Per reporter: Ding Zhouyang Wen Menghua Per intern reporter: Li Jianning Zhu Peng Per editor: Dong Xingsheng
After several twists and turns, SenseTime finally did not have to leave the goal of sprinting to the market until 2022.
On December 20, SenseTime resumed its public offering on the Hong Kong Stock Exchange, raising a total of about HK$6 billion and is expected to be officially listed on December 30. Named after the "Shang Tang" that subverted the history of the Xia Dynasty and opened up the Shang Zhou, it is not difficult to see the ambitions of this company.
Will artificial intelligence be a revolutionary technology that rewrites the history of human civilization? Compared with 2014, when SenseTime was founded, everyone now enjoys the various conveniences brought by artificial intelligence. Sweeping the face can open the door and pay, the assembly line robot arm punching holes and detection are more accurate than people, and the wisdom of security makes the crime nowhere to hide... But embarrassingly, compared with seven years ago, the enthusiasm of the capital market for artificial intelligence has cooled significantly.
Analyzing the financial situation of SenseTime, one of the "Artificial Intelligence (AI) Four Little Dragons", it is not difficult to understand the reasons for the change of face of capital. On the one hand, the annual expansion of business and growth in revenue, on the other hand, the high labor costs and huge losses for several consecutive years, the profitability of AI companies that were once highly sought after in the primary market is far away. The patience of the financiers involved through several rounds of financing is running out, and sprinting to the market is the only way out of the current predicament. How do AI companies that "graduate" from the primary market and burn money perform after coming to the secondary market?
"The AI industry is in a high-speed development of the boom track, is a state that can be observed and tracked. However, when it comes to specific stocks and targets, it needs to be treated calmly. Qin Heping, a senior internet expert at Tianfeng Securities Media, said in an interview with the Daily Economic News reporter.
Three-and-a-half-year loss of more than 24 billion
Unicorns are also "gold-devouring beasts"
On December 13, AI unicorn SenseTime, which is only one foot away from the Hong Kong IPO, announced that the global offering and listing will be postponed, and all application funds will be refunded to the applicant without interest. After being included in a U.S. Treasury Department's investment restriction list on December 10, SenseTime suspended the IPO process for a time.
Image source: SenseTime's official website
On December 20, SenseTime resumed its public offering. SenseTime will also become the first enterprise in the "FOUR AI Tigers" (SenseTime, Megvii, Yuncong, ANDU) to officially land on the capital market.
In 2014, Xu Li, co-founder of SenseTime, who was still engaged in computer vision research, met Tang Xiaoou, a professor at the University of Chinese in Hong Kong, who had been studying deep learning since 2001. A detailed chat involved two professors who "hid" in ivory towers and came up with the idea of turning technology into business. This year, the founding team released the self-developed DeepID series of face recognition algorithms, with an accuracy rate of 98.52%, exceeding the accuracy of human eye recognition for the first time in the world. This achievement shocked the world, and it also means that face recognition technology has opened the era of commercialization. Subsequently, IDG Capital came to invest tens of millions of dollars, and SenseTime was established.
Among the artificial intelligence companies that have mushroomed since then and the companies that rub the heat of artificial intelligence, SenseTime has always been considered a "hard technology" company in the field of artificial intelligence. SenseTime believes in the "longboard theory" and spends a lot of money recruiting phDs and professors, buying hardware, and building supercomputing platforms.
When AlphaGo, an AI robot that works on the principle of "deep learning", has defeated world Go champions Li Shishi and Ke Jie, the imagination of the capital market for AI has also reached a boiling point. In July 2017, SenseTime completed a $410 million Series B financing round, setting a record for a single round of financing for artificial intelligence in the world at that time.
However, the technical barriers exchanged for long-term research and development investment are still difficult to hide the large losses in the short term.
In this year's prospectus, SenseTime disclosed its revenue data for the first time in recent years, and from 2018 to the first half of 2021, the company's net loss totaled more than 24 billion yuan. The core reason for the loss is the huge manpower investment. "Ai-powered companies are the most labor-intensive smart companies." This joke is confirmed one by one in the prospectuses of the "AI Four Little Dragons", that when financing four companies that were more powerful than one, the financial data was more ugly than one, and SenseTime had the largest loss.
SenseTime said in the prospectus that the main reason for the loss was the loss of research and development expenditure and the fair value of preferred stock. The prospectus also expects a significant increase in net losses in 2021 for the same reasons.
SenseTime's current state is a reflection of the AI industry. WIND data shows that as of the close of trading on December 20, a total of 21 of the 35 artificial intelligence concept stock companies have risen year-to-date, but the profit situation is generally not ideal. In the first three quarters of 2021, the average attributable net profit of artificial intelligence concept stocks was 655 million, and only 8 of the 35 companies reached an average, of which Hikvision's net profit attributable to the mother was 10.966 billion yuan, far exceeding the second place Kunlun Wanwei's 2.246 billion yuan. In addition, nearly 70% of the companies' net profit attributable to the mother is less than 500 million yuan, of which 5 companies are in a state of loss, and the net profit attributable to the mother of 13 companies has declined year-on-year.
