Suiyuan Tech, backed by Tencent, files for IPO with a $612 billion valuation

Summary:

Class EE85 of Tsinghua University (formerly the Department of Radio Engineering, later renamed the Department of Electronic Engineering) holds for China’s semiconductor industry the same stature that the Yao Class does for China’s artificial intelligence industry. Zhao Lidong, founder of Enflame Technologies, graduated from this class. While his classmates had already surged into the capital market around 2017 and became China’s semiconductor billionaires, Zhao Lidong was the last from this elite Tsinghua cohort to enter the wealth-creation wave—but he is also destined to shoulder an even greater mission.

China's "Four Little Dragons" of domestic GPUs will today complete the final piece of the puzzle.

Suiyuan Technology opened subscription today at an issue price of RMB 142.18 per share, valuing the company at approximately RMB 61.2 billion—what justifies this valuation for a company that has lost nearly RMB 5 billion over three and a half years?

The answer is hidden in a name: Tencent. Prior to the offering, Tencent Technology held approximately 19.95% directly, and the Tencent ecosystem collectively held 20.26%, making it the largest institutional shareholder; additionally, in 2025, 83.79% of Suanshuan’s revenue came from Tencent.

Tencent’s confidence to invest six rounds over eight years and bet on its core AI business lies in the hands of two AMD veterans, Zhao Lidong and Zhang Yalin.

Today (September 2), Suiren Technology, one of China's "Four Little Dragons" of domestic GPUs, officially launched its subscription. At an issue price of RMB 142.18 per share, and with a lot size of 500 shares on the STAR Market, investors who win a lot will need to pay approximately RMB 71,100.

Preliminary inquiries included participation from 337 offline investors and 11,805 allocation objects; after excluding invalid bids and the highest bids, the total offline subscription amount reached 72.959 billion shares, equivalent to 2,648.93 times the initial offline offering size.

Looking at the strategic allocation list, Xiaomi, GigaDevice, Futong Microelectronics, ZTE, and the National Social Security Fund are all included.

What does this indicate? It shows that institutions aren't just talking up their confidence—they're actively buying in. But for retail investors, is it worthwhile to invest in a domestic GPU company that hasn't even gone public yet?

If you look only at the recent performance of the last few "seniors," the answer seems easy to guess.

On December 5, 2025, Moore Threads became the first company to list on the STAR Market, with an offering price of RMB 114.28 per share. On its first trading day, the stock peaked at a 425% increase, pushing its market capitalization above RMB 300 billion. For investors who subscribed to one lot of 500 shares, the maximum unrealized profit exceeded RMB 200,000.

Following closely behind, Muxi Shares saw even more explosive trading: on its first day, the price surged as much as 700%, with the maximum unrealized profit per lot nearing RMB 400,000, setting a new record for the highest single-lot profit on the first day of listing on the A-share market in nearly a decade.

BiRen Technology, which listed on the Hong Kong Stock Exchange on January 2 this year, did not replicate the surges of its two predecessors, but still rose 75.82% on its first day, with its market capitalization briefly exceeding HK$100 billion.

Among the four companies, Suiren Technology, the oldest but the last to go public, has finally stepped into the spotlight of the capital market after eight years of operation.

Chart | Performance of China's Four Major Domestic GPU Startups on Their First Day of Listing

The first three companies have consistently left the market with the impression that "domestically developed GPU IPOs = profit." Will Suoan continue this trend?

Looking at his performance, he has indeed been moving quite fast in recent years.

From 2023 to 2025, Suiren Technology's revenue amounted to RMB 301 million, RMB 722 million, and RMB 990 million, respectively, achieving a three-year CAGR of 81.36%.

This year's momentum is even stronger. In the first half of the year, the company achieved revenue of RMB 1.12 billion, matching last year's full-year revenue in just six months. The company also forecasts that revenue for the first nine months will reach RMB 2.3 to 3.0 billion, representing a year-over-year growth of up to 455%.

Revenue is rising, and losses are narrowing. For a GPU company that has been in operation for eight years, the market’s primary concern in the past was whether its products could be sold; today, the question has shifted to when it will become profitable.

The timeline provided in the prospectus indicates that the company is expected to achieve near break-even by 2026 and become profitable no later than 2027.

If it can continue at this pace, Suanshuan may become the first among these domestic AI chip companies to achieve annual profitability.

However, investors must calculate the valuation: at the offering price of RMB 142.18, Suiren Technology implies a forward static price-to-sales ratio of approximately 61.8x for 2025 on a diluted basis, which is below the industry average for comparable companies. At first glance, this figure may seem reasonable, but after the stock begins trading, the market price often tells a different story.

