AI Investment Should Focus on Fundamentals, Technology and Applications - AI Cannot Replace Investors
-
On September 20-21, 2018, the 12th China Investment Annual Limited Partner Summit, hosted by ChinaVenture Information and ChinaVenture Capital and co-organized by ChinaVenture, was held in Shenzhen. The summit, themed "Investment Turning Point," invited renowned limited partners (LPs) such as leading fund-of-funds, government-guided funds, bank asset management institutions, insurance capital, listed companies, and family funds, as well as top general partners (GPs) from China, to gather and deeply analyze the latest changes and trends in the industry. The event aimed to provide insights into the future direction of capital and build a top-tier platform for efficient GP-LP connections, communication, and fundraising.
AI's industrial applications and scenario implementations determine the future development of the AI industry. During the roundtable discussion titled "AI+: The Driving Force of the Future," ChinaVenture invited Mr. Feng Jie, General Manager of Leaguer Capital; Mr. Jin Chen, Managing Partner of Ventech China; Ms. Lan Lan, Partner of Shanghai Jinpu Investment; Mr. Liu Ming, Vice President of Hengtian Zhongyan; Ms. Li Fanghong, Partner and Managing Director of Chuangjing Capital; and Ms. Zhang Zhiwu, CEO of Northlight Photonics, to discuss the development of the AI industry. Mr. Guo Libo, Dean of ChinaVenture Research Institute, moderated the discussion.
The guests agreed that finance, healthcare, security, and retail—industries with substantial data reserves—are the four major scenarios for AI implementation. From an investment perspective, strategic layouts should be made at the fundamental, technological, and application levels. In the investment field, AI may provide data-driven assistance and references, but it will take time before machines can replace human investors.
Below is an excerpt from the roundtable discussion, edited by ChinaVenture:
Guo Libo: I am delighted to host the final session of today's AI forum. In the previous forum, we invited six investment experts to share their insights on AI investment trends and logic. For this session, we are honored to have five investment leaders and one AI entrepreneur to discuss AI empowerment and the future of "AI+."
Feng Jie: Good afternoon, everyone! I am Feng Jie from Leaguer Capital. As a Shenzhen-based VC with Tsinghua background, we have been focusing on early-stage technology-driven investments for nearly 20 years, primarily in electronics, materials, healthcare, advanced manufacturing, and new energy. Over 19 years, we have evolved from a pure VC into a comprehensive tech service provider integrating investment, industrial parks, and startup incubation. I look forward to further exchanges with you all. Thank you!
Jin Chen: I am Jin Chen from Ventech China. Originating in France, we are an independently funded and operated USD fund in China. Our investment focus includes internet, big data (covering AI, FinTech, and consumer sectors). We are long-term optimists about AI and have invested in AI and big data companies over the past two years, including L4 autonomous driving solutions, unmanned convenience stores, data visualization, intelligent databases, and sales lead generation tools. It’s a pleasure to be here today. Lan Lan: Hello everyone, I'm Lan Lan from Jinpu Investment. The Shanghai Financial Development Investment Fund was established in 2009 with approval from the State Council and National Development and Reform Commission. It was the only financial industry fund at that time primarily focused on and named after the financial sector. Our first phase had a scale of 11 billion yuan, completed in 2012, with investments in projects like China UnionPay, Guotai Junan Securities, and Industrial Bank. Our second phase is managed by a subsidiary of Shanghai Jinpu Investment, mainly investing in fintech projects including Ant Group. This year, we will continue to focus on industries empowered by artificial intelligence, including fintech, smart manufacturing, and smart communities. Thank you!
Li Fanghong: Hello everyone, I'm Li Fanghong from Chuangjing Capital. Chuangjing is a venture capital institution focused on growth-stage investments, covering sectors such as semiconductors, enterprise service software, education, medical services, and medical devices. Our past investments include companies like DJI and Qihoo 360. I'm very happy to have the opportunity to exchange ideas with colleagues today. Thank you!