When talking about the current situation of artificial intelligence "costing manpower and money", Liu Binbin, a researcher at the Shenlan Academy of Sciences, told every reporter: "We say that new energy vehicles reduce emission pollution, more refers to terminal vehicles, if you count the entire industrial chain is a complex formula." Similarly, the application of artificial intelligence is also a complex chain, in each node of the chain needs to continue to innovate and achieve a common vision, the software engineering of algorithms will reduce labor costs, the breakthrough of computing frameworks will reduce computing costs, resource conservation and the use of renewable energy are the efforts of the entire human being for the living environment. More importantly, we need to look at the return on investment. For human society, we can't give up electricity, computers, cars, etc. back to the agricultural era, although these industries will consume huge resources, when we think that the significance of the return is much greater than the input, the rest of the thing is to optimize. ”
The bubble has burst several rounds
AI is the empowerment of the industry, not the industry itself
In the summer of 1956, a lecture was held with the theme "How to simulate human intelligence with machines?" "The Academic Conference was held at Dartmouth University in the United States. Years later, the conference was recognized as the starting point for global AI research.
In 1997, when IBM's supercomputer Deep Blue defeated chess world champion Kasparov, people didn't know what that had to do with them.
In the past decade, when the dividend of the mobile Internet began to fade, artificial intelligence, as a medium for the integration of science and technology and the real world, brought opportunities for scene change.
"At present, artificial intelligence is in the application stage, gradually finding a landing point, and beginning to invest in the forefront of the industry, which can produce economic benefits, but it is still very limited." When talking about the development status of artificial intelligence, Zhu Xiaoyan, a professor in the Department of Computer Science of Tsinghua University who has been engaged in research in this field for more than 20 years, cautiously told every reporter.
Some scholars, even at once, do not say the scope of artificial intelligence. "Robot learning, automation, neuronal networks, it seems that everything can be counted as artificial intelligence." An industry conference, overwhelmed with so many AI experts. Some of our teachers joked that after decades of work, they suddenly found that I was originally doing artificial intelligence. A researcher who did not want to be named lamented to every reporter.
In Zhu Xiaoyan's perception, artificial intelligence has been pushed into the wind by capital and fell down heavily, which has been several ups and downs. "In March 2016, AlphaGo defeated two major Go champions, making more and more people pay attention to artificial intelligence. In those two years, all kinds of people poured in, and my office became a living room. Zhu Xiaoyan said, "I mainly give them Q&A, science popularization, artificial intelligence is not available in all fields." To figure out what problems your business needs to solve, is it really possible to use ARTIFICIAL intelligence technology? You can't just do it if the boss asks to do it. It's like falling in love, if you want to love me and be serious, it is not that parents say that you must fall in love, just fall in love. ”
"The AI is being held up every time and then falling down again. But this pullback will not be as tragic as the previous few times, because the general environment is much better, and everyone has indeed seen the landing of some technologies. Zhu Xiaoyan said, "Artificial intelligence is a technological boost, can empower the industry, the future it can also do a lot of things for human beings, but it is not an industry itself." The expectation of it must be related to its essence, so as not to be disappointed. ”
"In many cases, the public regards technology as a whole, in fact, the same technology application will show different levels of maturity in different scenarios." Therefore, AI must be the AI in the scene, and it is not comprehensive to talk about AI without the scene. Liu Binbin believes that the current artificial intelligence is indeed the "weak artificial intelligence" stage, more for the application of specific environments and scenarios, and the application of completely open scenes may take a long time to be perfect, which is also the focus of AI business landing.
What else can you do besides brush your face?
Stuck in more segmented scenarios
Whether it is the "AI Four Little Dragons" that are sprinting to the listing, or the old listed companies that have attached the new concept of artificial intelligence, the real mature application scenarios are still very limited.
Among them, the most widely used scenario is "AI + security". According to the data of Yiou Think Tank, the overall scale of China's security industry will exceed 800 billion yuan in 2020, of which the market size of "AI + security" is about 45.3 billion yuan, and the penetration rate of intelligent security is 5.5%.
As of the close of trading on December 20, the average market value of the 35 A-share artificial intelligence concept companies was 41.766 billion yuan, of which 3 companies had a market value of more than 100 billion yuan, and Hikvision (002415.SZ) became a well-deserved "leader" with a total market value of 479.580 billion.
Founded in 2001, Hikvision was initially positioned as a traditional video surveillance provider, and in 2011 it has taken the top spot in the global video surveillance market. Subsequently, Hikvision began to lay out in the fields of artificial intelligence, big data, cloud computing and other fields, combining AI technology with security. On December 13, hikvision's smart home camera company Fluorite Network submitted a prospectus on the Science and Technology Innovation Board to be accepted by the Shanghai Stock Exchange, officially opening the road to listing.
The "AI Four Little Dragons" are mainly engaged in computer vision, and they are mainly cut from the security scene, and their housekeeping skills "brush face technology" has been landed on a relatively large scale in the field of "big security" such as urban governance, park management, and traffic travel.
