On February 27, China's three leading computing chip companies—Cambricon Technologies Corporation Limited (688256.SH), Moore Threads Technology Co.,Ltd. (688795.SH), and Metax Integrated Circuits (Shanghai) Co., Ltd. (688802.SH)—released their 2025 performance reports on the same day. The data revealed triple-digit revenue growth for all three firms, with Cambricon achieving a net profit of 2.059 billion yuan, marking its first annual profit since listing. These nearly simultaneous reports send a strong signal: amid the ongoing artificial intelligence boom, domestic computing chip manufacturers are collectively crossing a critical threshold from research breakthroughs to commercial application.
The collective strong performance of domestic computing chip companies is underpinned by sustained explosive growth in global AI computing demand. At the same time as NVIDIA's financial report showed a 75% surge in its data center business, the rise of domestic players is quietly reshaping the global computing landscape.
Cambricon stands out as the brightest star. As the first AI chip stock on the STAR Market, its 2025 revenue reached 6.497 billion yuan, a staggering increase of 453.21% year-over-year. With a net profit of 2.059 billion yuan, the company successfully turned a profit, finally shedding the label of negative non-GAAP net profit it had carried for years since its IPO and reaching a genuine commercial inflection point. This historic breakthrough is primarily attributed to the large-scale deployment of its Siyuan 590 and 690 series cloud accelerator cards in large model training and inference scenarios. Notably, despite the annual profit, its fourth-quarter 2025 net profit declined by 19.8% quarter-over-quarter, marking a second consecutive quarterly drop. This indicates a period of volatility and adjustment following rapid growth, raising market concerns about the sustainability of its future profitability. An investment banking source close to Cambricon analyzed that this suggests the market is entering a new, more rational phase focused on long-term stability after a round of policy-driven concentrated procurement.
Moore Threads also demonstrated robust growth. The company reported 2025 revenue of 1.506 billion yuan, up 243.37% year-over-year. Although still unprofitable with a net loss of 1.024 billion yuan, its loss narrowed significantly by 36.7% compared to the previous year. The core driver was the mass production of its flagship integrated training-and-inference GPU computing card, the MTT S5000. This product has completed deep adaptation for domestic large models like GLM-5 and Qwen3.5, forming a "domestic chip + domestic model" technological closed loop.
Metax Integrated Circuits also delivered solid results. The company reported 2025 revenue of 1.644 billion yuan, an increase of 121.26% year-over-year. Its net loss was 781 million yuan, narrowing by 44.53% compared to the previous year. Through product lines such as "Xisi N," "Xiyun C," and "Xicai G," the company has achieved comprehensive coverage across four major computing segments: AI inference, AI training, graphics rendering, and scientific intelligence. It is accelerating the development of its next-generation flagship product, the C700, targeting performance comparable to NVIDIA's H100. Notably, Metax completed its IPO on the STAR Market, leading to a substantial expansion of its asset base. As of the end of 2025, the company's total assets reached 13.674 billion yuan, a growth of 251.56% from the beginning of the period, providing ample resources for subsequent high-intensity R&D.
"The three companies show a clear tiered characteristic. Cambricon, leveraging its first-mover advantage, has reached profitability first, while Metax and Moore Threads are on the verge of value realization," commented a semiconductor industry insider. "Their commonality lies in seizing the historic opportunity presented by the explosion of domestic large models, deeply embedding their products into the ecosystems of mainstream models like Zhipu AI and Alibaba's Tongyi, achieving a leap from 'usable' to 'effective.'" Their differentiation is reflected in their technical routes: Cambricon focuses on dedicated AI accelerators (ASIC), while Moore Threads and Metax primarily pursue general-purpose GPUs (GPGPU).
The performance surge of domestic computing chip companies occurs against a backdrop of unprecedented global demand for computing power. NVIDIA's latest financial report shows its Data Center segment revenue for Q4 fiscal 2026 reached $62.3 billion, accounting for over 91% of total revenue. NVIDIA CEO Jensen Huang asserted that in the AI era, computing equals revenue, outlining a vast market space for the entire industry. Simultaneously, the collective rise of domestic chips coincides with a rare strategic window of opportunity. "The escalating U.S. restrictions on AI chip exports to China have unexpectedly freed up market space for domestic companies. NVIDIA's China-specific H20 chip has yet to generate any revenue, prompting domestic cloud providers and AI companies to accelerate procurement of domestic alternatives," the semiconductor insider stated, suggesting this provides a valuable breathing space and development time for domestic solutions.
However, challenges remain formidable. The same insider pointed to three key areas of disparity: First, the technological gap. While domestic chips can serve as substitutes in specific scenarios, there is still a significant gap compared to international giants in terms of general computing capability and the maturity of software ecosystems (such as alternatives to CUDA). Second, profitability pressure. Apart from Cambricon, Moore Threads and Metax remain in a phase of high R&D investment and significant losses. Achieving the transition from "revenue growth" to "profit realization" is a common challenge for all. Finally, ecosystem building. The success of a chip depends not only on hardware performance but also on the ability to build a thriving developer community and application ecosystem, which requires long-term, sustained investment.
Industry observers believe the competition is now entering a more challenging phase. On one hand, application scenarios will expand from large model training to broader fields like biopharmaceuticals, weather forecasting, and smart manufacturing, demanding chips with stronger general computing capabilities. On the other hand, synergy across the industrial chain has become more critical than ever. Weaknesses in any segment, from EDA design tools to advanced process manufacturing, could become bottlenecks for the entire industry. Facing this situation, coordinated efforts from all sides are needed. The investment banking source mentioned that the investment community needs to view this "decades-in-the-making" industry with greater patience, avoiding wavering confidence due to short-term performance fluctuations. For policymakers, the focus of support should shift from mere hardware subsidies to systematic backing for foundational software, talent cultivation, and application ecosystem development.
As global AI computing demand surges like a tidal wave, domestic computing chip companies are moving from the laboratory to the real commercial battlefield. Their collective strong performance is a milestone, not a finish line. It declares that domestic computing chips have attained preliminary commercial viability but also heralds the arrival of even fiercer market competition. Their story is no longer just about breakthroughs in technical specifications, but about orders, customers, and real profits.
Comments