Zhong Tou Yushu but missed the follow-on: patient capital also gets stuck by fund cycles
Vertex was the sole institutional investor in Unitree's Pre-A+ round in 2020, but in the 2023 round it couldn't follow on because its own USD fund was expiring and it was raising a new fund. Doubling down tests not just vision, but the rhythm match between fund and project.
The video won't play here. Listen to the audio instead:
The argument · tap a timestamp to hear it
They started investing when chips were cheaper than potato chips
Vertex invested in semiconductors as early as 2011, and in 2012 invested in a new materials company making solid-state batteries. At the time, domestic VCs hardly looked at these because the internet was at its peak. Zheng Juncong told a joke: chips are called chips, and everyone thought chips were cheaper than potato chips, so no one invested, because chips relied on imports and sold for only one US dollar each. He judged that chip technology content was extremely high, from millions of lines of code to billions of lines of code, and all electronic products, cars, and the cloud needed them. The pitch for investing in chips also changed with the times: initially it was domestic production being tax-free, later it was iterating around the mobile phone supply chain, and now it's about being bottlenecked plus next-generation applications like optical computing.
— Zheng JuncongIC can pass by half, and there's a golden share
Vertex's investment committee does not require a unanimous vote; a majority suffices, and they even hold ad hoc meetings on weekends if a project needs to go to IC. More uniquely, there's a golden share: a partner has one chance a year to force investment in a company even if the IC does not approve. Votes are secret; you don't know whether a partner voted for or against, because they don't want everyone to do favors for each other and abandon independent judgment. Voting choices are also tied to project outcomes long-term; if a project fails but you voted for it, it will be traced back and counted in your score, and when distributing carry, it will be brought out to look at. This mechanism is designed to force independent judgment, not personal favors.
— Zheng JuncongThey don't chase trends because founders of hotly contested deals get arrogant
Vertex explicitly does not invest in projects that are particularly sought after in the market. Zheng Juncong gives two reasons: first, everyone competing drives up valuations; second, when entrepreneurs are chased, they focus on valuation, feel that investors are all flattering them, and won't settle down to build the company, so the success rate is actually lower. They value entrepreneurs who do solid technology, have patience, want to create value, and don't rely on hype. They have zero tolerance for financial fraud, because Singapore cares a lot about integrity and authenticity, and once fraud is discovered, they will definitely withdraw. This style also explains why they haven't invested in many money-burning internet companies.
— Zheng JuncongIf they could choose again on large models, they still wouldn't invest, because it's like the internet
Over the past four years, the undisputed investment hotspot in China has been AI large models, and Vertex didn't make a move in the first three or four years, which is very unusual among first-tier funds. Zheng Juncong's reasoning: large model startups in China have no advantage; the advantage will come from customers, data, and application scenarios, and it requires being big enough and able to burn a lot of money; it's essentially like the internet playbook, and the business model is still unclear. The two biggest internal voices were "this will definitely take off, China will definitely have large models" and "the model risk is high, it may continue to burn money like the internet, and existing large internet companies have a higher probability of success." He admits this discourages young investors who want to participate, but they held back from investing because they firmly believe hard tech is the advantage.
— Zheng JuncongSingapore certification: swap chips and software, then export
Zheng Juncong believes Singapore can help Chinese entrepreneurs solve two things: product internationalization and product security. Products made with China's manufacturing, design, and innovation capabilities are wanted worldwide, but when geopolitics comes into play, politicians will obstruct them on the grounds that "Chinese products are unsafe, robots might carry things and hit people on the street." Singapore, as a neutral, transparent, and reputable country, can achieve "Citified by Singapore"—chips certified here, or swapped for chips here, software here, and then exported from here, allowing products to enter other countries. He also criticizes "Singapore Washing": companies that move here just to slap on a label, without actual operations and R&D, are the term he least wants to hear.
— Zheng JuncongBoston Dynamics costs over a hundred thousand USD; Wang Xingxing sells for just over ten thousand
In 2019, Zheng Juncong went to Boston for a board meeting and visited Boston Dynamics. He wanted to buy a robot dog, but they said it wasn't for sale, and the cost was about a hundred thousand USD. After returning to China, he judged that China is strong in hardware, so someone must be doing it, and had his team search all Chinese robotics teams, finding Wang Xingxing. At the time, the robot dog was already very agile, could stand up and bow. He asked if it was controlled by AI, and Wang Xingxing said no, it used his own algorithm. When asked about cost, Wang Xingxing didn't mention cost, only said it currently sells for about ten thousand USD. The same dog, others cost over a hundred thousand, he sells for just over ten thousand; Zheng Juncong judged there was commercialization potential, and cost is a very important element in technological innovation.
— Zheng JuncongMissed the follow-on in Unitree: fund expired, not enough bullets
Vertex was the sole institutional investor in Unitree's Pre-A+ round in 2020, but the follow-on investment only came in September 2024, by which time Unitree had already produced the G1 bipedal robot and its valuation had soared. Why didn't they follow on earlier? Zheng Juncong reviews: when they invested, they saw a market of several billion, but didn't see that bipedal robots could be made; in the 2023 round, Matrix Partners and Shunwei came in, and they also wanted to invest, but their USD fund was just at the end of its term and they were raising a new fund, so they didn't have enough bullets to add to the position, and missed that round. He calls it a huge regret, and says this made them reflect; now the new fund requires starting to raise new capital once 65% is invested, reserving 20% to 30% for follow-on investments.
