Economic Observation Network Xu Mengyi/Wen Nearly a year after being included in the "Entity List" by the United States, the domestic GPU (graphics processor) unicorn Shanghai Bicheng Technology Co., Ltd. (hereinafter referred to as "Bicheng Technology") entered the listing counseling procedure. On September 12, the information on the official website of the China Securities Regulatory Commission showed that Bichen Technology had been registered with the Shanghai Securities Regulatory Bureau for counseling, and planned to make an initial public offering of shares and go public, and the counseling brokerage was Guotai Junan.

This means that, following the Cambrian (688256. SH) and Shanghai Suiyuan Technology Co., Ltd. (hereinafter referred to as "Suiyuan Technology"), another AI chip company began to start the IPO process.

On September 13, Wang Peng, an associate researcher at the Institute of Management of the Beijing Academy of Social Sciences, told the Economic Observer that the centralized listing of domestic AI GPU companies is an inevitable trend in the development of the industry, but it is also necessary to be wary of the risks brought by valuation bubbles.

He believes that relevant companies need to continue to strengthen technology research and development, expand application scenarios and improve industrial chain integration capabilities to meet the challenges of market competition.

The IPO process has been launched one after another

On October 17, 2023, 13 GPU entities, including Bicheng Technology, were added to the Bureau of Industry and Security (BIS) "Entity List" under the United States Department of Commerce. However, nearly a year after being included in the "Entity List", Bichen Technology announced its independent and original heterogeneous GPU collaborative training solution for the first time at the 2024 Global AI Chip Summit, which realizes the heterogeneous coexistence of domestic GPUs and NVIDIA GPUs, which is conducive to the accelerated landing and migration of domestic GPUs.

According to the information on the official website of Bichen Technology, up to now, Bichen Technology has completed the B round of financing, with a total financing amount of more than 5 billion yuan, and investors include Zhuhai Dahengqin Group, Gree Venture Capital, Ping An of China, New World Group, Country Garden Venture Capital, Qiming Venture Capital, IDG Capital, Walden International China Fund, Hillhouse Venture Capital, Source Code Capital, China Merchants Capital, CITIC Securities and other institutions. According to the "2024 Global Unicorn List" released by the Hurun Research Institute in April this year, Bichen Technology ranked 495th with a valuation of 15.5 billion yuan.

According to the research institute Verified Market Research, by 2025, the market size of China's GPGPU (General Computing Graphics Processing Unit) chip board will reach 45.8 billion yuan, which is 5 times the market size of 8.6 billion yuan in 2019; But on the other hand, the competition between domestic GPU companies is also becoming increasingly fierce - according to the information on the official website of the China Securities Regulatory Commission, on August 26, the listing counseling record of another GPU company, Suiyuan Technology, has also been accepted, and the counseling agency is CICC, which plans to IPO in A-shares.

Tianyancha information shows that in the six years since its establishment, Suiyuan Technology has carried out a total of ten rounds of financing, with a cumulative financing amount of nearly 7 billion yuan, and investors include the National Fund, Tencent Technology, Meitu, Zhen Fund, Red Dot China, etc. Among them, Tencent Technology has participated in six consecutive rounds of investment, becoming the largest shareholder of Suiyuan Technology, and is a major customer and business partner of Suiyuan Technology, accounting for 21.37% of the shares. According to the "2024 Global Unicorn List" released by the Hurun Research Institute, Suiyuan Technology is valued at 16 billion yuan, ranking 482nd in the world.

Wang Peng, an associate researcher at the Institute of Management of the Beijing Academy of Social Sciences, told the Economic Observer that it is not easy for these companies to "replace" Nvidia, and their products need to be the first to enter the supply chain under the fierce competition in the market.

Wang Peng believes that excellent domestic AI chip companies usually have strong industrial chain integration capabilities, and they can establish close cooperative relations with upstream and downstream enterprises, which can help companies reduce costs, improve efficiency and enhance market competitiveness.

According to the "2024-2029 China GPU Industry Market Status Survey and Development Trend Forecast Research Report" released by the China Business Industry Research Institute, the size of China's GPU market in 2023 will be 80.7 billion yuan, an increase of 32.78% over the previous year. In addition, the China Business Industry Research Institute predicts that the size of China's AI chip market will grow to 141.2 billion yuan in 2024, of which the GPU market size will increase to 107.3 billion yuan.

