01 Cambricon: The Leading AI Chipmaker in High Demand
As the leading domestic AI chip manufacturer, Cambricon has recently shown remarkable performance. On August 22, Cambricon’s closing price reached 1243.2 yuan, with a single-day increase of 20%, and a rolling price-to-earnings ratio as high as 4006.77 times.
The company’s Q1 2025 results showed explosive growth, with revenue reaching 1.111 billion yuan, a year-on-year increase of 4230%; net profit was 355 million yuan, achieving profitability for the first time since its listing10.
Founded in 2016, Cambricon originated from a 10-person academic team established in 2008 by the Institute of Computing Technology, Chinese Academy of Sciences, focusing on the intersection of processor architecture and AI10. The company specializes in the research and development of AI chip products, with its main business involving the R&D, design, and sales of core AI chips used in various cloud servers, edge computing devices, and terminal devices10.
02 Haiguang Information: A Strong Performer with High Growth
Haiguang Information has also performed well, with a closing price of 186.06 yuan on August 22, a single-day increase of 20%, and a rolling price-to-earnings ratio of 189.76 times5.
In the first half of the year, the company achieved total operating revenue of 5.464 billion yuan, a year-on-year increase of 45.21%; net profit attributable to shareholders was 1.201 billion yuan, a year-on-year increase of 40.78%1.
Haiguang’s advantage lies in its compatibility with the x86 instruction set, giving it a competitive edge in the bidding for domestic servers, with a market share of over 50% in domestic server CPUs1.
The company continues to invest heavily in R&D, with R&D expenditure reaching 3.446 billion yuan in 2024, accounting for about 37% of revenue1.
Haiguang is currently advancing a major asset restructuring to absorb and merge with Zhongke Shuguang, raising supporting funds, which will create a complete industrial chain of “chip design – complete machine manufacturing” with strong business synergy7.
03 Accelerated Domestic Substitution
Amid the accelerated domestic substitution, the demand for self-controllable AI chips from domestic cloud vendors and major internet companies is rapidly increasing4.
The DeepSeek-V3.1 model adopts the UE8M0 FP8 precision format, specifically designed for the next generation of domestic chips, indicating that future training and inference based on the DeepSeek model will increasingly utilize domestic AI chips3.
Meanwhile, NVIDIA’s H20 chip, specifically supplied for the Chinese market, has encountered difficulties. Reports indicate that NVIDIA has suspended the production of the H20 AI chip for the Chinese market, a decision influenced by regulatory pressures from both the US and China3.
The results of China Mobile’s 5 billion AI server procurement show that packages supporting Huawei’s Ascend “CANN ecosystem” dominate, accounting for about 70% of both procurement amount and equipment quantity. This result highlights the increasingly important role of domestic computing power ecosystems in communication infrastructure3.
04 Ecological Layout and Strategic Positioning
Cambricon’s Siyuan 590 chip has completed adaptation to DeepSeek, and with the development of domestic large models, the demand for high-performance AI inference chips is quite strong6.
The company’s 3.985 billion yuan private placement plan has been approved by the Shanghai Stock Exchange and will be implemented after registration with the CSRC, aimed at the R&D of large model chip platforms3.
Haiguang Information has also established the “Photosynthesis Organization” in collaboration with the industry chain, forming deep partnerships with nearly 5000 partners, continuously strengthening ecological barriers1.
The company’s DCU products are developed based on the ROCm platform, with good CUDA compatibility, widely used in AI computing power infrastructure1.
The development paths of the two companies reflect different strategic choices: Cambricon focuses more on AI chips themselves, while Haiguang emphasizes compatibility with existing ecosystems and integration of the industrial chain.
05 Investment Considerations Amid High Valuations
Although both companies are viewed positively by the market, their valuations are relatively high. Cambricon’s rolling price-to-earnings ratio exceeds 4000 times, far above the semiconductor industry’s average valuation of 121.19 times2.
Haiguang’s valuation, while relatively lower, still reaches a rolling price-to-earnings ratio of 189.76 times, above the industry average5. In contrast, NVIDIA’s average price-to-earnings ratio in the US stock market over the past five years is only 77.78 times9.
Investors need to pay attention to the alignment between the company’s fundamentals and valuations, as well as changes in the competitive landscape of the AI chip industry4.
With the acceleration of technological iteration and the relentless innovation of international giants, domestic chip companies face significant competitive pressure9.
As technology iterates rapidly and international giants continue to innovate, domestic chip companies face significant competitive pressure9.
This chip competition is not just a contest between companies, but also a struggle for technological strength between nations. The rise of domestic AI chip duos is rewriting the competitive landscape of the global AI chip market. Let’s look forward to the rise of domestic chips.