As a core demonstration factory, Wuhu Conch Cement has fully landed whole-process digital and intelligent transformation scenarios, achieving quality and efficiency gains, while national-level specialized and new little giant firm Keyi Intelligent has deployed AI equipment that detects defects and intelligently sorts products from a grain of rice to a shuttlecock, and from electronic components to auto parts.

Yan Tianxin, chairman of Keyi Intelligent, said the company's self-developed key technologies for AI-vision lightweight model inspection equipment overcame millisecond-level inspection on high-speed production lines and rapid model iteration across material categories, with measurement precision reaching 2 micrometers. Cloud-trained lightweight models are pushed to edge computing units on devices, cutting compute energy use without relying on large machine rooms.

On the research side, the traditional scientific paradigm is evolving toward AI for Science. At the University of Science and Technology of China's Science Intelligent Matter Creation Center, machine chemists run experiments 24 hours a day year-round, with data returned in real time to propose new hypotheses. Relying on the self-developed research agent, the center has realized a closed loop of large-model hypothesis generation, small-model verification, autonomous robot experimentation and iteration.

The Hefei Institutes of Physical Science has built the IIMedGPT omni-modal proactive health large model on domestic computing power, covering chronic disease rehabilitation, scientific fitness, community care and youth health. The institute has developed more than 40 types of proactive health intelligent devices across four categories, obtained over 20 Class-II medical device registrations, and deployed technologies in more than 50 tertiary hospitals and over 200 grassroots public health and sports service institutions, regularly serving more than 1 million people.

Gao Wen, director of the Pengcheng Laboratory, said China's computing network construction is advancing in three stages of computing aggregation, task scheduling and collaborative computing, with the goal of making computing power as convenient to call on demand across regions as electricity.