CCID report: China's AI inference data volume overtakes training for the first time

On October 31, 2026, the China Center for Information Industry Development (CCID) released a report titled 'Ten Major Trends of Innovation-Driven Development in New Domains and New Quality,' stating that China's AI inference data volume has surpassed training data volume for the first time, marking the country's shift from the model training and research phase into large-scale industrial application. IT Home covered the story on November 1, citing CCTV News.

According to the report, by June 2026, China's average daily token consumption exceeded 175 trillion, an increase of more than 1,000 times compared to early 2024. The iteration cycle of top-tier large models has also been shortened from several months to weeks or even days. The report further notes that as the 'AI + Manufacturing' initiative is implemented in depth, China has built over 500 outstanding-level smart factories, covering 31 provincial-level regions and spanning over 80% of major manufacturing categories. Domestic computing power and large models have been deployed in more than 1,000 application scenarios across 20 industries, including steel and automobiles, driving the expansion of new models such as personalized customization and full-lifecycle services.

36Kr reported on November 2 that the report was released during the 2026 New Domain and New Quality Innovation Competition, repeating the core figures. No other institutions have yet verified or updated these data, and the event currently remains at the report release stage.