Post-80s genius scientists reap 25 billion IPOs

For two years, four times before and after the IPO, the fourth paradigm of AI unicorns was finally about to achieve positive results, and successfully "promoted" to the third share of Hong Kong stock AI, and investors who accompanied them all the way waited for the dawn of return.

Founded in 2015, Fourth Paradigm has extremely bright labels: the largest decision-making AI company in China, the only AI company jointly invested by the five major state-owned banks, and also collected star capital such as Sequoia, Tencent, and Lenovo. Decision-making AI can be simply understood as the use of AI technology to mine data laws and assist human decision-making, as a leader in this field, the market share of the fourth paradigm is even higher than that of technology giants such as BAT.

On September 28, Fourth Paradigm officially landed on the Hong Kong Stock Exchange with an issue price of HK$55.6 per share. It was up 5.22% by midday, with a total market value of HK$27.15 billion.

This is the first AI company to land on the Hong Kong stock market in the past year after SenseTime was listed on the Hong Kong stock market in late 2021 and Innovation Qizhi in early 2022.

However, compared with the two "predecessors", the listing process of the fourth paradigm is somewhat "awkward". Since 2020, the market generally believes that the AI industry has entered a deep water area, and AI companies that were once highly sought after in the primary market have also begun to face the dilemma of financing crunch, and have turned to IPOs.

Prior to this successful listing, Fourth Paradigm had submitted its listing application to the HKEX in August 2021, February 2022 and September 2022, all of which failed to pass the hearing.

Until entering 2023, the rise of large models has brought new hope to the industry. In March this year, Fourth Paradigm began to show its large-model product "Style Theory" to the public, and in April it submitted its latest IPO documents to the Hong Kong Stock Exchange.

In this listing, the issue price of HK$55.6 of Fourth Paradigm was at the low value of the offering price range (HK$55.6 to HK$61.16). Therefore, some people have ridiculed that the listing of Fourth Paradigm is a "version of the IPO", and the total market value of Fourth Paradigm is about HK$25.8 billion, with a total net raising amount of about HK$836 million. Among them, cornerstone investors, including New China Capital, Beijing Zhongguancun and Montage Holdings, subscribed about HK$758 million.

In any case, after two years of dormancy, the company finally successfully landed in the secondary market and stood on a new capital stage.

An AI unicorn created by genius engineers after the 80s

In September 2014, post-80s AI engineer Dai Wenyuan and his wife Wu Ming registered and established the fourth paradigm.

The following year, Sequoia China, led by investment tycoon Shen Nanpeng, spent $4 million and became the sole investor of the fourth paradigm angel round. At that time, the AI industry was on the eve of the outbreak, and the fourth paradigm product route that had just begun to falter was not clear.

It is said that when Shen Nanpeng and Dai Wenyuan, the founder of the fourth paradigm, discussed investment matters, Dai Wenyuan asked the other party: "What questions do you want to ask me?" Shen Nanpeng's answer was: "I don't really care much about what you do." In the early stages of entrepreneurship, I care more about your team than about the specific business. ”

▲(Shen Nanpeng)

Shen Nanpeng's words are not unfounded, in the AI industry-academia community at that time, the team of the fourth paradigm was luxurious:

The founder Dai Wenyuan is the world's top AI scholar, graduated from Shanghai Jiao Tong University, studied for a doctorate at the Hong Kong University of Science and Technology, and won the ACM Competition (a major programming competition organized by the American Computer Association) known as the "Computer Olympiad".

After leaving academia, Dai Wenyuan joined Baidu and Huawei as a senior scientist. From 2009 to 2013, during his four years at Baidu, Dai Wenyuan optimized Baidu's search promotion service management platform "Phoenix Nest", which increased Baidu's revenue several times. As a result, Dai Wenyuan became the youngest senior scientist in Baidu's history, and was called "Dai Shen" by colleagues. During his time at Huawei, Dai Wenyuan also made great achievements and helped Huawei win large bank orders.

In the process of leaving the big factory to start his own business, Dai Wenyuan's fourth paradigm also attracted a lot of talents: Yang Qiang, the company's co-founder, is the first academician of the International AI Association in the Chinese community, which can be called the world's top AI researcher, and he is also Dai Wenyuan's doctoral supervisor at the Hong Kong University of Science and Technology; Chen Yuqiang, who joined the company in March 2015 and is currently the chief research scientist and non-executive director, has served as a senior R&D talent at Baidu and ByteDance; Liu Yiming, a core algorithm architecture engineer, has won a prize in the well-known international programming competition (Google Code Jam Top 500) hosted by Google...

Such a bright team does not only attract the attention of Sequoia. After being recognized by Shen Nanpeng's insight, the fourth paradigm continues to gain favor.

According to the prospectus, from August 2015 to July 2021, Fourth Paradigm completed a total of 11 rounds of financing. Among them, not only the "old friend" Sequoia has been running for multiple rounds, but also joined the five major banks (Industrial and Commercial Bank of China, Bank of China, Construction Bank, Agricultural Bank of China, Bank of Communications), Cisco, Lenovo, Tencent, National Manufacturing Transformation Fund, Goldman Sachs, Innovation Factory and other star capitals.