Liu Ming: Hello everyone, I'm Liu Ming from Hengtian Zhongyan. Hengtian Zhongyan is a leading asset management institution in the Chinese market. I am primarily responsible for all equity investment-related businesses, including equity investment fund-of-funds and several primary equity investment funds. Our core fund-of-funds mainly focuses on sectors that will represent China's economic transformation in the next 5-10 years, including healthcare, consumer goods, high-end manufacturing, education, and TMT. Our investment approach primarily involves investing in funds, supplemented by a small number of co-investments, mainly using the P+S+D method. Within the TMT sector, we pay close attention to areas with core technologies, new technologies, new models, and new application scenarios, such as artificial intelligence. I'm very pleased to participate in today's AI+ themed forum organized by ChinaVenture. Thank you!
Zhang Zhiwu: Hello everyone, I'm Zhang Zhiwu, founder and CEO of Beike Tianhui. Our company specializes in LiDAR technology, starting in 2005 as China's first LiDAR enterprise. Initially, LiDAR was used for high-precision measurement and mapping. In recent years, with the rise of autonomous driving and smart mobility, LiDAR has gained more attention. Since 2005, Beike Tianhui has faced international competition, as there were no other domestic LiDAR companies in the first decade—we competed with foreign companies for market share. Now, there are more LiDAR companies, making competition fiercer, but there are also more opportunities. Our headquarters is currently in Beijing, with our production base in Suzhou. This year, we established a software R&D center in Hefei, where we also located our chip R&D center. We are the first company globally to develop our own chips for LiDAR signal processing, starting our research in 2009 and applying the chips in 2015. Currently, we utilize many semiconductor technologies in LiDAR production. Our clients include leading automotive companies from the US, Germany, and Japan, as well as emerging AI companies, such as those in unmanned logistics and vertical scenarios like shared mobility. We are delighted to have this opportunity to share with everyone.
Guo Libo: Thank you, Mr. Zhang, and all the guests for your introductions. Let's move on to our first topic. Today, we mainly discuss AI+, which evolved from the concept of 'Internet+'. 'Internet+' has already flourished across various sectors of life, including education, healthcare, finance, etc. My question to our guests is: In which fields can AI+ be applied, and which fields hold the most promise?
Security, Finance, Healthcare, Retail: The Four Major Scenarios for AI
Feng Jie: Looking back at the AI field, there was a wave of enthusiasm decades ago when people believed the era of AI would arrive. Any new technology often starts with a lot of hype, followed by a period of low interest. Personally, I think the recent resurgence of AI is closely related to the internet era, which has generated massive amounts of data. Additionally, advancements in algorithms, hardware, and data processing capabilities have laid a solid foundation for AI applications.
We have been working on AI for a long time, progressing from the early to the intermediate stage. Over a decade ago, when I was in university, there was a focus on business intelligence, which involved using analytical tools to process commercial data for better decision-making—an early model or category of AI. As technology matures, AI tools and methods can liberate humans in many fields, potentially performing better than us. Take autonomous driving as an example: in the future, AI might be a safer option due to human factors like emotional fluctuations and fatigue, barring hacker interference. In industrial manufacturing, AI tools can enhance efficiency in executing plans and managing progress.
Currently, companies like Lihe are investing in AI applications in healthcare, security, and other fields, including early-stage foundational programs and chip development. Overall, I don’t believe AI companies focused solely on tools have great prospects. Instead, those integrating AI with specific scenarios will thrive. The same applies to the internet—pure internet tools have value, but the most successful internet companies use the internet to conduct business and transform industries. Similarly, AI and AI companies leveraging AI tools to conduct business in broader fields may have brighter prospects.
Guo Libo: Could you provide some examples? Feng Jie: Many unicorns are entering the security sector because the need for massive data analysis is particularly suitable for AI applications. Traditional security firms are also adapting through integration. Industries with vast datasets requiring long-term training—like security, healthcare, and finance—show great potential for AI applications. These have been hot investment areas in recent years due to visible opportunities.
Jin Chen: We view the AI industry chain in three layers: 1) Infrastructure (hardware, chips, sensors, cloud), 2) Software/algorithms, and 3) Applied technologies. "AI+" essentially combines AI solutions with daily or commercial scenarios. For example, our investment BingoBox is an unmanned convenience store leveraging AI for facial recognition, product placement optimization, and checkout. Its loss rate is negligible—if a customer forgets to pay for one item, backend systems identify it immediately, and staff follow up via WeChat Pay with 99% success. Traditional stores can't match this level of AI-driven efficiency and short-term ROI.
