Recently, a set of data on token usage has continued to flood the technology circle, triggering widespread attention and discussion from all walks of life. At the 2026 annual meeting of the China Development Forum held on March 23, it was revealed that the daily average token usage in China exceeded 14 trillion yuan in March this year, an increase of over a thousand times compared to 100 billion yuan at the beginning of 2024. In the same month, the token call volume of China's AI Big Model Week surpassed that of the United States for three consecutive weeks, becoming one of the countries with the highest AI application activity in the world. According to the latest data from OpenRouter, the world's largest AI model API aggregation platform, from March 16th to 22nd, the total number of token calls for AI models worldwide was 20.4 trillion, with China alone reaching 7.359 trillion, accounting for 36% of the global total. Token, as the smallest unit of measurement for AI models to process information, is a key indicator for measuring the activity and industrial value of AI models. The explosive growth of China's token usage directly reflects the development speed of China's AI industry and also reflects the profound changes in the global AI application pattern. China's leading achievement is not accidental, but the result of multiple factors working together. Policy guidance continues to increase. In August 2025, the State Council issued the "Opinions on Deepening the Implementation of the 'Artificial Intelligence+' Action, 'which clearly set three phased goals for 2027, 2030, and 2035. In January 2026, the Ministry of Industry and Information Technology and eight other departments jointly issued the Implementation Opinions of the "Artificial Intelligence+Manufacturing" Special Action, which clearly stated the core goal of launching 1000 high-level industrial intelligent agents by 2027. Following that, at the National People's Congress and Chinese People's Political Consultative Conference in March, the government work report proposed for the first time to "create a new form of intelligent economy" and explicitly stated to "promote the accelerated promotion of the new generation of intelligent terminals and agents". A series of policies are closely linked and progressively implemented, providing clear and sustained institutional guarantees for technological innovation, computing power construction, and scenario implementation. Continuous breakthroughs in technological innovation. Through optimizing the underlying architecture, the generation gap between domestic large models and top global technologies is gradually narrowing, and the inference cost and response speed are constantly improving, enabling them to complete the same complex tasks with fewer tokens. A series of domestic models represented by DeepSeek and MiniMax M2.5 not only reduce the cost of API usage to a low level, but also continue to approach and even surpass overseas closed source models in some segmented scenarios in terms of performance. This dual advantage of "high performance and low cost" directly reduces the entry threshold for developers and enterprises, providing the core driving force for the exponential growth of Token call volume. The computing power facilities are constantly improving. Computing power is the fundamental resource that supports the training and inference of large models. With the large-scale application of domestic chips in some scenarios and the continuous increase in the proportion of intelligent computing power, CCID Research Institute predicts that the proportion of intelligent computing power in China will exceed 35% by 2026, and the Chinese AI industry is getting rid of its dependence on a single computing power path. A more autonomous, diverse, and cost controllable computing power supply system provides a physical foundation for large-scale and high-frequency token calls, providing stable underlying support for daily call volumes exceeding 140 trillion tokens. Accelerated explosion of application scenarios. The significant increase in token usage is essentially a reflection of the deep integration of AI applications into socio-economic activities. On the one hand, tools represented by open-source AI agent OpenClaw can consume 10 to 100 times more tokens per interaction than traditional AI. Liu Xingliang, president of DCCI Internet Research Institute, said that the core value of such tools is to promote AI from "chatting" to "execution", so that AI can truly become a "digital employee". In essence, the enthusiasm of Chinese users for such "capable" AI tools is an urgent need for personal productivity upgrading. On the other hand, emerging scenarios such as video generation have become important growth poles for token consumption. Pan Helin, a member of the Information and Communication Economy Expert Committee of the Ministry of Industry and Information Technology, pointed out that after the launch of models such as Seedance 2.0 of DreamAI, it quickly sparked a creative boom on short video platforms, and social communication and content production mutually reinforced each other, continuously pushing up the scale of calls. From intelligent assistants to multimodal generation, the large-scale implementation of diverse scenarios constitutes a powerful engine for token consumption. Diversified synergy of industrial ecology. Unlike the single point breakthrough of the US AI industry relying on a few top companies, China has built a multi-level and widely participated ecological pattern of "open source community+technology enterprises+massive creators". Open source projects such as Open Source HarmonyOS and Open Source Euler continue to make breakthroughs in the field of basic software, demonstrating vibrant community vitality; ByteDance, Baidu, Alibaba, Tencent and other technology enterprises continued to increase their R&D efforts to promote the iterative upgrading of domestic large models and the launch of products; Massive individual creators and small and medium-sized enterprises rely on basic models to quickly transform AI into actual productivity. A closed-loop system of "research and development application feedback iteration" is formed among the three, and AI capabilities are widely implemented through platforms such as WeChat, DingTalk, and Feishu that reach billions of users. Ultimately, a virtuous cycle of "technology iteration cost optimization application popularization" is achieved, driving China's AI industry from quantity accumulation to systematic leadership in ecological collaboration. Standing at the beginning of the 15th Five Year Plan, looking ahead, the prospects are even more magnificent. The outline of the 15th Five Year Plan clearly proposes the comprehensive implementation of the "Artificial Intelligence+" action, empowering thousands of industries in all aspects. It is expected that by the end of the 15th Five Year Plan period, the scale of China's artificial intelligence related industries can exceed 10 trillion yuan and move towards a broader growth space. JPMorgan predicts that the consumption of AI inference tokens in China will increase from approximately 10 trillion in 2025 to approximately 390 trillion in 2030, a growth of approximately 370 times over five years. While there is enormous potential, challenges cannot be ignored: how to achieve quality improvement on the basis of quantitative growth, how to solve copyright issues in AI generated content, and how to build an integrated security governance system covering data, technology, networks, etc., are all issues that China's AI industry needs to continue to address. From the open source craze of "raising shrimp" among the public to the creative burst of AI videos on social media, from the two-fold increase in token usage in two years to the increasingly perfect industrial ecology, the development of artificial intelligence in China has already surpassed the stage of simply catching up with technology and moved towards a systematic breakthrough stage based on the local market and relying on a complete industrial system. The daily average usage of over 140 trillion tokens is not only a key turning point for artificial intelligence from technological accumulation to application explosion, but also a microcosm of the resonance between open source innovation and scenario implementation, as well as the deep integration of data elements and AI technology, becoming a clear indicator of the high-quality development of China's digital economy. We are experiencing and witnessing the 'Chinese moment' of artificial intelligence, and all of this is just the beginning. (New Society)