Summary:
On September 21, Hong Kong Stock Exchange documents showed that ESWIN passed its listing hearing. After three submissions, the second gamble of BOE founder Wang Dongsheng has finally reached the threshold of the Hong Kong Stock Exchange.
Produced by Phoenix Technology's "Fengyan Observation"
In 1993, 36-year-old Wang Dongsheng made his first gamble.
Beijing Electronics Tube Factory had been losing money for seven consecutive years. He led employees to pool 6.5 million yuan for a shareholding reform and founded "Beijing Orient Electronics Group Co., Ltd.," the predecessor of BOE.
The gamble he made was called "lack of screens."
At that time, China had no domestic LCD panel industry of its own, and the core screens of televisions and mobile phones all relied on imports. In 1994, BOE established a TFT-LCD project pre-research team, clearly entering liquid crystal displays; in 2003, it acquired the TFT-LCD business of South Korea's HYDIS for US$380 million and invested in building a fifth-generation production line in Beijing.
This was a typical asset-heavy, long-cycle technology route. Since then, BOE has continued to expand production and iterate its technology. By 2017, BOE ranked first globally in LCD panel shipments. Wang Dongsheng had his persistence and belief. "South Korea has Samsung, Japan has Panasonic and Sony. There is no reason we cannot create a world-class enterprise."
In June 2019, 62-year-old Wang Dongsheng stepped down as chairman of BOE.
No one believed he would truly retire. At that time, ZTE and Huawei were successively sanctioned, and "lack of chips" became a more painful real-world issue than "lack of screens." People in the industry joked: "You've done about enough with screens, why don't you do chips too."
So Wang Dongsheng made his second gamble.
On September 21, Hong Kong Stock Exchange documents showed that ESWIN passed its listing hearing, with CITIC Securities (Hong Kong) and China Securities International as joint sponsors. This means that after three submissions, Wang Dongsheng's second gamble has finally reached the threshold of the Hong Kong Stock Exchange.
If listed smoothly, ESWIN Computing will become the "first RISC-V stock" in the Hong Kong stock market, filling the long-standing gap of listed targets in this track.
But passing the hearing is only getting a ticket to enter. The real question remains: what exactly is Wang Dongsheng's second gamble, and whether the market is willing to follow.
Starting a business again: from "lack of screens" to "lack of chips"
In July 2019, Wang Dongsheng joined ESWIN. Regarding joining ESWIN, Wang Dongsheng clearly stated: based on long-term thinking about the integrated circuit field, he takes basic materials and chip underlying technologies as his focus, hoping to help the industry overcome difficulties on the manufacturing and product sides.
This has a typical Wang Dongsheng logic: "One should not do something just for the sake of doing it, but to solve difficult problems." At present, on the basic materials side, ESWIN Materials focuses on 12-inch silicon wafers and listed on the Shanghai Stock Exchange STAR Market in 2025; on the chip design and product side, it is ESWIN Computing.
In July 2026, ESWIN Computing submitted a prospectus to the Hong Kong Stock Exchange for the third time, aiming for a Hong Kong listing. In the narrative of the prospectus, ESWIN Computing tells an industrial story of "chip products based on the RISC-V architecture."
What is RISC-V?
In the world of chips, instruction set architecture is like the "language of chips"—it determines how software talks to hardware. Over the past few decades, general computing has mainly been dominated by two commercial architecture ecosystems, x86 and Arm. x86 has long occupied the PC and server markets, while Arm has formed a huge ecosystem in mobile and embedded fields.
RISC-V, however, is a "new language" born in 2010 in a laboratory at the University of California, Berkeley. It is an open, royalty-free instruction set standard, and any manufacturer can design its own processor based on it without purchasing ISA authorization from a single architecture owner.
Wang Dongsheng has expressed his judgment many times in public interviews: in the AI era, RISC-V will become a native computing architecture. On the one hand, it is green and open, continuously evolving and changing, and can better adapt to the needs of continuously optimizing software and algorithm models in the AI era. On the other hand, it can not only be compatible with the ecosystems of existing architectures through porting and adaptation, but can also serve as a native architecture to support emerging new applications and new scenarios.
Based on this judgment, Wang Dongsheng chose the RISC-V computing architecture. In recent years, the explosion of the chip industry has continued to validate Wang Dongsheng's forward-looking predictions. But, "choosing the RISC-V direction at that time required courage." Wang Dongsheng later recalled.
The problem is that openness does not equal ease. The truly difficult part of RISC-V is not just taping out a chip, but enabling it to form a sustainable software and hardware ecosystem in real products: whether compilers and toolchains are mature, whether operating systems and applications adapt smoothly, whether developers are willing to migrate, and whether chips, system software, and terminal products can continue to coordinate.
ESWIN's prospectus frankly admits: ARM and x86 already have mature ecosystems and complete software and hardware compatibility. In comparison, RISC-V is still developing, and ensuring seamless integration between different software platforms is challenging.
Therefore, Wang Dongsheng's second gamble is not really about broadly "making chips," but about whether RISC-V can move from an open instruction set to large-scale commercial application.
