Cambricon Technologies Corporation Limited, a globally renowned leader in AI chips, has been named in Forbes China's 2025 Top 50 Most Innovative Companies list under the computing infrastructure category. This recognition highlights Cambricon's role in advancing AI industry development through its cutting-edge intelligent chip technology.
The selection criteria for the list included innovation capability, corporate governance, growth potential, market competitiveness, and social impact. Cambricon stood out due to its strong technological foundation, robust financial performance, and continuous innovation.
The explosive growth of generative AI in 2025 has driven massive demand for AI computing power. As a fundamental computing platform for AI applications, intelligent chips have become a core pillar supporting smart industry development. Cambricon's general-purpose AI chips are designed for diverse AI scenarios, featuring a flexible instruction set and processor architecture optimized for AI algorithms. These chips efficiently support large-scale model training and inference, as well as multimodal AI tasks such as computer vision, speech processing, natural language processing, and recommendation systems.
Cambricon's solutions are compatible with mainstream open-source AI models, including DeepSeek, LLaMA, GPT, BLOOM, GLM, and Stable Diffusion, providing powerful computing support for industries such as cloud computing, energy, education, finance, telecommunications, healthcare, and internet services.
Since its inception, Cambricon has focused on AI chip R&D and technological innovation, mastering core technologies in intelligent chip design and system software development. The company has built a comprehensive intellectual property portfolio and launched multiple leading AI processors, including the Cambricon 1A, 1H, and 1M series for edge devices, as well as cloud and edge acceleration cards based on its MLU (Machine Learning Unit) chips.
Recently, Cambricon has expanded its technological adaptability and ecosystem partnerships. On September 29, the company successfully adapted DeepSeek's latest model, DeepSeek-V3.2-Exp, and open-sourced its vLLM-MLU inference engine. Through Triton operator development and BangC fusion optimization, Cambricon achieved industry-leading computational efficiency.
On September 30, Cambricon enabled GLM-4.6 to deploy on its domestic chips using FP8+Int4 hybrid quantization—a breakthrough solution that reduces inference costs while maintaining model accuracy. This advancement paves the way for localized large-model deployment on domestic chips.
In ecosystem collaboration, Cambricon recently signed a strategic agreement with SenseTime to jointly optimize hardware-software integration and foster an open AI industry ecosystem. The partnership will focus on domestic AI infrastructure, vertical market expansion, and global technology deployment.
With its large-scale industry adoption and continuous breakthroughs in technology and ecosystem development, Cambricon is reinforcing its leadership in AI chips. As the AI industry evolves, the company is poised to drive growth in domestic chip innovation and accelerate intelligent transformation across sectors.
Comments