Moore, MooXi, Biren, and Denglin! The "Four Little Dragons of Domestic GPUs" gather for IPO.

```

As the global computing power competition enters a white-hot stage, China's GPU industry is ushering in a historical collective breakthrough moment.

Following Moore Threads' listing on the STAR Market with a market value of over 300 billion yuan, Muxi also made a strong debut, Biren Technology has just passed the Hong Kong Stock Exchange hearing, and Iluvatar CoreX has filed for an IPO and is ready to go—"China's Four Little Dragons of GPUs" are gathering in the capital market.

Although the four companies are on the same track, their differentiated paths have become very clear: from the pursuit of platform strategies for full-featured ecosystems, to focusing on engineering capability for single-quarter profitability, and even to binding with national-level computing power hubs through cluster routes, each company is seizing chip market share through different angles.

In this fierce competition, the differentiated competition paths of various companies have become clear. Iluvatar CoreX, with its first-mover advantage of "earliest mass production, most stable supply," relying on its two mature product lines, Tiangai (training) and Zhikai (inference), has achieved diversification of its customer base and robust revenue growth.

Moore Threads, which has just gone public on the STAR Market, carries the banner of “ecosystem is king,” connecting B-end and C-end markets with its full-featured GPU strategy based on the MUSA architecture. Its grand ecosystem blueprint and market value of over 300 billion yuan show capital's high expectations for its platform-level value.

At the same time, the other two giants are showing strong momentum in computing power infrastructure and commercialization efficiency. Muxi dazzled the market by completing a successful listing with a market value of 200 billion yuan, and became the first among the Four Little Dragons to achieve single-quarter profitability in Q2 2025, breaking the industry norm of ongoing losses and becoming the commercialization leader.

In contrast, Biren Technology, soon to become the "first domestically produced GPU stock in Hong Kong", is backed by a "national team" of capital, demonstrating strength in large-scale cluster deployment and cooperation with operators.

Iluvatar CoreX: First-mover advantage + Full product line, Rapid decline in customer concentration

As one of the earliest enterprises to start general-purpose GPU design in China, Iluvatar CoreX has established a strong position in the supply chain with the first-mover advantage of "earliest mass production, most stable supply." The prospectus shows it is the first enterprise in China to achieve dual mass production of "training + inference" general-purpose GPUs, and the first Chinese design company to achieve this milestone with a 7nm process.

The completeness of the product line is the company's core moat. Currently, Iluvatar CoreX owns two mature product lines—Tiangai (training) and Zhikai (inference)—covering AI model training and inference scenarios, forming a complete general-purpose GPU portfolio. The company’s cumulative shipment volume has exceeded 53,000 units.

This dual-driven product strategy allows the company to meet all-scenario needs from complex model development to efficient production deployment. Financial data shows that over half of revenue in the past three and a half years came from high-priced training product series, demonstrating competitiveness in the core computing power market.

The prospectus shows the company's revenue is experiencing rapid growth, reaching 324 million yuan in the first half of 2025, surpassing the whole of 2023. At the same time, the client structure has significantly improved, with dependence on single clients greatly reduced. In 2022, the top five clients accounted for as much as 94.2% of revenue; by H1 2025, this has dropped sharply to 38.6%, and the total number of clients has grown from 22 to 106.

The rapid decline in customer concentration shows Iluvatar CoreX has moved beyond reliance on a handful of early major clients; product versatility and market recognition are rapidly penetrating the entire industry. In addition, more than one third of the R&D team has over ten years of experience, and the executives include seasoned experts from NVIDIA, AMD, and other international giants, further ensuring the stability of technology iteration.

Moore Threads: Ecosystem is king, platform value expected

Moore Threads, newly listed on the STAR Market with a valuation of over 300 billion yuan, is at the forefront of its sector. The high market valuation is largely paying for its immense ecosystem potential.

Unlike manufacturers focusing only on single compute scenarios, Moore Threads emphasizes a full-featured GPU strategy. The company has launched four generations of GPUs based on its self-developed MUSA architecture, achieving the breakthrough of supporting AI computation, graphics rendering, and scientific computing on a single chip. Its MTT S80 graphics card has single-precision floating-point computing power already approaching NVIDIA RTX3060, with some metrics at international advanced levels.