On September 13, Hong Yong, an associate researcher at the Institute of E-commerce of the Academy of International Trade and Economic Cooperation of the Ministry of Commerce, told the Economic Observer that domestic GPU companies have different advantages, for example, Biqian Technology has made significant breakthroughs in the field of high-performance computing with a strong R&D team and technology accumulation; Suiyuan Technology has a deep technical accumulation and product line layout in cloud training chips; Cambrian is known for its innovation in the design of AI processor architecture, and has core technologies and products with independent intellectual property rights.

Hong Yong also said that domestic AI chip companies need to face the deep technical and ecological barriers of giants such as NVIDIA, and in addition, the competition between domestic companies of the same type of commercial expansion and talent absorption is also testing the comprehensive ability of GPU unicorn companies, but these companies have their own focus on technology research and development, market positioning and strategic planning.

Achieving profitability is not easy

However, for domestic GPU companies, it is not easy to achieve profitable performance. Recently, Cambrian (688256. SH) released its latest performance announcement, and it has not been able to get rid of the loss dilemma after 4 years of listing, which has attracted market attention.

The financial report shows that in the first half of this year, Cambrian's revenue was only 64.7653 million yuan, a year-on-year decrease of 43.42%; The net profit attributable to the parent company was a loss of 530 million yuan, compared with a loss of 544 million yuan in the same period last year, and the non-net profit was a loss of 609 million yuan, compared with a loss of 641 million yuan in the same period last year.

From 2019 to 2023, Cambrian's net profit attributable to the parent company will lose 1.179 billion yuan, 435 million yuan, 825 million yuan, 1.257 billion yuan, and 848 million yuan respectively. From 2019 to the first half of 2024, Cambrian's net profit attributable to the parent company will have a cumulative loss of 5.074 billion yuan.

Regarding the loss, Cambrian said in its 2024 semi-annual report that the company's revenue decreased year-on-year due to unfavorable supply chain factors such as the "entity list". In addition, the company has not yet achieved profitability and has accumulated uncovered losses, mainly due to the continuous investment in R&D to ensure the high-quality iteration of smart chip products and basic system software platforms, and to maintain technological advantages in the highly competitive market.

Bai Wenxi, vice chairman of the China Enterprise Capital Alliance and a distinguished tutor of the Graduate School of the Chinese Academy of Social Sciences, told the Economic Observer that the excessive hype of AI technology in the market has led to the overvaluation of some companies, and the actual technological innovation and application value are not as expected by the market. In addition, some AI chip companies have the problem of over-reliance on financing and lack of self-profitability, which has also brought a valuation bubble to a certain extent.

According to public information, from 2020 to the first half of 2024, Cambrian's cumulative R&D expenses will be 4.992 billion yuan. In the first half of 2024, Cambrian's R&D expenses will be 447 million yuan, a year-on-year decrease of 7.27%, accounting for 690.92% of operating income.

Guo Tao, an expert member of the Internet Society of China and an angel investor, told the Economic Observer that there are problems such as high barriers and difficult ecological construction in GPU technology, and domestic GPU manufacturers need sufficient funds to support R&D investment in GPU chips.

In addition, he believes that the fierce competition pattern of the domestic AI chip industry and the rapid technological upgrading also increase the uncertainty of the valuation bubble of GPU companies, and if the relevant companies cannot maintain their technological leadership or adapt to market changes, their valuations may be significantly affected.

At the same time, the global market demand for AI chips is continuing to expand. According to the latest forecast report of technical analysis firm Gartner, the global sales revenue of artificial intelligence chips will increase by 33% from last year to $71 billion in 2024 and will continue to grow by 29% to $92 billion in 2025.

Wang Peng believes that different domestic AI chip companies have their own focus on application scenarios. For example, some companies focus on areas such as intelligent driving and intelligent security, while others are more focused on areas such as cloud computing and big data processing. This differentiated competition will help domestic GPU companies better meet the needs of the global market and further expand their market share.