AI Unicorn Backed by Sequoia for 6 Rounds is Going Public

[baoshi.rao]

China's largest decision-making AI company has officially landed on the Hong Kong stock market.

An AI tech expert and investor power couple built an IPO in 9 years.

After nearly two years, the Hong Kong stock market is about to welcome another AI unicorn.

On September 28, 4Paradigm, China's largest decision-making AI company, officially listed on the Hong Kong Stock Exchange. It is reported that 4Paradigm priced its IPO at the lower end of the guidance range at HK$55.60 per share, issuing 18.396 million shares with a market capitalization of HK$25.8 billion. The net proceeds amounted to HK$835.5 million, with cornerstone investors subscribing to 13.6396 million shares totaling HK$757 million.

On its first trading day, 4Paradigm opened at HK$63.10 per share, up 13.49% from the issue price, with a market cap exceeding HK$29 billion.

Unlike the naming conventions of other AI companies, "4Paradigm" originates from a speech about the four paradigms of scientific development by Turing Award winner and relational database pioneer Jim Gray before his disappearance in 2007.

Founded in 2014, 4Paradigm was established by the power couple of tech expert Dai Wenyuan and investor Wu Ming. By application domain, 4Paradigm positions itself as an AI decision-making platform company, which is relatively more technically challenging. Over its 9-year journey, the company has attracted backing from prominent institutions like Sequoia Capital and Sinovation Ventures, reaching a pre-IPO valuation of nearly $3 billion.

Behind 4Paradigm's four IPO filings lies the impending expiration of shareholder withdrawal clauses. Riding the new wave brought by AI large models, 4Paradigm seized the opportunity to achieve its IPO. However, what new directions will its future commercialization take? After all, like its AI peers, 4Paradigm has yet to reach a clearer profitability inflection point.

01
AI Expert Founds Unicorn Startup Valued Over $25.8 Billion

Born in 1983, Dai Wenyuan just turned 40 this year. He started programming in Logo language on computer lab machines at a very young age. His childhood dream was to join the national team and represent China in the International Olympiad in Informatics. He was so determined to study computer science that he even gave up a guaranteed admission to Tsinghua University due to major preferences. Fortunately, Shanghai Jiao Tong University's ACM class, which rivals Tsinghua's Yao class, extended an olive branch to him.

In 2005, Dai won the world championship at the ACM International Collegiate Programming Contest global finals, often called the "Olympics of Computing." Interestingly, when later asked why he entered the AI field, Dai candidly admitted it wasn't due to foresight. When his ACM class had to choose research directions, he was abroad for competitions and couldn't participate in the selection. By the time he returned, all the popular fields were taken, leaving only the then-niche area of artificial intelligence.

After graduating from Shanghai Jiao Tong University, Dai pursued further studies at Hong Kong University of Science and Technology under Professor Qiang Yang, the first Chinese scholar elected as an AAAI Fellow. During his graduate studies, Dai published multiple papers at top AI conferences thanks to his strong foundation in informatics algorithms. His papers on transfer learning were among the world's top three most cited in the field.

After leaving academia, Dai joined Baidu as head of the "Baidu Phoenix Nest" system, becoming the company's youngest scientist at the time. He led the team that built China's largest machine learning system and received Baidu's highest award. It was during this period that he met Wu Ming, a senior investment manager at Baidu who later became his wife. Dai then joined Huawei's Noah's Ark Lab as Principal Scientist, where he received the "President Award."

In 2014, as a wave of AI entrepreneurship emerged, Dai chose to resign and start his own business. In September of that year, 4Paradigm was officially established, and in December, it released the first commercial product using its core technology - version 1.0 of the AutoML framework-based Prophet Platform.

Of course, 4Paradigm's launch wasn't a solo effort by Dai. Just two months after founding the company, he brought in his graduate advisor Qiang Yang as Chief Scientific Advisor. He also recruited Chen Yuqiang, his junior from Shanghai Jiao Tong University and former Baidu colleague, as co-founder and Chief Scientist.

According to Jim Gray's explanation of the four paradigms of human scientific development, the fourth paradigm involves machines replacing humans in discovering patterns within massive datasets. Since its inception, The Fourth Paradigm has been committed to empowering enterprises with AI transformation and advancing AI development. Focusing on decision-making AI applications, it has chosen the path of "platform-centric artificial intelligence" to help businesses achieve rapid and scalable AI transformation.

The prospectus reveals that after listing, Dai Wenyuan and his wife will directly and indirectly control approximately 38.84% of The Fourth Paradigm's issued share capital. With a market capitalization of HKD 25.8 billion, the couple's net worth will exceed HKD 10 billion post-listing.

02
Sequoia Invested in Six Rounds, Achieving a 3x Return in 8 Years

At a time when the commercial application of AI was unclear, The Fourth Paradigm managed to attract capital market attention due to Dai Wenyuan's reputation. Initially co-owned by Dai Wenyuan and his wife Wu Ming, the company secured a USD 4 million Series A funding from Sequoia Capital in August 2015, just a year after its establishment. It continued to receive subsequent rounds of financing, gradually building an impressive roster of shareholders.

By the time of its IPO, The Fourth Paradigm had completed 11 funding rounds, raising a total of USD 983 million and RMB 436 million, exceeding USD 1 billion in total financing. The Series D round alone amounted to USD 700 million. Investors included prominent institutions such as Sequoia, Innovation Works, Songhe Capital, Poly Capital, Yuexiu Industrial Fund, Cornerstone Capital, Lenovo Capital, Cisco, CITIC Industrial Fund, Primavera Capital, and Goldman Sachs, as well as state-backed funds like China Development Bank, China Reform Holdings, and the National Manufacturing Transformation Fund, alongside tech giants like Tencent.

