EqualOcean interviewed with Xia Zhijin (夏志进), general partner at Vertex Ventures China (祥峰投资) on July 15, 2019.
Temasek Holdings subsidiary Vertex Ventures China invests in high-growth technology, retail, and internet start-ups across mainland China.
Mr. Xia Zhijin joined the firm in 2010. Before joining the team, he was a research scientist at Thomson Innovation Laboratory in Beijing and an analyst in Taishan Capital. Mr. Xia holds a M.E. and a B.E. in Electrical Engineering from Tsinghua University.
Here are the key takeaways from the interview:
2019: AI commercialization is the topic
Xia Zhijin: In retrospect, AI development peaked in 2016-2017, where VC activities were vigorous. Actually, the investment activities have started since 2013 and 2014, when audio and image recognition products such as those of iFLYTEK (科大讯飞) and Xuebajun (学霸君) came to people's attention.
At the earlier development stage in 2015 and 2016, people were not certain about the direction of AI’s application, and therefore most of the venture investments occurred at the time were basic technology-oriented. While in recent two years certain AI technologies are mature enough and more open-source algorithms are available, the focus turns to the application in every possible vertical market.
We noticed two general trends in 2019. From an investment perspective, fundraising activities are clearly slowing down, especially for RMB funds. Venture activities are continuing, but they tend to make safer investment decisions by investing in larger, more mature companies. From a startup perspective, certain technologies are ready to be put into practice and they are looking for approaches in landing AI. They tend to think more about how to make money, to develop a vertical market, and to find user demand and pain points, etc.
Two directions for AI investment: underlying technology and applications
Opportunities for chipmakers are at the edge
Xia Zhijin: Chips provide a foundation for AI development, at the bottom of the AI value chain. This is a very competitive area, with incumbents such as Huawei (华为) and Horizon Robotics (地平线) and also many small to medium-sized companies.
Tech giants such as Huawei and Baidu have already occupied the cloud server chip market. These cloud IaaS providers either make chips themselves or closely associated with chip-making partners. This is a saturated market and newcomers could hardly compete with them.
However, opportunities for medium to small-sized players remain at the edge, providing chips used in terminal devices. Horizon Robotics is such a company who enables tier 1 suppliers and OEMs in autonomous driving sector.
Many newcomers target the embedded processors of terminal devices because the area requires high customization. Demand varies regarding application scenarios. Based on the different scenarios, there are segmented needs regarding computing power, electric power, and cost. For example, chips for terminal devices used in surveillance cameras are different from those used in smart locks. For surveillance cameras, chips could also be designed differently due to different levels of data to be processed in different environment.
Making these investment decisions, we consider the technical background of the team and the product positioning strategies.
AI application can hardly stand along
Xia Zhijin: No company is pure AI company. Currently, only cloud platforms or business intelligence companies could provide standardized, generic use of AI, while most of the firms are working on consolidation with industries and providing customized solutions.
Our investment logic is to replace human labor with cost-saving technology. Therefore, we choose the areas that are the most labor-intensive, where people do repetitive work. Security, customer services, and warehouse logistics are such labor-intensive areas.
For example, there are 5 million people conducting customer services across the country. While corporate software such as an HR system could save only a few headcounts, smart customer services could save hundreds of them. Besides, In the past ten years, the wage growth rate of blue-collar workers is much more than that of white-collar. Our 2019 investment XiaoduoAI (晓多机器人) could reduce labor cost by 20% to 50%.
Our layout in security area is a "one step behind" strategy. It is a long process in earning government trust and building relationships, so we focus more on the underlying technology. We invested in Tuya Technology (图鸭科技), a company focusing on deep learning-enabled image
AI application in healthcare is definitely preceding slower due to three reasons. First, personal medical data in China are highly separated in different hospitals on multiple platforms. Second, development of algorithms requires repeated iterative learning and training, while the process in healthcare is quite slow due to the complexity of policy issue and authorization procedures. Third, a doctor takes multipart considerations in determining symptom, while current AI could only take care of it from a single focused point.
It's impossible to have AI replace teachers completely. Although many education-related products are employing AI technologies such as Xuebajun using image recognition to correct and revise homework. But they are essentially education companies, rather than AI companies.
Unmanned retail store is far from developed, though some AI applications are helping retailers in the value chain to improve efficiency. Smart fitting room is a good illustration. Underware brand Neiwai (内外) adopted intelligent fitting rooms which improve the conversion rate of purchase from 15% to 60%.
AI commercialization situations by industries
Xia Zhijin: Giant tech firms are the primary beneficiaries of using AI. These companies have already had structured data collection and technology background. AI as an important part of the technology provides a lucrative return to tech giants in consumer internet. For example, Baidu, one of the earliest AI players, applied AI in the advertising business to generate revenues and intelligent recommendation to attract customers.
So far, many companies in security area could fall into revenue bracket of CNY 1 - 2 billion. It's reasonable for some BI companies and fintech companies to achieve CNY 100 - 400 million. Providers for smart customer services would achieve CNY 60 - 80 million. Revenues that autonomous driving companies generate mostly come from ADAS products and co-research fundings.