Behind the plunge in SenseTime’s stock price: Deep learning technology has reached its ceiling in terms of application

The sharp drop in SenseTime’s stock price reflects market sentiment and also bursts the market bubble in the valuation of the AI ​​(artificial intelligence) industry.

On June 30, SenseTime, Asia's "first AI stock", suffered a black Thursday. Affected by the lifting of the lock-up period for the company's pre-IPO investors and cornerstone investors, SenseTime's share price fell 46.77% in a single day. As of the close of the day, the share price closed at HK$3.13.

Although SenseTime anticipated the market sentiment in advance, Chairman and CEO Xu Li, Chief Scientist Wang Xiaogang, Executive Director Xu Bing and other co-founders announced an additional lock-up period for their shares at the opening, but these measures still failed to reverse the downward trend of SenseTime's stock price.

The lifting of shareholder lock-up restrictions can always cause fluctuations in the stock prices of listed companies, but a sharp drop like SenseTime is not common. Why is the capital market so bearish on SenseTime, the leader of the domestic AI industry?

Before competing in the capital market, SenseTime had always been a hot commodity that investors were eager to pursue. Public information shows that it had completed 12 rounds of financing before its listing, raising a total of $5.2 billion in funds.

However, there is no free lunch in the world. Just like we pay interest on loans, the money raised by SenseTime also comes at a price. Investors value the capital premium after SenseTime goes public. Therefore, while SenseTime continues to raise funds, it also ties its fate closely to the investors' demands for profit and cash out.

In December 2021, SenseTime officially landed on the Hong Kong Stock Exchange. Although its issue price was not high, the company's stock price soon doubled due to its great reputation, and SenseTime's total market value also reached 300 billion.

Looking at SenseTime’s outstanding performance in the capital market, the only hurdle that investors need to overcome before they can obtain the expected capital premium is the lock-up period.

The number of shares in SenseTime's initial public offering was 1.725 billion. However, there were not many tradable shares that retail investors could subscribe to at that time, which also laid a heavy foreshadowing for the sharp drop in SenseTime's stock price after the lifting of the ban on pre-IPO investors and cornerstone investors.

For investors who bought SenseTime before the listing, the lifting of the ban means that their investment in the primary market has been successful. But for investors who bought SenseTime after the listing, the lifting of the ban increases the possibility of them becoming leeks. Based on their respective demands, the trigger for the sharp drop in SenseTime's stock price is the concentrated opposition reaction of the above two different positions.

It is worth noting that with the sharp drop in SenseTime’s stock price, investors have once again begun to discuss whether there is a bubble in AI valuation. Some investors bluntly stated that the sharp drop in SenseTime’s stock price means that the AI ​​valuation bubble has burst, and its stock price may continue to fall in the future.

As Asia's largest artificial intelligence company, SenseTime has not yet achieved profitability since its establishment. In the past four years, SenseTime's annual losses from continuing operations have been 3.433 billion yuan, 4.968 billion yuan, 12.158 billion yuan and 17.177 billion yuan, respectively, showing a trend of increasing year by year.

Although SenseTime is already a listed company with a market value of hundreds of billions, in terms of business maturity, it has not yet been able to prove its strength to the market and investors with solid products and services.

Currently, SenseTime is still facing the problems of immaturity of new businesses and huge amounts of accounts receivable.

SenseTime's main businesses include smart business, smart city, smart life and smart car. Among them, smart city and smart business contribute the vast majority of revenue. In 2021, SenseTime's smart business and smart city revenues were 1.958 billion yuan and 2.143 billion yuan respectively, and even if the revenues of smart life and smart cars are added together, they are not as much as the revenues of any of the first two businesses.

If we were to describe SenseTime’s business structure, the term “top-heavy” seems to be a good description.

Why did SenseTime form such a business structure? To a large extent, this is closely related to the highly limited application scenarios of AI technology. If AI technology is to empower real-world scenarios, it must first build an algorithm model that is adapted to it. However, a carefully built model cannot cure all diseases, and any slight changes in the real-world scenario will affect the efficiency of the model.