National Wave: As a USD fund manager, what differences exist between USD and RMB investments in AI?
Jin Chen: No major barriers exist. Some projects start with RMB structures, making later USD participation harder. YinTai Capital invests early-stage (Series A/B) and can adapt to VIE structures if needed.
Ant Financial: More Than AI, a Tech Service Provider
National Wave: Lan Lan, since Jinpu invested in Ant Financial—an AI-driven company—what was the investment logic? Which industries do you see as most viable for "AI+"? Lan Lan: We know that data is the cornerstone of AI technology. In the process of applying AI technology in real-world scenarios, we focus on two key points: First, whether there is high-quality and large-scale data—both quality and quantity are crucial—and whether this data can streamline the entire workflow and be updated regularly. Second, whether the specific application domain has clearly defined boundaries. From an industry perspective, the financial sector aligns well with these characteristics, which is why AI applications initially flourished in finance. The financial industry has vast amounts of accumulated customer data, and it faces significant challenges such as credit security, fraud prevention, and risk control.
When evaluating Ant Group, we see that it has evolved beyond FinTech, now positioning itself as a 'Technology-as-a-Service' enterprise. This global tech service provider role means our due diligence and investment focus more on their future technological output. This output implies they will gradually reduce their involvement in lending and wealth management businesses, shifting toward becoming a pure technology provider. Their AI-driven solutions for risk control, credit scoring models, and anti-fraud will be offered to other institutions for collaboration.
We also invested in Suning Financial Services, which shares a similar concept. The key difference is Suning's offline physical scenarios, such as facial recognition in Suning stores, which generates massive data for conversion and traffic enhancement. We prioritized these two companies because they are pioneers in applying AI, especially in the AI + finance domain.
This year, Jinpu will establish a new Smart Tech fund, focusing on AI-driven industries like smart communities, smart manufacturing, and foundational technologies for other application scenarios. We believe smart communities, smart manufacturing, and IoT will be major trends in the future.
Foundation, Technology, Application: The Logic of Fund Investment in AI
National Bo: Thank you, Lan Lan, for sharing. Li Zong, what are your thoughts on AI+?
Li Fanghong: We consider several factors. The industry must have massive data as a foundation, followed by suitable algorithms. Algorithms extract features from data to form patterns—similar to Newton's laws, where complex derivations lead to simple, elegant formulas. The industry must exhibit this characteristic, where data-driven induction and summarization yield clear logic. In the process of algorithm development, some hardware assistance may be leveraged, such as chips. Whether the cost of these chips is controllable and whether they can fully unleash the efficiency of algorithms are key considerations for us. In the scenarios mentioned by several executives earlier, we've observed some major directions, including autonomous driving, new retail, and healthcare.
Within these areas, there are actually some specific segments that can be broken down. For example, in medical imaging, pathological slice images and X-ray images require different technical applications. Similarly, autonomous driving involves complex decision-making processes, including perception and decision-making. During the perception phase, different hardware and algorithm implementations are needed. Currently, we're focusing on highly specialized industrial applications. For instance, high-end machine tools use extremely expensive cutting blades that wear out after some time, leading to production halts and significant costs for enterprises. By applying AI, we can make some predictions to avoid such losses.
Guo Libo: Mr. Liu, as you mentioned, Hengtian Zhongyan operates as a fund of funds while also making direct investments, including some AI unicorn projects. Could you share your investment logic and your perspective on the future roadmap of AI+ development?
Liu Ming: Our investment logic primarily involves selecting major industries and sectors. After all, as a fund of funds, we start from a broad perspective. We are optimistic about the entire AI field over the next 5-10 years. Equity investment is a long-term behavior, so we need to ensure the sector has strong development opportunities within this timeframe before making our layout.
In the actual investment process, we first focus on identifying the GPs with the strongest investment capabilities in this field for our investment layout. Simultaneously, we collaborate with these top GPs to explore underlying direct investment opportunities. In China's AI sector, development has entered a period of rapid growth in recent years. Statistics show that the combined scale of AI enterprises in Beijing, Shanghai, and Shenzhen accounts for nearly 6% of the global AI market - already a significant figure, with rapid growth.
