware supply chain and a market of 1.4 billion users. Application deployment will surely be faster. Stacy (Yingce Capital): Look at the founder's profile. If you understand Chinese users, compete in China. If you understand Western culture and can win American business customers, go to the United States. And AI is developing so fast now that the starting point matters less—liblib started in Beijing and became international in less than a year. Q2: Is now a good time to invest in AI or start an AI company? Fan Youxian: Definitely. The large-model infrastructure has been built. After so much money has been burned, everyone is waiting for a breakout application to recoup that investment. We talk about the lobster taking off—well, isn't that exactly what has arrived? Stacy: I'm quite excited at this moment. OpenClaw has opened an entirely new paradigm: interacting through IM and remotely controlling hardware expands what we can imagine for productivity. It's a good time. Director Kong (Qidian Capital): You can do it. But be clear about the approach you're taking. A purely traffic-driven APP? Don't do it. If you're strong in serving businesses, you can earn money first. If you're strong in supply chains, get moving. Q3: What exactly has OpenClaw opened up? Stacy: It has opened up a new agent architecture—you need only a phone to remotely control multiple computers doing work. It also opens up control over local data permissions. But it is still some distance from true mass-market consumer adoption. Asking my grandparents to deploy OpenClaw would be practically a fantasy. The real breakout moment will come when a consumer product makes it a single-step process. Q4: What directions would you advise entrepreneurs not to pursue? Stacy: Iterating and upgrading the original Open Cloud agent framework is within the big companies' range. Tencent, ByteDance, DeepSeek and MiniMax will all make their own versions. Ordinary entrepreneurs need to think this through quickly. Director Kong: Purely traffic-driven APP businesses. You need very strong traffic resources first; without them, don't do it. And I'm very pessimistic about purely subscription-based software. I've tried SaaS twice and never succeeded. Q5: Are there really still opportunities in vertical applications? Stacy: Yes. I've personally paid for Manus, Genspark, Perplexity, MiniMax and Doubao… I've tried a whole pile of PPT tools, but in the end I still paid Gamma, over US$200 a year. In that one vertical of making PPT presentations, it has become the best in the world. Go extremely vertical, and users will pay. Director Kong: I'm very pessimistic about the application layer. Large models will gradually swallow it. The money earned at that layer will get thinner and thinner. Q6: Which city should you start a company in? Director Kong: For intelligent hardware, the first stop is Shenzhen. For overseas marketing, the first stop is Hangzhou. Shanghai… I live here, ha ha. The money is here, so I put Shanghai on the list. Fan Youxian: Each place has its strengths. Zhangjiang suits medical technology, chips and medium- to long-term technology development; Shenzhen suits new ToC technologies. Q7: What should ordinary programmers do in the AI era? Director Kong: Recently, I've been getting companies I've invested in to fire people. Programmers implementing routine functions are having a painful time in this wave—where can they go to do ordinary functional work? I ran an experiment: a project that used to take five people was handled by two people comprehensively using AI. It was genuinely fast. I suggest they come to talks and first learn how to operate AI. Be anxious if you must, but first get AI working for you. Q8: The Token Economy—What Is the Best Business Model? Director Kong: Frankly, the best business is “reselling Token.” Get Token from big companies, package it wholesale and then subcontract and sell it to users. It's like being a dealer in the telecom-operator era: you can't work out how you spent that 100-plus yuan—and that is the profit. Fan Youxian: Agreed. Look at MiniMax's valuation rising 6-fold. Why? Isn't it because Token usage exploded? The lobster is hot, everyone is using it, so more Token gets sold. Q9: Where Will AI Ultimately Take Humanity? Fan Youxian: AI ultimately frees up labor. What do we do once labor is freed? Enjoy life, have our emotional needs met, smile ourselves awake even in our sleep. So the best ToC applications for AI are emotional companionship, pet companionship and the transformation of education. The emotional needs of 1.4 billion people—how big a market is that? The industrial era turned people into machines. In the AI era, thank AI for liberating humanity: at last we can return to our humanity and enjoy what makes us human. Q10: If you had 100 yuan, how would you divide a software budget (models / Infra/ applications)? Fan Youxian: Forget the first two. Invest all 100 yuan in founders building deeply deployed vertical applications. Find a genius and back them. Director Kong: I don't look at the application layer. Large models devour everything. Stacy: It isn't a question of how to spend the money, but of investment sequence. First you need intelligent enough models and good infrastructure; only then can good applications grow. Wu Wei: Have 100 yuan? I'd buy existing ByteDance shares. Article source: Unique Research · ClawdChat Hardcore Party Roundtable --- Original publication: https://uniqueresearch.substack.com/p/src-20260329-01html On-site reading page: https://ffcap.cn/en/research/src-20260329-01html