In the process of snowballing financing, the valuation of the fourth paradigm has also risen. As of the completion of the Series C financing in March 2019, the valuation of Fourth Paradigm has exceeded US$1 billion, and it has become a unicorn enterprise.

According to the last round of financing before the listing completed in July 2021, the valuation of Fourth Paradigm reached US$2.95 billion (more than 20 billion yuan). Compared to the $27 million valuation after the completion of the angel round in 2015, the valuation of Fourth Paradigm has increased by about 108 times in less than 6 years.

In the process of financing snowballing, Dai Wenyuan's personal net worth has also risen.

According to the prospectus, Dai Wenyuan and his wife Wu Ming, as well as five shareholding platforms Beijing Xinzhi, Paradigm Investment, Paradigm Yinyuan, Paradigm Surprise and Paradigm Tianqin, are listed as a group of controlling shareholders of Fourth Paradigm.

Prior to listing, Dai Wenyuan directly controlled or through his wife and shareholding platform, controlling a total of 40.44% of the company's shares.

In the "2022 Hurun U40 Young Entrepreneur List", Dai Wenyuan and his wife entered the list for the first time with a wealth of 8.5 billion yuan, ranking 18th.

After the IPO, Dai Wenyuan's total control of shares decreased to 38.84%. Based on the total market value of HK$27.15 billion at midday on the first day of listing, Dai Wenyuan's latest net worth is about HK$10.545 billion.

A different way of decision-making AI leaders

An investor in the AI field told "City Boundary" that in the outbreak of the domestic AI industry in 2017 and 2018, the main flashpoint was concentrated in the field of computer vision (CV, Computer Vision).

Unlike the same group of friends who took CV as the main channel, the fourth paradigm has taken a different approach from the beginning and chose to delineate its business scope in the field of "decision-making AI".

The downstream applications of CV technology include face recognition, object recognition and other aspects. At that time, star companies such as the "AI Four Tigers (SenseTime, Megvii, Yuncong, and YITU)" emerged as leaders in this regard. According to the 2017 research report of Guojin Securities, more than 40% of AI companies in the domestic market at that time were concentrated in the CV field.

In contrast, the "decision-making AI" selected by the fourth paradigm refers to the algorithm through full mining and segmentation of massive data, deriving insights, predicting events and putting forward relevant decision-making suggestions, and then having an important impact on the decision-making ability of enterprises and users, providing value for enterprises and users. For example, the user recommendation algorithms used by platforms such as Douyin, Taobao, and Xiaohongshu fall into this category.

In Dai Wenyuan's vision, the fourth paradigm is to make a tool product that can lower the threshold for AI use, and the ultimate ideal is to achieve "AI For Everyone", so that everyone can become a participant, user and creator of artificial intelligence.

In its early days, Fourth Paradigm quickly opened up its popularity by quickly reaching key users.

According to the prospectus, in 2015, the second year of the company's establishment, Fourth Paradigm has begun to provide services for "the first benchmark user in the financial industry" and develop "smart product recommended applications"; In 2016, Fourth Paradigm began to provide services to other financial institutions such as securities companies; Since 2017, it has been working with one of China's largest state-owned banks to build a risk control AI model based on data such as the bank's user history transaction information...

At that time, although decision-making AI was not a mainstream choice in the industry, the fourth paradigm was not completely unrivaled in this field. More and more competitors such as Internet giants, cross-border CV companies, and emerging AI players soon began to target this market cake. The "stronghold" financial market of the fourth paradigm is undoubtedly their place of contention.

For example, in 2016, Tencent Cloud and Baidu Cloud successively announced the opening of financial cloud solution services; In 2017, Wang Jin, former chief risk officer of Baidu Finance, left to start a business and founded the financial AI solution company "Ronghui Jinke"; In 2018, Innovation Qizhi, incubated by Kai-Fu Lee Innovation Factory, was also established; Also in mid-2018, Cloudwalk Technology began to vigorously develop loan solutions for small and medium-sized enterprises...

Faced with a strong and numerous number of opponents, Fourth Paradigm tried to expand its revenue sources and turned its attention to other markets such as energy and electricity.

According to the prospectus, in 2018, Fourth Paradigm began serving customers in the retail and energy sectors. In order to quickly expand customer groups and expand revenue, Fourth Paradigm has also embarked on the road of external acquisition.

In 2021, Fourth Paradigm spent 198 million yuan to acquire a 66% stake in Guangzhou Jianxin. Guangzhou Jianxin's main customers are concentrated in the energy and power industry, mainly engaged in the sales of application software and hardware products, application development and consulting services to customers. In the same year, Fourth Paradigm also acquired a total of 56.84% equity interest in Li Technology, a software developer that provides one-stop operation and maintenance automation solutions for customers in the financial industry, for a total consideration of 270 million yuan through acquisition and capital increase.

In June 2022, Fourth Paradigm extended its tentacles to the manufacturing field and spent RMB 119 million to acquire a 79.66% stake in the manufacturing informatization and industrial automation solution company AP Gonghua and its subsidiaries.