With each round of financing, The Fourth Paradigm's valuation soared. Starting at USD 26.67 million, it grew to USD 2.952 billion by its last funding round in June 2021. The company crossed the unicorn threshold during its Series C round in 2018, when its valuation surpassed USD 1 billion.

It is noteworthy that Sequoia Capital was the largest institutional shareholder of 4Paradigm before its IPO, holding an 11.7% stake. Over the span of five years from August 2015 to November 2020, Sequoia invested in six funding rounds.

According to the prospectus, Sequoia invested $4 million in Series A, $14 million in Series A1, RMB 19.1889 million in Series A2, $6.169 million in Series B1, $30 million in Series C, and $60 million in Series D, totaling approximately $111 million (about HK$868 million). Based on 4Paradigm's IPO valuation of HK$25.8 billion, Sequoia's stake was worth around HK$2.8 billion. This means that over its eight-year investment, Sequoia achieved a return exceeding 3x.

Ji Yue, a partner at Sequoia China, once remarked on 4Paradigm years ago: 'Sequoia's early investment in 4Paradigm was heavily influenced by the core capabilities of the founders and the team. From the outset, we had very high expectations.' Notably, Dai Wenyuan, 4Paradigm's founder, and Neil Shen, founder of Sequoia China, are both alumni of Shanghai Jiao Tong University. Dai's wife, Wu Ming, a co-founder and investor in 4Paradigm, joined Sequoia in 2018.

Additionally, 4Paradigm initially focused on AI applications in the financial sector. Its investor lineup includes major financial institutions such as ICBC, Bank of China, China Construction Bank, Agricultural Bank of China, and Bank of Communications.

03

'Forced IPO': How Will 4Paradigm Address Its Challenges?

After 11 rounds of financing, 4Paradigm's shareholder base has expanded significantly, but concerns have also emerged.

It is reported that 4Paradigm had agreements with investors stipulating that if the company failed to complete an IPO within 48 months after the completion of Series D financing or within 18 months of signing the termination agreement, withdrew its IPO application, failed to refile within three months after the application lapsed, or did not receive IPO approval, investors could exercise their right to withdraw their investments.

The final deadline for 4Paradigm's IPO was the end of 2023. Consequently, the company submitted IPO applications to the Hong Kong Stock Exchange four times starting from August 2021.

However, going public does not directly solve all of 4Paradigm's problems. According to its prospectus, the company's revenues for 2020, 2021, 2022, and Q1 2023 were ¥942 million, ¥2.018 billion, ¥3.083 billion, and ¥644 million, respectively, showing a declining growth trend. Its net losses for the same periods were ¥750 million, ¥1.802 billion, ¥1.653 billion, and ¥304 million, respectively.

4Paradigm's main expenditure lies in its high R&D costs. From 2020 to Q1 2023, its R&D expenses were ¥566 million, ¥1.25 billion, ¥1.65 billion, and ¥241 million, accounting for 60.0%, 61.9%, and 53.5% of its revenue during those periods, respectively.

According to a CIC report, 4Paradigm ranked first in China's platform-centric decision-making AI market by revenue in 2022, holding a 22.6% market share. The demand for platform-centric solutions primarily comes from top-tier large enterprises or government entities across various industries, which have extensive application scenarios and highly customized AI adoption needs. Currently, 4Paradigm serves leading companies in sectors including finance, retail, manufacturing, energy and power, telecommunications, transportation, technology, education, media, and healthcare.

Losses are a common challenge for AI companies. For AI startups before the large model wave, institutions often evaluate them based on their ability to implement AI in practical scenarios. As a result, 4Paradigm is continuously expanding its boundaries.

This expansion is evident in its development trajectory. In 2021, to strengthen its AI applications in the energy and power sector and IT operations management, 4Paradigm spent hundreds of millions of yuan to acquire two companies with over a decade of experience: Jianxin Seller and Ideal Technology. These acquisitions also led to changes in 4Paradigm's revenue structure.

From 2020 to 2022, revenue from the financial sector grew from ¥292 million to ¥520 million, but its share of total revenue dropped from 31% to 16.9%, making it the second-largest sector. Meanwhile, revenue from the energy and power sector increased from ¥138 million to ¥627 million, with its share rising from 14.7% to 20.3%, gradually becoming the largest business segment.

To broaden its application scenarios, 4Paradigm has turned to investments. In August 2020, Beijing Paradigm Private Fund Management Co., Ltd. was established, and in June 2021, Paradigm Fund registered with the Asset Management Association of China. Today, Paradigm Fund has attracted multiple LPs, marking 4Paradigm's entry into venture capital.

Tianyancha data shows that as of September 28, 2023, Paradigm Fund has made eight investments, covering fields such as artificial intelligence, enterprise services, big data, IoT, and advanced manufacturing.

In the current AI wave driven by large models, Fourth Paradigm has kept pace by launching SageGPT, an enterprise-grade generative AI product tailored for business scenarios, in March this year.

According to reports, SageGPT features multimodal interaction capabilities and enterprise-level AI tool characteristics, specifically designed for corporate users. It can connect to end-users' existing business systems, AI applications, and internal databases, processing queries and tasks in formats such as video, images, voice, and text. These queries and tasks are then transmitted to the end-users' existing business systems, with responses output in the required format. However, SageGPT is still in the early stages of commercialization.

As of the first quarter of 2023, Fourth Paradigm had cash and cash equivalents of 1.425 billion yuan. The current IPO fundraising will further increase cash reserves, providing support for R&D and future investments.