China's AI Industry Shifts from Training to Large-Scale Inference
A report from the CCID reveals that Chinese AI inference data volume has surpassed training data, marking a transition toward widespread industrial deployment.
The China Center for Information Industry Development (CCID) has reported that artificial intelligence in China has entered a phase of large-scale industrial application. Data indicates that inference volume—the active use of models—now exceeds training data volume, a shift that researchers associate with the transition from laboratory research to commercial utility.
As of June 2026, daily token usage reached over 175 trillion, representing a growth of more than 1,000 times since the beginning of 2024. This surge accompanies a significant compression in model iteration cycles, which have moved from months down to a matter of days. Experts note this shift reflects a move away from model development toward integrated, real-world utility.
The integration of AI into manufacturing sectors is supporting this trend, with over 500 smart factories established across China. Domestic computing power and large models are currently deployed in more than 1,000 use cases across 20 industries, including automotive and steel production. With the total user base for generative AI in China exceeding 700 million as of mid-2026, the technology's reach now spans more than half the population.