AI, Investment, Robotics | WAVES New Wave 2025
WAVES New Wave 2025 invites you to embark on the "New Era" of China's venture capital.
This is the new era of China's venture capital. Currently, the Chinese venture capital market is not only at the turning point of the cycle bottoming out but also in the in - depth period of structural transformation. In the new ecosystem dominated by policies and with high concentration of state - owned assets and capital, only by conforming to the trend and making flexible adjustments can we seize certain opportunities in the face of uncertainties.
From June 11th to 12th, at the Liangzhu Culture and Art Center in Hangzhou, the 36Kr WAVES New Wave 2025 Conference, themed "New Era", will gather top investors in the venture capital field, founders of emerging enterprises, as well as scientists, creators, and scholars deeply involved in technology, innovation, and business. They will jointly discuss cutting - edge issues such as AI technological innovation, the wave of globalization, and value re - evaluation, dissect their views on business ideals and the future world, and jointly discuss, search for, and move towards the "New Era" of China's venture capital.
On the morning of June 11th, a themed round - table discussion on AI, investment, and robots was held in the investors' venue. Wang Chenhui, Managing Director of Volcanic Stone Capital; Wang Guolong, Partner of Inovance Industrial Investment; Cao Fangning, Founding Partner of Multidimensional Capital; Hu Zhu, Partner of Ginkgo Valley Capital; Luan Yiming, Partner of Shunchuang Industrial Investment Fund; and Zhao Peizhou, Partner of Xiaomiao Langcheng participated in the discussion. This round - table was hosted by Shi Jiaxiang, an author from Anyong.
The following is the transcript of the dialogue:
Shi Jiaxiang: Today's panel is a bit crowded. From the number of people on our panel, we can see the importance of embodied intelligence and AI hardware in the current primary market. We've selected all of you here because you've discovered star projects. So, I'd like to invite each of you to introduce yourselves first, including your institution and your recent areas of focus.
Wang Chenhui: Hello everyone. I'm from Volcanic Stone Capital. In 2016, several partners from IDG founded Volcanic Stone Capital. Robotics has always been an investment direction for our team. For example, our team was the only institutional investor in Ecovacs before its listing.
At Volcanic Stone Capital, we've invested in Geek+ for commercial robots, Fourier Intelligence for general - purpose robots, and some earlier - stage companies, such as GJ Robot for entertainment and consumer robots and Jiaao for medical robots. Robotics is an area that our fund has been focusing on and investing in for a long time.
Back to today's topic, currently, people are quite concerned about embodied intelligence. Our views on this are clear. We believe that the entire field of embodied intelligence has not yet reached its "iPhone moment". It is now in a period of rapid development of technological iteration and commercial scenario implementation. For such companies, we focus on three points: the ability to form a data closed - loop, the ability to reduce hardware costs, and the ability to implement commercial scenarios.
Wang Guolong: Thanks to 36Kr for the invitation. I'm very glad to be in Hangzhou today. I graduated from Zhejiang University, so coming to Hangzhou is like being in a semi - home field for me. I'm Wang Guolong from Inovance Industrial Investment. We are the only CVC platform under the listed company Inovance Technology. We were established in 2017 and started operating as a fund in 2021. Since our establishment, our investment direction has been quite focused. One is around Inovance's business needs, and the other is around strategic layout directions.
Based on this, we've derived corresponding investment directions and made orderly adjustments according to the development and changes in various industries. First, the upstream of Inovance, including chips, components, and sensors. Second, the downstream of Inovance covers all industries, so all kinds of advanced manufacturing equipment are in our area of concern. Third, Inovance United Power is the absolute leader in the third - party automotive three - electric sector, so we've also made in - depth investments around key automotive technologies and materials. Fourth, we believe that the future will be software - defined industry, so we have systematic coverage and investment in the fields of industrial software and AI. In addition, we're also continuously exploring relatively cutting - edge directions. For example, around the low - altitude economy, we recently completed the acquisition of a company focusing on upstream avionics and flight control software and hardware systems.
Back to today's topic of embodied intelligence, for Inovance, we can self - manufacture 50% - 60% of the BOM at the level of the embodied hardware body. Inovance's humanoid robot team is also positioned as a supplier of core components for the embodied body. At the investment level, we've previously invested in six - axis force sensors, which are also key components of the embodied body. In addition, we've been continuously concerned about the brain, cerebellum, dexterous operation, and lower - limb control.
Cao Fangning: Hello everyone. I'm Cao Fangning, the founder of Multidimensional Capital. Multidimensional Capital has been operating in the industry for ten years. Since its establishment in 2015, we've completed approximately 300 rounds of financing for more than 200 companies, with transactions close to $50 billion. Our overall style is relatively more focused on the mid - to - late stages, and we cover a wide range of industries, including semiconductors, aerospace in the past, and embodied intelligence, AI, and healthcare today. Regarding AI hardware and embodied intelligence, from last year to this year, we've served about a dozen leading companies, such as Songyan Power and Hangzhou's Weifen Zhifei. In the field of AI glasses, we've served Rokid, as well as a series of upstream and downstream enterprises in the industrial chain, such as Lingchu and Daimeng. So, we've accumulated some experience and industrial knowledge about the current dynamics of the primary market for AI hardware and robots, and I'll share our views with you later.