Image source: Photo.com
But scale does not equal maturity, Deng Zhidong, director of the Visual Intelligence Research Center of tsinghua university's Institute of Artificial Intelligence, told every reporter: "There are still semantic understanding challenges in the fields of vision, natural language processing, speech recognition and so on. To achieve human-level understanding, a breakthrough is needed for the next generation of artificial intelligence, namely cognitive intelligence and general artificial intelligence. Current AI is a data-driven weak artificial intelligence whose main mission is to make machines see, hear, and read more clearly. ”
In more complex industries such as autonomous driving and medical treatment, breaking through the limitations of weak artificial intelligence is an inevitable goal, but its path is not achieved overnight.
Taking automatic driving as an example, the current domestic "Automobile Driving Automation Classification" has been introduced, with a total of 6 levels, from the L0 level of all-artificial driving to the L5 level of fully automatic driving. Deng Zhidong said that L2 and L3 self-driving cars are mass-produced, and Volkswagen may be more concerned about when the large-scale commercialization of RoboTaxi (referring to self-driving taxis) such as L4 will arrive. When talking about fully automated L5 autonomous driving, Deng Zhidong said that the industry is not actually paying attention at present. "Before there is a real breakthrough in cognitive intelligence, L5-level unmanned driving should be far away." L5's commercial value is too great to be a truly disruptive technology, but it is inherently limited by the breakthrough development of cognitive intelligence at the human level. ”
The development of AI is human-led, but there will be an evolution mechanism that is different from human intelligence in the process of deep learning. "Data intelligence is a new species, which may not evolve with human intelligence mechanisms, but it can bring industrial value." Just like we can't ask large aircraft to adopt the same principle of flight as birds. Deng Zhidong said.
The depth and breadth of the current AI industry penetration are continuing to advance, but the pain points and difficulties of the industry are also obvious. Deng Zhidong believes that the key is the selection of subdivided application scenarios and the polishing of data. It is not conceivable that some kind of panacea of artificial intelligence will suddenly arrive. "Now the AI bubble is very strong, and everywhere they say that their products are 'smart'. Just like many service robots, there is not much breakthrough in perception and decision-making, and the user experience is poor, more like 'giant babies'. In fact, there is no AI, more of an enterprise PR. ”
The primary market tends to be saturated
Who pays for high valuations?
According to the Southwest Securities Research Report, since the beginning of 2019, the number of early investment and financing of AI enterprises in China has decreased significantly, while the proportion of investment in the middle and late period has increased significantly, and the proportion of C round investment and financing has exceeded 50%. In 2020, the average single financing amount of AI companies will increase from 100 million yuan to 330 million yuan.
Gao Dongxu, founder and chief analyst of China Entertainment Think Tank, told every reporter that with the continuous development of artificial intelligence, there have been few breakthrough new technologies, so the primary market tends to be saturated, but also prompted companies with financing needs to seek to absorb more funds in the secondary market.
After several rounds of financing in the primary market, the high valuation of artificial intelligence companies is a sword hanging over the heads of these AI unicorns.
"This year, on the whole, the artificial intelligence track and the entire technology industry are not so hot, and under the trend of strong domestic supervision, the overall valuation of China's technology industry sector represented by Internet companies is actually downward." The newly promulgated Data Security Law and the Personal Information Protection Law have put forward stricter regulatory requirements for businesses involving face collection, which has affected some artificial intelligence products to a certain extent. Qin Heping told every reporter, but in the long run, the standardization of the data environment is conducive to the benign development of the industry.
In Qin Heping's view, from the outlet to the gradual decline is the stage that artificial intelligence will inevitably go through in the capital market. "Any new technology comes out, there will be a stage in the capital market to chase up and kill the fall, and all the funds will pounce on it." However, this year, the capital market's attention to artificial intelligence has declined compared with previous years, and the valuation of artificial intelligence companies in the secondary market will tend to be rational. In the future, the combination of artificial intelligence and traditional industries will become deeper and deeper, and the capital market will continue to maintain high vigilance and high attention to artificial intelligence. ”
"Before a hundred flowers bloomed, you thought that writing a program could do artificial intelligence, but now the threshold has definitely become higher." Yang Ge, founding partner of Xinghan Capital, said, "From 2020 to 2025, which company can make standardized and versatile artificial intelligence chips, who is the leader in promoting the overall development of AI." This year I invested in two AI chip companies. ”
Yang Ge believes that in addition to focusing on the technology itself, investment in artificial intelligence should also pay attention to the concept of the market. "I came from a purely technical background, and then I transformed into an investment. I also have a lot of technical ideals and ambitions, but I think if investors focus entirely on technology and don't focus on concepts, there will be a big bias. Because technology alone cannot form a new market, the concept is a transitional bridge connecting technology and the market, which makes the social group have a figurative cognition of a relatively vague feeling, which in turn provides a very important soil for the development of technology. ”
"Digitalization, informatization, intelligence, these technologies will certainly be used by almost all industries, but whether they can enter the eyes of investors and whether they can make a big enterprise according to the logic of capital is another matter." Zhu Xiaoyan said.
Daily economic news