Summary in Plain Language
Suiyuan Technology, a domestic AI chip manufacturer that has been in business for 8 and a half years, officially listed on the STAR Market in September. The initial offering price was 142.18 yuan, but the stock price soared to 410 yuan right at the opening, with a daily increase of nearly 190%. The market value on the first day exceeded 170 billion yuan, making it the latest addition to the wave of domestic AI chip companies going public in the past year.
Unlike most of its competitors, Suiyuan did not follow the common route of being compatible with NVIDIA’s CUDA ecosystem. Instead, it chose a customized architecture optimized for AI computing. Thanks to Tencent’s extensive investment from the early stages and support with actual use cases, Suiyuan started by focusing on AI inference chips, which have a lower entry barrier and higher demand. It has already sold nearly 200,000 units of these chips, and its revenue in the first half of 2026 increased by 279% year-on-year. With the 6.1 billion yuan raised from this listing, Suiyuan faces three critical challenges: acquiring new major customers beyond Tencent, mass-producing more challenging AI training chips, and gradually reducing losses to achieve profitability. Only by overcoming these challenges can it establish a solid foothold in the highly competitive domestic AI chip market.
---
Detailed Explanation
1. The recent wave of domestic AI chip companies going public is not about cashing out immediately; it represents a critical phase of the industry’s “money-burning” phase for market dominance
Many people might think that the successive listings of companies like Moore Threads, Muxi, Birun, Tianshu ZhiXin, and Suiyuan are just about raising funds. However, that’s not the case. In the past seven to eight years, the main goal of these companies was to develop the chips themselves, and venture capital was sufficient for research and development. But now the industry has moved on to a stage where chips need to be sold in large quantities, production capacity must reach millions, and the next generation of chips must be developed every 1-2 years. The pace of spending on this has accelerated significantly. Global spending on AI infrastructure is expected to reach 487 billion US dollars by 2026, and customers often purchase thousands of chips for large systems. Companies need to prepare production capacity, build after-sales teams, and invest in the next generation of chips, which means spending billions of dollars each year. The initial venture capital is no longer enough.
Of the 6.1 billion yuan raised by Suiyuan, over 2.7 billion will be invested in the research and development of the fifth and sixth-generation chips, while the remaining 3.3 billion will be used to develop a technology system that integrates software and hardware. Raising this money from the stock market allows Suiyuan to maintain the founders’ equity and build confidence with its partners and major customers, essentially securing a spot in the final rounds of the industry competition.
2. Suiyuan’s decision to avoid compatibility with NVIDIA’s CUDA is not a whim; it’s a strategic move to bypass a bottleneck
Some may wonder why Suiyuan would choose a different approach instead of using NVIDIA’s established ecosystem. Here’s an analogy: NVIDIA’s CUDA is like the “Android system” for the entire AI industry. All developers are accustomed to writing code on this system. If a company wants to make its chips compatible with CUDA, it’s like creating a new phone that must run all Android apps smoothly. This would mean constantly updating according to NVIDIA’s changes, making it impossible to optimize the chips for domestic customers’ specific needs. Suiyuan’s custom DSA architecture, on the other hand, is like creating a “customized knife” for AI computing, removing all unnecessary features and focusing all computing power on AI tasks. This approach is more efficient, especially in current use cases where AI is widely used for tasks like chatting, video ad recommendations, and virtual meetings. Suiyuan’s second-generation chips have sold over 30,000 units, and the third generation has already shipped 160,000 units, proving that this approach can lead to large-scale orders.
3. Leveraging Tencent’s support is a significant advantage, but it also poses a risk
Suiyuan’s deep partnership with Tencent is unique in the industry. Tencent has been investing in the company since its inception, with a total investment of over 2 billion yuan and currently holding a 20% stake. Last year, 80% of Suiyuan’s revenue came from Tencent, making it both the owner and the largest customer. This model offers significant benefits: other companies struggle to find real-world use cases for their chips and can only test them in laboratories. Tencent, with its vast user base in social media, gaming, meetings, and search, allows Suiyuan to test its chips in actual applications, accelerating product development by 2-3 times. However, this also means that Suiyuan’s revenue is heavily dependent on Tencent. Nearly half of the domestic AI market is controlled by Tencent, Alibaba, and ByteDance, so any change in their chip strategies could significantly impact Suiyuan’s performance. Suiyuan’s gross margin has dropped from 40% to 32% due to discounts for major customers and rising storage chip prices. Relying solely on Tencent for orders makes it difficult to increase profits.
4. The domestic AI chip market is no longer a blue ocean; it’s a red ocean dominated by a few players
Although NVIDIA holds 55% of the domestic market, domestic companies like Huawei have already captured 41% of the market share. Other players, such as Cambricon, Alibaba, and ByteDance, are also developing their own chips. Customers now purchase large quantities of chips for complex systems, requiring additional services like customized software, high-speed interconnectivity, and 24/7 maintenance. Suiyuan’s R&D expenses exceed its revenue, and even with Tencent’s support, its profit margins are under pressure. The industry consensus is that only the top players, capable of continuously investing in chip development and securing large orders, will survive. Small companies will be eliminated within three years. The funds raised from this listing will help Suiyuan compete for a place among the leading players.
5. Shanghai’s role in fostering AI chip companies
Shanghai has seen the largest number of AI chip companies go public, thanks to its well-developed industrial chain. The city’s integrated circuit industry is worth nearly 600 billion yuan and employs over 1,200 companies. Companies like Suiyuan, which focus on chip design and production, rely heavily on this ecosystem. In other regions, companies may have to travel across the country to find foundries, packaging and testing services, and EDA tools. In Shanghai, all these services are available within a short drive, significantly accelerating product development. Local industrial funds and government support also contribute to Suiyuan’s success. This comprehensive support system has helped the company go public in just eight years and achieve annual revenue of over 1 billion yuan.
In summary, Suiyuan’s listing is a strategic move in a highly competitive market where only the best players, those able to invest in continuous innovation and secure large orders, will survive. The success of companies like Suiyuan is a testament to Shanghai’s strong industrial foundation.