en.Wedoany.com Reported - In 2026, Cambricon will focus on the development needs of large models, prioritizing the iteration of intelligent processor micro-architectures, the research and development of a series of chip products covering different task scenarios, and the study of key technologies in advanced packaging, in order to seize the dual opportunities presented by the rapid development of the artificial intelligence industry and the dawn of the agent era. This intelligent chip company plans to optimize around the training and inference scenarios of language large models, image large models, multimodal large models, and vertical domain large models, accelerating the chip development process to adapt to market demands.

Zhang Yaqin, an academician of the Chinese Academy of Engineering, Chair Professor at Tsinghua University, and Dean of the Institute for AI Industry Research (AIR) at Tsinghua University, previously pointed out three major macro trends in the current development of artificial intelligence: moving from "generative AI" to "Agent AI," covering fields such as industry, consumer, and pharmaceuticals, with 2026 regarded as the "first year of Agent AI"; transitioning from "information intelligence" to "physical intelligence" and "biological intelligence," where robots and autonomous driving are applications of physical AI; and shifting from "AI" to "AI+," where AI becomes an economic tool integrated into various industries, requiring companies to manage with an AI mindset. Industry consensus holds that robust computing power is the key infrastructure supporting large models in the agent era to move from text to physical scenarios and achieve large-scale deployment.

Leveraging core technologies such as intelligent processor instruction sets, intelligent processor micro-architectures, intelligent chip programming languages, and intelligent chip mathematical libraries, Cambricon has developed a series of intelligent chip products and platform-based foundational system software that integrates cloud, edge, and terminal, combines software and hardware, merges training and inference, and features a unified ecosystem. The company believes that these core technologies, characterized by high barriers, difficult research and development, and wide applications, hold technological, economic, and ecological value for the integrated circuit industry and the artificial intelligence industry.

In terms of specific research and development plans, Cambricon will focus on the diverse needs of large models, accelerating the development of chips for large model training, chips for large language model inference, chips for multimodal inference, and switching chips for large model demands. At the same time, it will build a software platform for large models, integrating three core modules: a flexible compilation system, a training platform, and an inference platform, to enhance the usability and scenario adaptability of intelligent chips. Additionally, the company will conduct research on key technologies in advanced packaging for large model needs, establishing a specialized advanced packaging technology platform to adapt to the packaging requirements of differentiated products across various scenarios.

This article is compiled by Wedoany. All AI citations must indicate the source as "Wedoany". If there is any infringement or other issues, please notify us promptly, and we will modify or delete it accordingly. Email: news@wedoany.com