Key facts
- CCID said China's inference data volume surpassed its training data volume for the first time.
- CCID said average daily token calls in China exceeded 175 trillion as of June 2026.
- CCID said that figure was more than 1,000 times the level in early 2024.
- CCID said top-tier large model iteration cycles shortened from several months to weeks or days.
- CCID said China built over 500 excellence-level smart factories across 31 provincial-level regions.
China's volume of inference data generated by artificial intelligence systems surpassed the volume of data used to train those systems for the first time, according to a report by the China Center for Information Industry Development (CCID), which operates under China's Ministry of Industry and Information Technology.
CCID released the report, on 10 major trends in innovation and development within new domains and new-quality areas, at the opening ceremony of the 2026 New Domain and New Quality Innovation Competition's finale in Qingdao, in east China's Shandong Province, on a Saturday. The body does not give the calendar date of that release.
The report said China's average daily token calls had exceeded 175 trillion as of June 2026, an increase of more than 1,000 times compared with early 2024. It said the iteration cycle for top-tier large models had shortened from several months to a few weeks or even days.
On manufacturing, CCID said China had established over 500 excellence-level smart factories covering more than 80 percent of major manufacturing sub-sectors across 31 provincial-level regions. Domestically developed computing power and large-scale AI models had been applied in more than 20 industries, including steel and automotive, in over 1,000 scenarios, it said.
Liu Wenqiang, vice president of CCID, told the ceremony that artificial intelligence is reshaping industrial systems and has become a general-purpose technology leading systemic changes across domains.