Not long ago, Unitree Robotics, which had just gone public, experienced such a dramatic fluctuation. Prior to its listing, the company’s market capitalization was approximately RMB 61 billion, with market optimism projecting around RMB 200 billion. However, on August 19, after opening, its stock price surged to RMB 1,100, corresponding to a market cap of RMB 444.9 billion. Following this peak, the stock price steadily declined; by the open on September 2, the market cap had retreated to approximately RMB 221 billion, nearly halving from its highest point.

Of course, this alone cannot be used to predict how Suanyuan will perform, but it does at least show that the price of a new stock on its listing day may be entirely different from what institutions consider its fair value.

Back to Suiyuan, is a P/S ratio of 61.8x expensive? Ultimately, it depends on whether its future performance can keep up.

On Suanyuan, one name always comes up: Tencent.

The story of these two companies begins in 2018. As tech tensions between China and the U.S. escalated, China’s major internet giants collectively experienced growing anxiety over their dependence on NVIDIA.

As one of China’s largest buyers of AI computing power, Tencent requires substantial computing resources for WeChat, ad recommendations, content security, and later, the HunYuan large model—so it must secure stable domestic suppliers as soon as possible.

At that time, Moore Threads, Muxi, and Biren Technology had not yet been founded, and there were very few mature options available on the market—Sunway was one of the very few teams that had already begun this work. Therefore, when Sunway Technology was established in March 2018, Tencent led its Pre-A round of funding.

In addition, Zhao Lidong, a graduate of Tsinghua University’s Class EE85, is also one of the reasons Tencent was willing to make such a bold bet. In 1992, after graduating from Tsinghua as a high-tech talent, Zhao moved to the United States to further his studies and spent many years working in Silicon Valley on semiconductor chips, holding key positions such as Senior Director of AMD’s Computing Division and Senior Director of Product Engineering. He played a central role in establishing AMD’s China R&D center. In 2014, when China launched its first major semiconductor fund, Zhao resigned and returned to China to join Unigroup Communications Technology Group Co., Ltd. as Vice President, overseeing semiconductor investments.

In 2018, another pivotal turning point arrived when Zhao Lidong left his job at Unigroup and brought on another AMD veteran, Zhang Yalin.

Zhang Yalin has been working in chip design since 2000, later joining AMD, where he served as Senior Chip Manager and Technical Director of the China R&D Center. As one of AMD’s primary global leads for chip development, he spearheaded the design and mass production of multiple products, including the main chip for Microsoft’s Xbox One.

In its second year of establishment, Tencent and Enflame began collaborating. Over the next six years, Enflame’s chips progressed from peripheral use cases such as Tencent Meeting and OCR recognition to becoming integral to Tencent’s core AI operations; by the end of 2024, the Wan-card inference cluster in Qingyang, Gansu, was deployed, with Enflame’s S60 emerging as one of its core computing products—the first time a domestically developed AI chip had been tested at a ten-thousand-card scale in real-world business applications.

This partnership also became a defining hallmark of Enflame: Tencent was both its earliest investor and one of its most important customers. But every situation has two sides.

For a chip startup, having a major client willing to partner in experimenting with core business operations can be lifesaving. Suiren’s eight-year journey, with four generations of architecture and five chips all achieving first-pass silicon success, was made possible only through rigorous, real-world testing and refinement.

But this is also a cause for concern for other investors. In 2025, revenue from Tencent’s direct sales and affiliated sales models accounted for 83.79% of Suanyuan’s total revenue, compared to just 33.34% in 2023. In other words, the more resources Suanyuan has received from Tencent over the years, the deeper its dependence on Tencent has become.

In particular, Tencent’s dual role as both the largest institutional shareholder and the largest customer has led to ongoing market skepticism about the company’s ability to independently acquire customers. In its prospectus, Suanguan Technology even explicitly stated: “We expect the high proportion of sales to Tencent to continue in the future.”

Suiyuan also acknowledged that its performance would be directly impacted if Tencent's procurement strategy undergoes significant changes or if new customer acquisition falls short of expectations.

Additionally, there is another concern regarding Suiren’s product structure: most of its current products are primarily designed for inference scenarios. Although the fourth-generation training and inference integrated product, the L600, has been taped out, it has not yet entered large-scale mass production or delivery. This means that Suiren has not yet established a strong foothold in the larger market of large model training.

The four little dragons are all together; the story of China’s domestic GPUs has entered the second half—going public is just getting a ticket to the game; the second half is about who can truly take market share away from NVIDIA.

Suanyuan holds a strong hand: its proprietary DSA architecture, independence from the CUDA ecosystem, a major client in Tencent, and a clear path to profitability already on the table.

But these cards must ultimately become orders and profits to give the market a sustained reason to be bullish.

This article is from the WeChat official account "Phoenix Tech," authored by Phoenix Tech.