ESWIN Computing's "Customer A"
ESWIN Computing's products are mainly divided into two categories according to application scenarios:
The first category is intelligent terminal chips, specializing in home, office, and portable devices, enabling intelligent terminals to manage screen-based input/output and process multimedia signals. Prospectus data shows that human-machine interaction chip products among intelligent terminal chips contribute to ESWIN's base, achieving revenue of 1.856 billion yuan in 2025, accounting for 76.3% of total revenue.
In other words, what has long supported ESWIN Computing is mainly chips related to "screens."
However, this base is under pressure from prices and profit margins. The average selling price of human-machine interaction chips fell from 20.9 yuan in 2023 to 12.1 yuan in 2025, and further to 11.6 yuan in the first quarter of 2026; gross margin fell from 18.8% in 2023 to 16.2% in 2025, and further to 7.3% in the first quarter of 2026.
It is worth mentioning that the gross margin of the chip design industry is usually between 40% and 60%, with companies like Nvidia even reaching over 70%. A gross margin of 16.2% is almost "flying close to the ground" in the industry.
The second category is interconnect and computing chips, targeting automobiles, robots, and industrial equipment. Cars need to recognize road conditions, robots need to grasp objects, and factory equipment needs to automatically adjust processes. This type of chip is the hardware foundation that provides them with "senses" and "decision-making power."
This part is undoubtedly the story Wang Dongsheng most wants to tell. Revenue from interconnect and computing chip products reached 383 million yuan in 2025, a year-on-year increase of 103%, making it ESWIN's fastest-growing business, accounting for 15.8% of total revenue. The gross margin of this business also offers more room for imagination: the gross margin of computing chips was 26.2% in 2025, and interconnect chips had already reached 32.5% in the first quarter of 2026. But obviously, this part is still small in scale and cannot yet support the overall profitability of ESWIN Computing.
The prospectus discloses the company's largest customer anonymously as "Customer A," with the following description: Customer A is a leading IoT supplier registered and established in China, providing intelligent interaction products and professional services, and listed on the Shenzhen Stock Exchange. The industry generally infers that this customer is Wang Dongsheng's former employer—BOE. Prospectus data shows that prospectus data shows that from 2023 to 2025, revenue from the largest customer accounted for 82.1%, 76.8%, and 64.6%, respectively.
Undoubtedly, ESWIN Computing must open up customer markets beyond "Customer A." In Q1 2026, Customer I—a subsidiary of a computing power operator—contributed 194 million yuan in a single quarter, briefly becoming ESWIN Computing's largest customer. But this revenue was delivered on a project basis. And this does not mean the customer structure has completed diversification: in the same period, the top five customers still contributed 83.1% of revenue in total.
Wang Dongsheng cannot wait for the ecosystem to grow naturally.
In 2024, ESWIN launched the RISAA platform—a technology platform that packages RISC-V cores, dedicated IP, and open software, allowing developers to avoid building chips from scratch. He also proposed a bigger concept called RDI (RISC-V Digital Infrastructure), hoping to bring upstream and downstream players in the industry together to work on it.
Of course, platforms and concepts are one thing. Whether the RISC-V ecosystem can grow still depends on whether developers are willing to join and whether scenarios can be implemented.
The Unsolved Loss Problem
Opening ESWIN Computing's financial statements, the key word is still loss.
From 2023 to 2025, ESWIN Computing's operating revenue was 1.752 billion yuan, 2.025 billion yuan, and 2.431 billion yuan, respectively, continuing to grow; net losses during the same period were 1.837 billion yuan, 1.547 billion yuan, and 1.516 billion yuan, respectively. By the first quarter of 2026, the company achieved revenue of 494 million yuan but still lost 375 million yuan.
In the primary market, ESWIN Computing obtained multiple rounds of financing. According to the prospectus, before the IPO it had completed rounds A through D, with cumulative amounts exceeding 9.2 billion yuan. Investors include IDG Capital, Legend Capital, CITIC Securities, and others.
But chipmaking is a marathon run while bleeding.
R&D investment remains high, but the absolute amount is narrowing year by year. From 2023 to 2025, the company's R&D expenses were 1.445 billion yuan, 1.337 billion yuan, and 1.042 billion yuan, respectively, accounting for 82.5%, 66.0%, and 42.8% of revenue in the corresponding periods.
The operating cash flow data better reveals the consumption of this long run. From 2023 to 2025, ESWIN Computing's net cash flow from operating activities was -1.251 billion yuan, -781 million yuan, and -1.171 billion yuan, respectively, with a cumulative net outflow of 3.203 billion yuan over the three years.
ESWIN Computing has knocked on the door of the secondary market three times: it first submitted an application on May 30, 2025, submitted again on January 30, 2026, and submitted for the third time on July 31, 2026. For a chip company that is still losing money and needs continuous R&D investment, the significance of listing is more about buying time for itself before commercialization is fully realized.
Until September 21, 2026, this door is finally expected to open.
Even if it lists smoothly, ESWIN Computing still faces multiple tests: whether RISC-V can cross the gap between ecosystem and commercialization from an open architecture; whether it can truly reduce dependence on display chips and a small number of major customers; and whether the speed of large-scale realization can outrun the patience of capital.
Wang Dongsheng's second gamble has not yet reached the moment to reveal its cards.
This article comes from the WeChat public account: Phoenix Technology , author: Phoenix Technology, editor: Dong Yuqing