After some acquisition operations, the revenue structure of the fourth paradigm has shown diversified characteristics. In 2022, the energy and power industry has become the largest source of revenue for the fourth paradigm, accounting for 20.3%, and the revenue from the financial industry accounting for 16.9%, ranking second.

At the same time, from 2018 to 2022, the number of customers of Fourth Paradigm continued to rise, with 38, 79, 156, 245 and 409 respectively.

It is difficult to solve the commercialization dilemma

According to the China Insights Consulting report, in terms of revenue in 2022, the top five in the domestic platform-centric decision-making artificial intelligence market accounted for a total of 56.1% of the market share, of which the fourth paradigm ranked first with a share of 22.6%, far exceeding the second place with a share of 12.3%.

The above report does not list the specific names of the second, third, fourth and fifth market participants, but according to the relevant descriptions, it can be inferred that Company A is most likely Baidu, followed by Alibaba, Huawei, and Tencent.

Although it is firmly in the leading position of domestic decision-making AI and continues to improve its revenue structure, objectively speaking, the financial situation of the fourth paradigm is still difficult to be optimistic.

Overall, the revenue of the fourth paradigm in 2020, 2021 and 2022 has increased year by year, with 942 million yuan, 2.018 billion yuan and 3.083 billion yuan respectively. However, from the perspective of profit, the company is still in the red, with losses of 750 million yuan, 1.802 billion yuan and 1.653 billion yuan in 2020, 2021 and 2022, respectively, with a total loss of more than 4.2 billion yuan in three years.

The fourth paradigm summarizes the reasons for losses as reasons such as higher R&D expenditure. According to the prospectus, the company's R&D expenses in 2020, 2021 and 2022 were 566 million yuan, 1.249 billion yuan and 1.65 billion yuan respectively; The R&D expense ratio reached 60%, 61.9% and 53.5% respectively.

It should be pointed out that under the vigorous research and development and multiple external acquisitions, the current capital reserves of the fourth paradigm are quite sufficient. As of the end of 2022, the Company had cash and cash equivalents on its books of $1,327 million.

Still, persistent losses are undoubtedly draining investors' patience, which explains the urgency behind Fourth Paradigm's 4-year hit IPO.

According to the statistics of the Securities Times, as of July 2023, among the companies to be listed on the Hong Kong Stock Exchange, there are 28 companies involved in AI or digitalization, of which 19 have a loss in net profit as of 2022, and the remaining 7 are only slightly profitable.

Lin Wei, a researcher at Shenzhen Innovation Investment Group Co., Ltd., told the city: "VC/PE's tolerance for unprofitable companies, including the AI industry, is decreasing, and the requirements for earnings conversion plans and valuation control are increasing." ”

To continue to gain favor with investment, the fourth paradigm needs to tell a new story. The wave of big models emerging at the end of 2022 is undoubtedly a great opportunity.

In March 2023, Fourth Paradigm launched its self-developed large-model product "SageGPT", positioning itself as an enterprise-level generative artificial intelligence product designed for business scenarios. So far, the large model product has been updated to version 3.0.

▲ (Fourth paradigm large model product display, photographed by the city at the 2023 Shanghai World Artificial Intelligence Conference)

However, in the large model market, the challenges facing the fourth paradigm are also not to be underestimated.

Although many large model products, including Formula Theory, are in the early stages of commercialization. However, this market is already extremely crowded and competitive, and various companies have entered the market, including SenseTime, Chuangxinqizhi, Internet giants such as Baidu and Alibaba, and AI start-ups such as Lanzhou Technology and Mobvoi.

And almost all large model market participants are locked in the enterprise market as the target market as the fourth normal form. A staff member of a large model company told "City Boundary": "Now is a critical period to develop the market, as far as I know, in order to win customers, almost every company will provide customers with a period of free use." ”

On the other hand, the development and iteration of large model products also require a very high level of R&D investment. According to Guosheng Securities, the cost of training GPT-3 for a large model with 175 billion parameters is as high as about 1.4 million US dollars, and for some models with larger parameters, the training cost may be between 2 million US dollars and 12 million US dollars.

An employee of Fourth Paradigm once told the "city boundary": "In terms of cost, of course, the one-time investment of large-model services will seem to be large, but if it is evenly distributed to everyone and each unit of working time, it is actually cost-effective." "In other words, it's a track with a very long payback period.

For Fourth Paradigm at this stage, the listing has added valuable ammunition and provided an exit channel for eager investors, but it is only the beginning.

With the gradual completion of relevant policies, the competition for the market by large model enterprises has entered a critical period, and the problems related to commercialization will be placed in front of it again.

Bibliography:

1. "38-year-old genius scientist at the helm, three and a half years of accumulated losses of 3 billion, the fourth paradigm sprint to Hong Kong stock IPO", titanium media

2. "The "Fourth Paradigm" Helps Traditional Industries Embrace Artificial Intelligence - "Artificial Intelligence for All"", Economic Daily Chen Jing

Author | Dong Wenshu

Edit | Sun Chunfang

Operations | Liu Shan