
About Cambrian
Cambricon Technologies Co., Ltd. (Stock Code: 688256) was established on March 15, 2016, and is headquartered in Haidian District, Beijing. It was jointly founded by Chen Tianshi’s team and the Institute of Computing Technology of the Chinese Academy of Sciences. As a global pioneer in the field of intelligent chips, Cambricon focuses on the research, development, and technological innovation of artificial intelligence chips. Adopting a fabless model, the company provides integrated intelligent chip and system software solutions spanning cloud, edge, and endpoint applications. The company’s core business spans servers, edge computing devices, and end-user devices. Its products are widely used in sectors such as the internet, finance, transportation, energy, and manufacturing, and the company leads several national-level intelligent computing projects. As of August 2025, CambriconMarket CapitalizationIt surpassed 600 billion yuan, becoming one of the tech companies with the highest market capitalization on the A-share market, and has been hailed as the “Chinese version of NVIDIA.”
Development history: from technology catching up to leading independently
The development of Cambrian can be divided into three key stages:
1. Start-up period (2016-2019): terminal IP licensing with Huawei
Cambrian was initially founded with terminal intelligent processor IP licensing as its core business, and the world's first commercial terminal AI processor was launched in 2017Cambrian 1AEquipped with Huawei's Kirin 970 chip, it was applied to Mate 10 series cell phones, realizing the first commercialization of AI chips. Since then, Cambrian 1H and 1M series processors have been cumulatively integrated into more than 100 million intelligent terminal devices. However, with Huawei Hesi's shift to self-developed chips, Cambrian's terminal business revenue plummeted, and the proportion of this business fell to less than 101 TP4T in 2019 from 99.691 TP4T in 2018.
2. Transformation Pain Period (2020-2023): Government Programs and Cloud Exploration
In the face of terminal business shrinkage, Cambrian turned to the government intelligent computing cluster system business, and the revenue share of this business in 2019 amounted to 66.72%. but by the impact of policy fluctuations, the revenue in the first half of 2020 was only 100-200,000 yuan. During the same period, the company accelerated the layout of the cloud chip, in 2020, the launch of the Siyuan 270 cloud chip, in 2021, the release of the Siyuan 290, but the market performance was flat. 2022, Cambrian was included in the United States "entity list", the supply chain is facing challenges, but the process of technological autonomy accelerated.
3. Explosive period (2024-present): cloud-based chips dominate growth
In 2024, Cambrian seized the window of NVIDIA's A100/H100 chip ban on China to launch theSiwon 590Cloud chip, using 7nm process, performance close to the international mainstream level, energy efficiency ratio beyond the NVIDIA H20 chip, in the large model inference scenario latency reduction of 30%, cost reduction of 45%. the chip quickly gained the recognition of the Internet, finance and other areas of the head of the customer, to promote the explosive growth of the cloud business revenue: in 2024, the cloud chip revenues amounted to 1.166 billion yuan, year-on-year Surge 1187.78%; 2025 Q1 revenue accounted for 99.3%, becoming the company's core growth engine.
Products and Services: Integrated Layout of Cloud Edge and End
Cambrian has built a matrix of AI chip products covering all scenarios, forming a synergistic ecosystem of "cloud training + edge reasoning + terminal applications":
1. Cloud Product Line
- Siwon 590: Seventh-generation cloud AI chip, using self-developed MLUarch03 architecture, supporting FP16/TF32 mixed precision training, single card performance is comparable to NVIDIA A100, has entered the domestic supply chain.
- MLU370 Series Accelerator Card: Including MLU370-X8 (training card), MLU370-S4 (inference card), etc., supporting multi-card interconnection to meet the high-density arithmetic demand of data centers.
- Genius 1000 Intelligent Accelerator: Integration of 8 MLU chips, supporting the construction of arithmetic clusters across the entire machine, suitable for ultra-large-scale AI model training.
2. Edge Product Line
- Siwon 220: Edge-side AI chips, supporting low-power scenarios, with cumulative sales exceeding one million chips, used in intelligent transportation, security monitoring and other fields.
- MLU220-M.2 Accelerator Card: Localized AI reasoning capabilities for IoT devices.
3. IP Licensing and the Software Ecosystem
- MLU Instruction Set License: Licensing of self-developed intelligent processor architectures to customers, which have been integrated into more than 100 million end devices.
- Basic system software platform: Provide compilers, drivers, development toolchain, etc., support TensorFlow, PyTorch and other mainstream frameworks to reduce the threshold of customer development.
Core technology: full-stack self-research and ecological barriers
Cambrian's core competitiveness stems from a full-stack self-research technology system:
1. Smart Processor Instruction Set and Microarchitecture
The company independently researches and develops MLU instruction set to get rid of the dependence on ARM and x86 architectures, and realize the independent control of instruction set, micro-architecture and programming language. SiYuan 590 adopts Chiplet technology, which improves arithmetic density and reduces manufacturing cost through multi-chip interconnection.
2. Unified ecology and hardware and software synergy
Cambrian has built a cloud-side-end integrated hardware and software ecosystem, and its chip products are deeply adapted to the Zhongke Dawning server and Huawei Hongmeng system, supporting multimodal AI tasks and large-scale model deployment. Its software platform is compatible with mainstream AI frameworks, and its development efficiency is 30% higher than that of international competitors.
3. Intellectual property accumulation
As of 2025, Cambrian has accumulated 1,599 authorized patents, covering the entire process of chip design, manufacturing, packaging and testing, forming a technological barrier. 2024 R&D investment reached 1.072 billion yuan, accounting for 91.3% of revenue, and the R&D team accounted for 75.6%, continuing to promote technological innovation.
Development Prospects: Domestic Substitution and Globalization Opportunities
1. Accelerated domestic substitution
In the context of NVIDIA's high-end chip ban on China, Cambrian has become the main alternative to domestic AI arithmetic demand. Its cloud chips have been applied to the large model training of Internet giants such as ByteDance and Alibaba, and the coverage of the financial sector has exceeded 60%. It is expected that the size of the domestic AI chip market will reach 50 billion yuan in 2025, and Cambrian's market share is expected to exceed 10%.
2. Technology Iteration and Ecological Expansion
Cambrian plans to raise 3.985 billion yuan through a fixed-price increase for the research and development of large model chip platforms and software platforms to further improve product performance and ecological openness. Meanwhile, the company is exploring new areas such as automotive chips and robotics chips to expand its growth boundaries.
3. Globalization challenges and opportunities
Despite being included in the "Entity List" by the United States, Cambrian reduces risks through technological autonomy and supply chain diversification. Its products have entered the Southeast Asian and Middle Eastern markets, and are expected to seize international market share through its cost-effective advantage in the future.
4. Valuation and risk
As of August 2025, Cambrian's dynamic price-earnings ratio is over 500 times, much higher than the industry average, and the market has extremely high expectations for its future growth. The company needs to be alert to the risks of single large customer dependence, inventory turnover pressure and cash flow fluctuations. If it can continue to fulfill its performance commitments, Cambrian is expected to become an important player in the global AI chip field.
data statistics
Relevant Navigation