In this process, we categorize AI enterprises into three types: foundational layer AI companies, technological layer AI companies, and application layer AI companies. For the technological layer, we focus on excellent targets centered around chips and AI development foundations, particularly in big data and cloud computing - the two core technological ends. For the application layer, we primarily look at AI technology companies with technology and algorithms at their core, such as China's prominent AI companies like Megvii and SenseTime, which combine both technological and application aspects. We monitor these according to different categories.
Through fund investments, we've laid out in companies like Cambricon specializing in chips, as well as those in big data and cloud computing, including Chinese cloud providers beyond Alibaba, Tencent, and Baidu, such as Kingsoft Cloud. Regarding AI+ scenario applications, as the host mentioned, it's quite similar to Internet+. Currently, the more successful and promising AI+ sectors with near-term breakthroughs include AI+ finance, as several guests have mentioned. For example, nowadays everyone uses mobile banking, including China Merchants Bank's mobile banking, which incorporates many AI features, such as robo-advisors. The initial stages will continue to improve gradually. For instance, AI + logistics is a hot topic, leading to smart logistics. There's also AI + healthcare. The advancement of these technologies has significantly improved areas like diagnostics.
From 2010 to 2016, the accuracy of machine recognition increased from 70% to 95%. We believe that as this accuracy continues to improve, AI + healthcare will have even more application prospects. A guest from Lenovo Capital also talked about smart cars, which involve a lot of autonomous driving technologies, including various sensors and automatic recognition, all requiring high-level AI applications. These scenarios have great potential for implementation in the next 5-10 years. At the same time, we not only focus on these scenarios but also pay attention to the underlying core infrastructure and technological implementation, which we have never overlooked. Beyond scenario applications, we have always been attentive to investment opportunities in technology.
The price of autonomous driving sensors will drop significantly.
Guo Libo: Thank you, Mr. Liu, for your technical insights. Mr. Zhang, could you share your understanding of AI + in your industry and the future trends of AI + in your field?
Zhang Zhiwu: Our industry is closely related to autonomous driving and mapping. As the ancients said, being smart means having sharp ears and keen eyes, which cannot be achieved without sensors. The role of sensors is to collect and accumulate massive amounts of data. From our observations, the sources of data in the future will be very diverse. For example, high-precision map data is currently collected by professional mapping companies. However, with the widespread use of sensors, roads that once only had streetlights are now filled with cameras.
Yesterday, at Alibaba's Yunqi Conference, I talked about the integration of roads and vehicles, such as installing LiDAR on roads. In the future, high-precision maps might come from traditional mapping or even vehicles themselves. Currently, we are working with a British car manufacturer to equip vehicles with six LiDAR units. In the future, vehicles like those from Uber or Didi might also be equipped with various LiDAR, millimeter-wave radars, and cameras. Thus, our lives will be filled with sensors, which means abundant data sources. All this data will be integrated and serve as a driving force.
Right now, sensors are expensive, and many people say LiDAR is too costly. The truth is, none of the raw materials for LiDAR components are particularly expensive. The real cost lies in R&D. For example, we have our own chips for LiDAR processing. Because the production volume is not very large and orders are limited, the cost per chip is high. However, when we produce millions or tens of millions of LiDAR units, the cost per chip could drop to just a few dollars. Based on this logic, we predict that LiDAR will not be expensive in the future. Currently priced at over 100,000 RMB, the cost of LiDAR will significantly decrease as production scales up. I tell users not to dismiss LiDAR just because it's expensive now—if you wait until it becomes affordable to develop your algorithms and solutions, you might already be behind. AI is a highly dynamic process, from data collection and processing to algorithms and solutions. In this industry, the bottleneck of BOM costs will soon be broken, accelerating AI advancements beyond our expectations—even beyond what professionals in the field anticipate. The application scenarios are evolving faster than imagined, with breakthroughs emerging rapidly across many sectors.
