China Positions the AI Token as Central Metric for Data Factor Economy

China has elevated the AI 'token'—the fundamental building block of large language model processing—to center stage at its premier national big data forum. Opening on August 28 in Guiyang, the 2026 China International Big Data Industry Expo centered on tokens as a practical mechanism to monetize data assets and measure computing output. Official data highlighted China's daily token consumption exceeding 500 trillion, propelled by a surge in autonomous AI agents and multimodal content, signaling a strategic transition from model training to large-scale inference.

This shift marks a turning point in China's AI ecosystem from foundational pre-training toward practical inference and operational deployment. Elevating the token to a standardized measure across computing power, algorithmic output, and data resources lays the groundwork for uniform commercial pricing, value settlement, and marketized data-factor trading.

The token—the foundational unit of text and data processed by artificial intelligence models—has moved from an engineering metric to the forefront of China's national digital strategy. At the 2026 China International Big Data Industry Expo held in Guiyang on August 28, the concept was established as the event's core theme, framed as an emerging pathway to unlock the commercial value of data factors. The expo drew 372 domestic and foreign enterprises and more than 16,000 attendees to Guiyang, often dubbed 'China's Big Data Valley.' According to People's Daily, enterprise-led or enterprise-participating events accounted for more than 60 percent of the agenda, reflecting strong corporate engagement across cloud computing, infrastructure, and model development. Speaking at the expo, National Data Bureau Director Liu Liehong conceptualized the token as an explicit vehicle of data value, a metric of computing efficiency, and a standardized measure of AI capability. Expanding on remarks delivered earlier this year at the China Development Forum, Liu characterized tokens as a commercial settlement unit linking technical capacity to market demand—analogizing raw data to input materials, computing clusters to factories, and tokens to finished industrial output. The emphasis on tokens follows a rapid acceleration in domestic AI utilization. According to an August 29 report by state broadcaster CCTV cited by People's Daily, China's daily token consumption has surpassed 500 trillion. Data from global model aggregation platforms such as OpenRouter cited in the report also pointed to strong international usage, with domestic models like DeepSeek and Alibaba Cloud's Tongyi Qianwen leading global platform invocation volume for 15 consecutive weeks. Industry participants detailed how automated workflows are driving this consumption curve. Alibaba Cloud Vice President An Xiaopeng noted that in February, automated token calls initiated by AI agents surpassed manual human prompts for the first time, reaching five times the human volume by August. Academic and think-tank experts also pointed to generative entertainment as an amplifier; Development Research Center specialist Ma Jun noted domestic micro-drama production surged from 20,000 episodes last year to 800,000 in the first half of this year, each demanding intensive multimodal processing. Addressing the cost burdens of massive inference, Dai Guohao, co-founder of AI infrastructure startup Infinence (Wuwen Xinqung), showcased heterogeneous inference across distributed clusters that reduced operating costs by an order of magnitude. Meanwhile, Liu Yi, vice president of iSoftStone, likened token infrastructure to power plants requiring 99.9 percent service reliability, underscoring the shift toward industrializing AI delivery across China's economy.