At present, SenseTime’s AI technology is most mature in the fields of smart business and smart cities.

Taking smart cities as an example, as of the end of last year, SenseTime's smart city platform City Ark had been equipped with 22,425 AI models. In contrast, although SenseTime has been planning its smart life and smart car businesses for many years, the related businesses are still in the critical stage and it will take a long time to become a new growth pole.

SenseTime’s future may be good, but due to the need to continuously invest in the research and development of new businesses, continued losses will also be a high probability event.

In addition, judging from SenseTime's financial report, there is also a problem of paper gold on its books. In 2021, although SenseTime's operating income was 4.7 billion yuan, its net cash used in operating activities during the same period was -2.485 billion yuan. This reflects a problem that although SenseTime has made a lot of money, the money has not entered its pocket.

Where did the money go? According to the financial statements it submitted, the money was spent by SenseTime on the one hand, and became accounts receivable on the other hand.

In 2021, SenseTime's R&D expenditure increased by 47.3% year-on-year to 3.614 billion yuan. However, it is criticized that the increased R&D expenditure was not used for technical research, but for expanding the team and improving treatment. As of the end of last year, SenseTime had a total of 6,113 employees, of which 70% were R&D personnel. Affected by the above factors, SenseTime's total expenditure last year also increased.

However, compared to spending so much money, what is more worrying for Shang Tang is that the money earned cannot go into his own pocket.

SenseTime's customer base is mainly G-end and B-end. Based on the good credit status of the G-end, the company often records its revenue as accounts receivable and gives a certain payment period. Because of this operation, SenseTime's books often show an increase in revenue and accounts receivable, but the money has not been received.

Last year, due to the repeated impact of the epidemic, SenseTime's B-end customers also began to hang up their accounts. As a result, the company's third-party accounts receivable also surged from 3.562 billion yuan last year to 6.018 billion yuan, a year-on-year increase of 68.95%. These paper wealth also affected SenseTime's own hematopoietic ability.

In fact, whether it is the top-heavy business structure or the erosion of the company's hematopoietic ability, it will increase investors' suspicion of SenseTime. The prospect of AI track is good, but vision is vision and business is business. SenseTime has not yet proven itself at the business level at least.

Similarly, the problems encountered by SenseTime also plague other domestic AI companies. On the one hand, these companies have to prove their value to investors, and on the other hand, they have to continue to burn money to study technology. However, they have also ignored the limitations of AI development and the dual problems of a single commercial model.

In 2019, Zhang Bei, an artificial intelligence expert and academician of the Chinese Academy of Sciences, said meaningfully that technological improvements would be difficult to completely solve the fundamental defects of artificial intelligence at the current stage. Except for the "miracles" in the three fields of speech recognition, image recognition and Go, this "miracle" has not been reproduced in other fields, and the boundaries and conditions of its technological application have gradually become clear.

In fact, even from a global perspective, AI companies face the challenge of a narrow track. AI technology is advanced, but it is also full of low adaptability like a "dragon-slaying sword". As Academician Zhang said, the activeness of artificial intelligence in the second decade of this century benefited from breakthroughs in the field of deep learning, and deep learning technology, from an application perspective, has reached its ceiling.

Before it could shine, it had already hit the ceiling of the industry. For SenseTime, which had just gone public, perhaps it had already known the answer to this story, and the large-scale withdrawal of the capital market was like manually giving a thumbs up to this answer.

Now, the only people who are still immersed in this story are probably the stock investors who suffered nearly 50% plunge in one day.

As a winner of Toutiao's Qingyun Plan and Baijiahao's Bai+ Plan, the 2019 Baidu Digital Author of the Year, the Baijiahao's Most Popular Author in the Technology Field, the 2019 Sogou Technology and Culture Author, and the 2021 Baijiahao Quarterly Influential Creator, he has won many awards, including the 2013 Sohu Best Industry Media Person, the 2015 China New Media Entrepreneurship Competition Beijing Third Place, the 2015 Guangmang Experience Award, the 2015 China New Media Entrepreneurship Competition Finals Third Place, and the 2018 Baidu Dynamic Annual Powerful Celebrity.