Hu Zhu: Hello everyone. I'm Hu Zhu from Ginkgo Valley Capital. Our institution was established in 2013 and focuses on investments in cutting - edge fields such as semiconductors, humanoid robots, biomedicine, and artificial intelligence. If any of you here are entrepreneurs in these fields, you can come to me. We're willing to be your companions on the entrepreneurial journey. Thank you.
Luan Yiming: Hello everyone. I'm Luan Yiming from Shunchuang Industrial Investment. Compared with all of you here, our institution is much younger. It was established in 2023. So far, we've set up 11 funds with a total scale of about 3 billion yuan. Our main investment directions are related to the nature of our company. Our institution is a state - owned enterprise in Shunyi District, Beijing, a proper state - owned enterprise. On the one hand, most of the money we manage comes from state - owned assets in Shunyi District, including the self - owned funds of state - owned enterprises and various levels of fiscal funds in Shunyi District. In addition, there is also some money raised from the market. Our investment directions are mainly the leading industries in Shunyi District, such as new energy vehicles, aerospace, third - generation semiconductors, biomedicine, and high - end robot manufacturing.
Some of our previously invested projects may be quite relevant to today's topic. For example, Mainline Technology, which focuses on autonomous trucks, and Yue Shi Robotics, which is engaged in cold - chain logistics. These enterprises that have industrial synergy with Shunyi District are definitely of great interest to us. If there are enterprises whose directions are in line and are considering setting up in Shunyi District, Beijing, as a state - owned enterprise in Shunyi District, Beijing, after becoming partners, we're very willing to help enterprises empower their industries as much as possible.
Zhao Peizhou: Hello everyone. I'm Zhao Peizhou, a partner of Xiaomiao Langcheng. Xiaomiao Langcheng is a market - oriented institution under the Shanghai Zizhu High - tech Park. Since our establishment more than 10 years ago, we've invested in more than 130 early - to - mid - stage hard - tech enterprises. Our main investment directions are two major sectors: artificial intelligence and advanced manufacturing. In particular, in the future investment direction of our fund, artificial intelligence will still be one of our most important investment tracks.
Regarding the field of artificial intelligence, we mainly focus on four major sectors. One is infrastructure, including some peripheral facilities related to data centers and computing power chips. The second is humanoid robots and embodied intelligence, including the entire industrial chain. The third is various AI applications. The fourth is AI - driven intelligent hardware.
In the past, we've invested in about 5 enterprises in the general embodied intelligence and humanoid robot body, including the industrial chain. They are Fei Xia Robotics, Qiongche Intelligence, Songyan Power, West Lake Robotics, and Qianjue Robotics. Fortunately, we invested in these 5 enterprises at a relatively early stage. After our investment, they've all completed at least two rounds of financing.
People label us as an early - stage investment institution. For emerging industries, especially the new industries driven by artificial intelligence now, we hope to be the first - round investment institution for entrepreneurs. Of course, for enterprises in the development or maturity stage, we'll also make investment layouts covering the mid - to - late stages and even the entire stage.
Shi Jiaxiang: We all know that embodied intelligence is a field with high technical complexity and great challenges in scenario implementation. May I ask how you balance technological foresight and commercialization in investment?
Wang Chenhui: For a company, technological leadership and commercial implementation are like two legs of a person. You can't walk with only one leg. So, at a certain stage, technology may lead commercialization, but ultimately, it has to serve commercial scenarios.
Taking the field of embodied intelligence as an example, we now see two scenarios for embodied intelligence. One is for specific scenarios, which we call professional intelligent robots, and the other is consumer - grade robots for general scenarios. The technologies they need are different from the current mainstream technical solutions. The former focuses more on scenarios, needs to find suitable specific scenarios in industrial manufacturing or commercial services, provide specific products, and even needs to customize hardware and models to a certain extent. So, the intelligence level is positively correlated with the complexity of the scenario.
Back to the model, for example, the so - called hierarchical model in the market before needs to separate the control of the brain and the cerebellum. It's more suitable for technologies that require modularization, strong interpretability, and less generalization ability. The latter, the consumer - grade humanoid robots for general scenarios, are designed to solve long - term complex tasks in business or consumer scenarios and require natural human - machine interaction. So, such scenarios require models with stronger generalization ability to handle multi - modal input, precise interaction force control, etc. So, it's indeed more suitable for end - to - end models like Google's RT series. This kind of model requires a large amount of data input and strong computing power. But from the current technological implementation perspective, we think it still needs a 3 - 5 - year technological verification cycle. So, I think it still comes back to the point that technology ultimately has to serve commercial scenarios. It depends on what scenarios you want to target and then choose the matching technical route.
