云天励飞董事长陈宁：聚焦AI芯片战略 推动三大场景规模化落地

Core Viewpoint - The company, CloudWalk Technology, announced a strategic focus on AI chips during the 2025 World Artificial Intelligence Conference, aiming to develop a domestic computing "accelerator" centered around edge computing, cloud large model inference, and embodied intelligence [2] Group 1: Company Background and Development - CloudWalk Technology was established in 2014 and went public on the Shanghai Stock Exchange's Sci-Tech Innovation Board in 2023, with a focus on large-scale AI application [3] - The company has developed multiple innovative products across consumer, enterprise, and industry sectors over the past 11 years, starting with the NPU to reduce AI algorithm computation costs [3] - The second-generation NPU chip was launched in 2018, widely used in smart cameras, police terminals, robots, and industrial intelligence detection [3] - In 2023, the company released its third-generation chip, DeepEdge10, designed for large model inference needs, featuring a modular "computing block" architecture [4] Group 2: Market Outlook and Trends - The company anticipates a significant increase in inference computing demand as AI transitions from a training era to an inference era, with 2025 marking a pivotal point in AI development [5] - The inference chip market is expected to grow exponentially, outpacing the training chip market, which is currently valued in the hundreds of billions [6] - CloudWalk Technology predicts that AI will redefine various electronic products, enabling them to possess "active working" capabilities and interact like humans [6] Group 3: Focus Areas and Applications - The company is concentrating on three key scenarios: edge computing, cloud large model inference acceleration, and embodied intelligence, aiming to build a cost-effective AI inference chip technology and product system [7] - In edge computing, the company has implemented numerous demonstration projects and gained recognition from leading industry clients, focusing on low power consumption and high performance [7] - For cloud inference acceleration, the company has developed a reasoning acceleration card based on domestic technology, enhancing chip adaptability and product iteration speed [7] - In the field of embodied intelligence, partnerships with over ten robotics manufacturers have been established, with edge chip products entering deployment stages [8]