This full-featured layout allows Moore Threads to cover diverse needs from AI intelligence to personal computers. The prospectus shows the company sold 5 AI clusters in H1 2025, demonstrating product competitiveness and replacement potential in compute-heavy "bottleneck" areas like AI centers.

Although not profitable yet, Moore Threads shows strong growth momentum. The company expects 2025 revenue to reach as high as 1.498 billion yuan, with a YoY growth of over 240%, and gross margin has increased to about 70%. According to the plan, the company expects to reach profitability as early as 2027.

Investors are essentially paying for Moore Threads' long-term ecosystem imagination. As one of the very few GPU enterprises in China with simultaneous B-end and C-end deployment and support for all computing precision, once Moore Threads forms an ecosystem due to wide application adaptation, its platform will have a powerful amplification effect. This “ecosystem is king” path is the core of NVIDIA’s own growth logic.

Biren Technology: Backed by the National Team, first choice for computing power hubs

Biren Technology, which has passed its HKEX hearing, has taken a unique path deeply tied to the “national computing power upgrade dividend.”

On December 17, according to the HKEX, Biren Technology passed the hearing and is about to become "the first domestically produced GPU stock in Hong Kong." The prospectus shows that the company’s 2024 revenue rose to 337 million yuan, with a two-year compound growth rate of up to 2,500%.

The company is backed by a luxury "national team" capital lineup, including Shanghai Guotou Leading Fund, Guangzhou Industrial Investment, and Ping An Group. This shareholder background gives it a natural advantage in undertaking national-level computing centers and supercomputer projects.

In terms of technology, Biren Technology focuses on “kilocard-level” clusters, chiplet, and optical interconnect technologies, specializing in solving system-level challenges in large-scale computing deployment. Data shows that in 2024, the company established deep partnership with China’s three major telecom operators and successfully delivered a 1.8-billion-yuan, 1024-GPU intelligent computing cluster project. This big order delivery ability proves its product maturity for carrier-grade high-reliability scenarios.

Although the revenue scale is still climbing, Biren Technology’s growth is explosive, with a compound annual growth rate of up to 2,500%. As its Birly series products deepen their applications in AI large model training and inference—especially rapid adaptation to cutting-edge models like DeepSeek—Biren Technology is steadily establishing itself as a core supplier of computing infrastructure.

Muxi: Fastest commercialization, first to achieve single-quarter profitability

Compared to the other three, Muxi shows exceptionally strong engineering and commercialization capability. With deep AMD technical roots, the company saw a surge in commercialization just four years after its founding, breaking the industry’s prevailing losses norm.

According to information from December 17, Muxi’s opening price was 700.00 yuan, a surge of 568.83% from its IPO price. Its current market cap is 281.4 billion yuan, becoming the second listed domestic GPU company after Moore Threads. As a national-level "specialized new little giant," it took less than 6 months from IPO application to listing,

More crucially, Muxi has reached a financial turning point ahead of peers—in Q2 2025, it achieved parent net profit of 46.619 million yuan, the earliest among the Four Little Dragons to post single-quarter profitability. From January to September 2025, losses narrowed significantly, with net profit at -346 million yuan.

Muxi’s approach is pragmatic and highly efficient. Rather than chasing concepts, it relies on the AMD team’s strong product deployment ability to quickly launch three major product lines covering AI computing, general computing, and graphics rendering. In H1 2025 alone, it achieved several times revenue growth and expects full-year revenue to exceed 1.5 billion yuan.

The company’s highly efficient execution comes from its technical team; core executives Chen Weiliang, Peng Li, Yang Jian, etc., all served long-term at AMD Shanghai and have strong product engineering capability. This technical foundation drives the rapid iteration of the three major product lines: the domestically supplied Xiyun C600 chip is expected to enter risk production by year-end, with the flagship Xiyun C700 aiming to match NVIDIA’s H100. The company’s cumulative GPU sales have exceeded 25,000 units, and it has completed deployments on multiple national and telecom AI computing platforms.

Risk Warning and DisclaimerThe market carries risks; investment requires caution. This article does not constitute individual investment advice, nor does it consider the special investment objectives, financial situation, or needs of individual users. Users should consider whether any opinions, views, or conclusions in this article fit their own situations. Investment based on this is at your own risk. ```