Wang Guolong: Whether it's embodied intelligence or other technology - startup enterprises, they usually follow the principle of selling one generation, researching and developing the next generation, and pre - researching the generation after that, which basically corresponds to the stages of market promotion, customer verification, and concept verification. For startups, technological innovation ability must be the core competitiveness. Technological foresight is more about making reserves to maintain the company's continuous iteration ability. However, when facing actual user scenarios, in fact, startups need to deliver end - to - end solutions to users. Taking the upper - limb operation of the embodied body as an example, the technical routes include the VLA hierarchical model and reinforcement learning. However, there are actually seemingly traditional but relatively robust technical routes, such as MPC and PID control. So, in the product R & D stage, it's necessary to maintain foresight, but after entering the real scenario, various methods are needed to solve practical problems.
Cao Fangning: I really like to compare this wave of opportunities in embodied intelligence with the previous wave of opportunities in autonomous driving. We saw that during the entrepreneurial wave of autonomous driving ten years ago, the key factors for success included both the ability to understand the foresight of technological R & D and the flexibility of the team in real commercial solutions. At the same time, being able to cooperate with commercial scenarios better based on the direction of industrial upgrading was also essential. These two aspects are indispensable. In the future development of this wave of embodied intelligence startups, a similar logic will be followed. That is, your R & D and technological thinking must serve your commercial implementation. Of course, in the early stage, the scenarios you choose and the partners you choose will determine your technical R & D direction. We saw that in the past, in the field of autonomous driving, many companies have emerged in technologies such as fusion perception, lidar, and visual computing. I believe that the same will happen to embodied intelligence projects. No matter what technical direction you choose today, when your scenario application solution matches your technical direction, your solution will ultimately succeed.
Hu Zhu: I'll use a case to discuss the topic of "whether the technical solution of an embodied intelligence project needs a clear implementation scenario". This case is also one that General Manager Cao participated in. Weifen Zhifei is a project led by Professor Gao Fei from the Control College of Zhejiang University. When we made the investment, it was based on our long - term and systematic tracking of the dynamics of the Control College and the Computer College of Zhejiang University. When we co - created this project with Professor Gao Fei, it was even praised by Elon Musk. The core technology of the project is the ability of drone swarms to achieve autonomous control in GPS - denied environments. As for whether it has found its application scenario at present, I think we shouldn't set a limit for it. Usually, precisely because there is no preset application scenario, it can find a broader development space in the future, its own vast expanse of stars and seas.
Luan Yiming: Regarding the AI direction of robots and embodied intelligence, we think there are three dimensions for a project. One is whether the technology itself is advanced. The other is whether the team has the ability to engineer the technology. After the engineering ability is in place and the product is made, is there anyone willing to pay for it? How is the commercial implementation? We may look at it from these three dimensions. If we look at it in sequence, if the technology is well - developed, the product can be made, and then we go looking for someone to buy it, it's inevitable that we'll be in a situation where we see every problem as a nail because we have a hammer. The company thinks it can apply to many scenarios, but in fact, during the commercial process, the enterprise can't spare so much energy to find the corresponding scenarios and fully promote their implementation. Especially when the enterprise thinks it has strong reactivity and general applicability, the challenges it faces are also great.
So, in this kind of direction, we prefer to see some enterprises that are already doing commercialization and then come back to look for robot - related technologies and scenarios. Or, we provide good scenarios and you solve specific problems for us. This kind of thing is what we'd like to see. Of course, maybe because of our own preferences, we may prefer that the product has been implemented commercially and there are people willing to pay for it. In this way, we can control the risk. Although it will limit our expected profit return, for risk, we think it's relatively controllable, and it will be easier for enterprises to get funding. Overall, this is the general trend. Whether it's embodied intelligence or AI, considering the manufacturing industry more, we still attach great importance to whether it can solve some specific problems.
Zhao Peizhou: This question is about the balance between technology and commercialization. From a macro perspective, we should do as much commercialization as possible, but we still need to look at the details and combine different industries. We mainly need to split and look at the two industries of humanoid robots and general embodied intelligence.
Humanoid robots are hardware - oriented entities. In the past, the technical barriers of humanoid robots were high, and there were relatively few people researching this industry. Now, it has become a huge trend in society and among capital. We've noticed that whether it's hardware technology, joint technology, or cerebellum control, motion control, and the threshold of reinforcement learning have all been significantly reduced. It's expected that in two years, there will be a large overflow of talent, and the hardware threshold barriers will also be significantly reduced. So, for humanoid robot enterprises, commercialization is a must. To put it bluntly, if they can't generate hundreds of millions of revenue during this wave of dividends this year and next year, and transform the initial barriers from joints and motion control to the next wave of higher - end technologies, break through the scenario barriers and achieve a certain scale, these enterprises will be out.
Regarding general intelligence, I hold the opposite view. I think we should do subtraction in commercialization because general intelligence must generate value. The value driven by this wave of AI must be that a robot, no matter what form it takes, whether it's a wheeled robot with a robotic arm or a humanoid robot, can perform different