AI Cannot Replace Investors Yet, but It Can Assist in Data Analysis
Guo Libo: Thank you for sharing, Mr. Zhang. AI + investment hasn’t been mentioned yet. Nowadays, more and more secondary market investments, including primary market robo-advisors, are leveraging AI. Goldman Sachs, for example, has over 10,000 employees, with at least 30% working in technology. We’ve learned that Google also has a product using AI to assist in portfolio decision-making. Since many of us here are investors, how do you think AI will impact our industry, especially secondary market investments?
Feng Jie: As Mr. Guo mentioned, the secondary market deals with relatively standardized products, making electronic trading with big data an excellent foundation for AI applications. In the primary market, however, much information has traditionally been non-standardized or non-public, often inaccessible through conventional channels. With the increasing digitization and internet penetration of society, including past behaviors of companies—especially the founders’ actions and life trajectories—much of this data is now traceable.
A foreign scientist once said that if one could access all your personal life information, they could predict your future behavior, as many actions follow logical patterns. A VC in Silicon Valley has fully adopted AI-driven investment, relying on models and data to learn from all available books in Silicon Valley—investing in anyone who asks for funding.
In China, however, AI is currently more applied in data collection and process automation. The core of investment still heavily relies on human judgment, and there aren’t yet effective AI tools to assess people based on data. As investors, we must recognize that AI is the future. For our industry, we should be cautious—the venture capital sector might one day be disrupted by AI. We should embrace and integrate AI technology into investment practices sooner rather than later.
Guo Libo: Anything is possible. I understand dollar funds differ from RMB funds, as the former don’t employ as many analysts. Mr. Jin, do you think analysts could be replaced by AI? Could investments someday be made solely by partners and machines? Jin Chen: I hope so, as Director Guo said, investment work is quite demanding. This industry is full of smart people. Our approach is to spend a significant amount of time reviewing every project in a field once we identify it, even if 90% of them won’t be considered for investment. So, our days are indeed quite tough. However, by doing this, we ensure the quality of our project investments. Based on this principle, we can, to some extent, automate the analytical work.
As Director Feng mentioned earlier about assessing people, we currently rely mainly on conversations, background checks, and various verifications. Conversely, the team aspect is where we struggle the most. Everyone says investment is about betting on people, and when we review our past investments, the best-performing projects overwhelmingly have outstanding teams. Yet, during the investment process, especially for early-stage projects, we only meet the team a few times, making it difficult to determine whether they are truly exceptional.
However, with the application of big data, people’s information will become more comprehensive in the future. Through analytical algorithms, we may predict 60-70% of their next steps or estimate their success rate in a given endeavor. I believe this is possible. From a holistic perspective, I hope that one day, a robot and I can jointly take responsibility for managing a fund.
Lan Lan: I’ve looked into some AI + investment projects. Currently, some initiatives specifically target the primary market, using AI to help analysts better manage funds and gather industry insights. We’ve reviewed a few projects that perform quite well, and major investment institutions have already adopted them. We’ve also implemented such a system, charging an annual fee. AI + investment helps free up manpower—particularly for technical tasks.
As Director Jin mentioned, investing in a project largely depends on evaluating people. The investment process is incredibly time-consuming—it’s not just about gathering industry information, rankings, or data. We gain a lot of insights by interacting with management and employees, combined with our own investment experience. So far, AI cannot replicate this.
However, for basic analyst tasks—such as collecting industry reports, financial data, and other public information—AI can certainly assist. This helps us save manpower, improve efficiency, and deliver better service. I recall a saying: AI doesn’t replace anyone; it makes your work more efficient, allowing you to focus your time on more meaningful tasks.
Guo Libo: During the investment process, Director Li, have you consciously or unconsciously used related technologies to enhance your investment capabilities? Are you applying such methods in decision-making now? Li Fanghong: Some aspects of the investment sector have achieved a certain degree of artificial intelligence. For instance, quantitative funds in secondary markets could be considered a form of AI, as they meet certain criteria by processing massive datasets with proprietary models and algorithms.
However, we've encountered obstacles in primary market investments. First, as mentioned by others, data-related challenges persist. We've experimented with various databases but consistently face issues of incompleteness and inaccuracies, making data-driven decisions problematic. Second, many project elements resist quantification—like key legal risk points in investment terms. These currently defy AI implementation, often requiring iterative negotiation between parties.