— Zheng JuncongWang Xingxing raised funds slowly, but that honed mass production consistency
Zheng Juncong believes Wang Xingxing doesn't need to be better at fundraising; the current fundraising pace is actually the best. If he had been very good at fundraising back then and raised a lot in the next round, the playbook might have been different—flashy, more money-burning, no discipline in running the company, not nitpicking cost details, and the models produced might not be profitable. After AI arrived, many companies want to copy Unitree, thinking the barrier isn't high, it's just algorithms, and they can catch up quickly with AI. Zheng Juncong says everyone can get algorithms to about the same level, but the real gap is mass production and consistency, which is built up from all past investments and time, and no other company in the market has his mass production consistency.
— Zheng JuncongA crude way to judge a bubble: look at who is starting companies
Asked whether China's embodied intelligence has a bubble, Zheng Juncong says no, because the market is still expanding and in its creation phase. He gives a rough indicator for judging a bubble: when the entrepreneurs you attract are not real entrepreneurs, but people who were in big companies without entrepreneurial spirit, who see others succeed and raise funds, and jump out saying they want to start a company too—that's the point when a bubble appears. Now he sees entrepreneurs who are "I've done this and that," very strong, capable people. He quotes Lee Kuan Yew saying the market needs bubbles, like beer without foam isn't tasty, but too much foam isn't good either; it should be controlled at an appropriate proportion.
— Zheng JuncongIn their own words · checked verbatim
Because chips are called chip CHIP, and then they thought, what chip chip is cheaper than potato chip, even more chip chip than potato chip, so no one invested.
因为芯片叫chip CHIP 然后他们觉得说什么chip chip比potatochip还便宜 就比potatochip还chip chip 就没有人去投
Zheng Juncong7:08
Golden share means I must invest. You have one choice a year. If I use the golden share, I must invest in this company.
goldenshare就是我一定要投 对你一年有一次的选择 如果我使用了goldenshare 我一定要投这家企业
Zheng Juncong18:16
So I think this is a bit like the internet playbook, so we think this business model can still be seen clearly, so we won't invest.
所以我觉得 这个有点像互联网的打法 所以我们觉得 这个商业模式 还能看清楚 所以我们就不会去投资
Zheng Juncong22:19
He said I now sell for about ten thousand USD. He didn't mention cost, because the cost is much lower. Yes, so I thought, wow, amazing, this is China's capability.
他说我现在卖大概一万多美金 他没有说成本 是因为成本更低的多 对 所以我哇 好厉害 这是中国的能力
Zheng Juncong1:02:48
But our fund at the time, I remember, was at the end of our USD fund's term, the tail end. We were raising a new fund, so we didn't have enough bullets to add to the investment in that round, so we missed that round.
但我们的基金呢当时我记得是在经到我们美金的尾期 尾升了 我们正在融新的基金 所以就没有那么多子弹去加码投资在那一轮 所以那一轮就miss掉了
Zheng Juncong1:20:56
When I see that it's not my classmate who succeeded in starting a company, I also want to start one. I happen to have a colleague who also left to start a company. I think I'm better than him, I want to start a company. This is, I think, a bubble, the point when a bubble appears.
当我看到说 不是我这个同学他创业成功了 我也想创业 我刚好我同事也出来创业了 我会觉得我比他厉害 我要创业 这就是 我觉得就是泡沫 泡沫出现的节点
Zheng Juncong1:51:20
But those who can score 10 are investors who can lead the industry to invest in new industries, drive the industry. They see that this industry might be an opportunity in the future, maybe the person is still starting a company, and they go find people to start companies and invest in them.
但能打10分的那些是可以带领行业去投资新的行业 驱动这个行业的投资人 就是他看到可能这行业未来这个是机会 可能连人还在创业 他就去找人去创业去投资的企业
Zheng Juncong2:25:47
Figures
| GIC plus Temasek assets under management | About 1.2 trillion | 10:11 |
| Vertex USD fund term | 10 years, extendable by two years to 12 years | 10:11 |
| Yunyinggu from investment to IPO | 14 years | 10:11 |
| Vertex's USD fund size raised in 2023 | 390 million USD plus 400 million USD | 1:38:13 |
| Threshold for Vertex's new fund to start raising the next fund | 65% invested | 1:21:56 |
| Proportion reserved for follow-on investments in Vertex's new fund | 20% to 30% | 1:21:56 |
| Proportion Vertex's new fund plans to invest in embodied intelligence | At least 50% | 1:52:21 |
| World model teams Zheng Juncong personally visited | 6 | 2:02:33 |
| World model teams the team has looked at | At least 30 | 2:02:33 |
| Zheng Juncong's self-rating as an investor | 7 | 2:25:47 |
Glossary
- golden share
- Vertex's investment committee mechanism: a partner can, once a year, force investment in a project rejected by the IC.
- DPI
- Distributions to Paid-In Capital: the money LPs actually get back divided by the capital invested; the hardest metric for sovereign funds evaluating GPs.
- Singapore Washing
- Refers to Chinese companies that only move their headquarters to Singapore for labeling, without actual operations and R&D.
- eVTOL
- Electric vertical take-off and landing aircraft: a core category in the low-altitude economy, using multiple electric rotors instead of a helicopter's single engine.
- AI for Science
- Using AI to research new materials, chemical structures, and other problems humans cannot calculate.
How to listen
Suitable for founders and investors focused on hard tech, robotics, and the full cycle of USD fund raising, investing, managing, and exiting, especially those who want to know how a sovereign-backed VC makes independent judgments and how it reviews missed follow-ons.
From 27:35 to 33:28, it discusses Singapore clan associations, the Zhongyuan Festival, and Chinese cohesion, which is weakly related to investment judgment.