Focusing on the field of full-scale general-purpose humanoid robots, we leverage our full-stack proprietary technology to drive the large-scale application of robots from laboratories into diverse scenarios such as industrial, commercial, and domestic settings.

AAC Technologies
Specialized in providing acoustic, optical, haptic feedback, sensors and semiconductors and other perceptual experience solutions for consumer electronics, automotive electronics and emerging fields, and driving the interaction upgrade of smart devices with innovative technologies.

Huichuan Technology
Focusing on automation, digitalization and intelligence in the industrial field, we concentrate on the core technologies of "information layer, control layer, drive layer, execution layer and sensing layer".

SigmaStar
Focusing on the development and design of intelligent vision processing chips, the company has deeply laid out edge AI computing scenarios such as intelligent security, in-vehicle vision, robotics and smart home.

Yuanli Semiconductor
A company focusing on AI Chiplet technology, dedicated to providing cost-effective end-side large model inference arithmetic solutions for the AI Agent era through a “building block” architecture.

Exton
A leading provider of intelligent manufacturing solutions, specializing in the research, development, production and sales of high-end intelligent machinery and equipment, industrial robots and core control components.

Pilot Intelligence
Specialized in the research and development, production, sales and service of high-end automation equipment.

GuoChuang Software
A leading high-tech listed company in the field of software and information technology services, focusing on data intelligence, high trustworthy software technology, providing digital intelligent transformation solutions for operators, governmental enterprises, etc., and making a mark in the field of intelligent manufacturing.
No comments...