Guo Libo: President Liu, your operations are more complex. While not yet at the AI stage, are you employing advanced methods in investment management?
Liu Ming: Currently, AI+ only provides auxiliary functions. I don't foresee AI assisting primary market investment decisions in the near term. Equity investments here are probabilistic—early-stage strategies like 'investing in people' remain valid. However, corporate development rules gain importance across stages. For early-phase projects, success hinges on countless micro-decisions (left or right turns) that shape growth trajectories. While AI can support, final investment calls require human judgment.
Guo Libo: Still premature. President Zhang, as a fundraising company, what investor profiles do you prefer?
Zhang Zhiwu: As investees, we increasingly value investors' advisory role in shareholder/board meetings. Their strategic insights—especially regarding major decisions—complement our manufacturing-focused, cost-conscious mindset. In today's landscape, certain initiatives demand big-picture thinking. Investors help us pinpoint strategic resource allocation with surgical clarity. When selecting investment institutions, it's a process of mutual understanding and communication. I believe it's crucial whether the investment institution truly understands the industry. Secondly, whether they grasp the industry's development patterns is equally important. If they are knowledgeable about the industry and its trends, beyond just capital, their intellectual input can have an intangible impact on us. Additionally, having some strategic resources would be perfect. These are my expectations for investors.
Mid-tier AI Companies Have Lower Valuations and Are Worth Attention and Investment
National Bo: Thank you, Mr. Zhang, for sharing. Time is almost up. Since this is an LP summit with many LPs present, here’s a final simple question. Data suggests that the next 1-2 years will be a challenging adjustment period for the VC/PE industry, which may trigger a series of chain reactions on the valuations of downstream companies. What advice do the panelists have for LPs regarding AI investment in the next 1-2 years?
Feng Jie: In investment, especially venture capital, to some extent, you always need to chase the fastest-growing companies. If you believe AI is the future trend and a tool that will be used across all industries, then this sector is worth continuous investment. However, since AI is highly technology-driven, any fund investing in AI should not focus solely on late-stage investments. It’s essential to have strong sensitivity and attention to early-stage opportunities. That’s my advice.
Jin Chen: In the AI sector, top-tier companies have very high valuations, sparking discussions about an AI bubble. But beyond these top players, we see many mid-tier, tail-end, or early-stage companies in the AI field with relatively low valuations because growth in this sector is slower compared to 2C companies like Pinduoduo. If there are concerns about China’s economy over the next two years, the expectations for growth may decrease, but the demand for efficiency improvements will rise significantly. In this context, AI has great potential. From this perspective, both GPs and LPs should pay more attention to the development and investment opportunities in the AI field.
Lan Lan: I agree with Mr. Jin’s view. When investing in AI technology companies, we must avoid the bubble and focus more on their technology and the feasibility of their business models. In currently hot sectors like visual recognition and image recognition, valuations may already be high. Instead, we can explore other sub-sectors or vertical AI technology companies to find projects with better cost-performance ratios. The key is to look broadly, think deeply, and not limit ourselves to just one or two industries.
Li Fanghong: The ability to price assets accurately is crucial, whether for AI companies or other industries. Once we can more precisely determine an asset’s value, patience is often required. When an industry is affected by economic fluctuations, it’s not necessarily a bad thing. Be patient and wait for prices to reach a reasonable range before investing. Liu Ming: As we are in asset management, my suggestion is that LP investment decisions must be diversified—not concentrated in a single project, investment stage, or sector. The second piece of advice is that for non-professional investors, finding reliable GPs and competent managers is crucial. If you are optimistic about an industry, such as AI, invest with the GP who has the strongest investment capability in that sector. Third, even when selecting GPs or managers, diversification is key, which is why we operate through fund-of-funds. In early-stage sectors in China, the core of investment strategy is diversification.
Zhang Zhiwu: Since we are not in investment, I just think that in the AI field—a deep-tech sector—investment and valuation should be reasonable. More importantly, focus on the team's technical expertise and their ability to monetize and implement. This is very important.