China’s daily token consumption has increased more than 300-fold over the past 18 months. This is due to the rapid growth in AI applications in the country, said a senior official on Thursday. Liu Liehong said at a State Council Press Conference that the daily token usage had risen from 100 billion to 30 trillion by June. Liu also emphasized progress in building high quality datasets. He noted that China had created over 35,000 datasets with a total of 400 petabytes by June. AI model training has sparked a demand for data trading. The cumulative transactions have reached nearly 4 billion yuan (552 million dollars). Beijing Data Exchange accounts for almost 80% of these trades, up 10% from last year. Liu said that Chinese-language data played a crucial role in model development. Most domestic AI models are now trained on datasets with more than 60% Chinese, and in some cases up to 80%.[
