Explosive Token Growth Signals New Stage for China’s AI

China’s data resources are powering artificial intelligence at unprecedented scale, according to the National Data Resources Survey Report (2025) released at the 9th Digital China Construction Summit on April 29. The report revealed that national daily token calls surged from just over one trillion at the start of 2025 to 100 trillion by year‑end, with cumulative annual calls reaching about 21,100 trillion. This exponential growth has become a vivid symbol of the accelerated pace of Digital China.

Data and computing power form the backbone of this transformation. In 2025, China built more than 110,000 high‑quality datasets totaling over 908 PB, spanning healthcare, industry and education. Data annotation exceeded 85 PB, generating RMB 18.3 billion in output. Intelligent computing capacity reached 1.59 million PFlops, with eight hub regions contributing over 80% of new national computing power. A national integrated computing network pilot platform was launched, covering more than 80% of provinces and municipalities.

AI applications are rapidly permeating industries. The report noted that the added value of core digital‑economy industries exceeded 10.5% of GDP, while the AI core industry surpassed RMB 1.2 trillion. AI is accelerating the development of digital economy, governance, culture, society and ecological civilization. In digital society, the national smart education platform reached 178 million users across more than 200 countries and regions. Internet healthcare users totaled 410 million, representing 36.5% of netizens. Medical insurance imaging cloud data reached 274 million entries, and cross‑province retrieval was launched, embedding digital technology into more public‑service scenarios.

Industrial capacity is also being reshaped. In 2025, China strengthened investment and financing services, launching the National Venture Capital Guidance Fund to support integrated circuits and AI. Fixed‑asset investment in high‑tech services rose 3.5% year‑on‑year, with information services investment up 28.4%. The report highlighted that total data used for AI training and inference reached 199.48 EB, up 42.86% year‑on‑year. For the first time, inference data volume exceeded training data at 101.34 EB, marking a new stage where training and inference are equally critical and application‑driven iteration defines progress.