A milestone in AI adoption
Inference is the process of running trained AI models to respond to user requests, which involves processing data units called tokens. Token calls through public cloud providers are an indicator of AI usage. The report did not disclose how much data is used to train AI systems in China.
According to the China Academy of Information and Communications Technology, including private cloud usage, China’s average daily token consumption was about 500 trillion as of June.
The National Data Administration estimates computing demand for inference could eventually be three times that for training.

















