The National Data Administration recently officially designated the Chinese translation of Token as "Ciyuan", clarifying its role as a value anchor in the intelligent era and a "settlement unit" that connects technology supply and commercial demand. In recent days, the discussion on lexical elements has remained hot. What is the essence and value of lexical elements? What industrial impacts will the word element economy it generates bring? What are the risks and challenges that require high attention and systematic response? The exponential growth of call volume is the smallest basic unit for processing information in large models, where a single Chinese character, word, and punctuation can all be considered as one word element. The language that big models face is global, with Chinese, English, Arabic, code, mathematical formulas, and other content coexisting. "Zhang Ting, the product manager of Baidu Qianfan platform, believes that lexical elements are equivalent to a universal" greatest common divisor "that allows models to process all languages and symbols in a unified way, becoming the basic unit of measurement for artificial intelligence interaction. In daily use, the consumption of lexical elements can be seen everywhere. Having AI write an 800 word essay, including prompt words and complete output, would consume approximately 1000 to 1500 word elements; Having AI analyze a 10 page contract may consume 5000 to 10000 word elements. In recent years, artificial intelligence has accelerated its entry into daily life, and many people have become accustomed to using AI to write articles, search for information, and do translation. Currently, mainstream AI platforms provide free services to ordinary users, which can generally meet their daily work and life needs. But for enterprise users and developers, they need to pay fees based on the number of times the word element is called or the actual usage. At the beginning of this year, the AI intelligent agent tool "OpenClaw" quickly became popular, driving multiple domestic AI platforms to launch similar products one after another, and the number of keyword calls showed explosive growth. At a recent press conference of the State Council Information Office, Liu Liehong, Director of the National Data Administration, introduced that as of March this year, China's daily average use of keywords has exceeded 140 trillion yuan, an increase of more than 1000 times compared to 100 billion yuan at the beginning of 2024, and an increase of more than 40% in just three months compared to 100 trillion yuan at the end of 2025. The leapfrog growth in the use of lexical elements reflects that China's artificial intelligence industry has entered a period of rapid growth, with deepening application scenarios and significantly enhanced industrial competitiveness. In the view of Bai Runxuan, an analyst at the Artificial Intelligence and Big Data Research Center of CCID Consulting, the surge in the use of keyword tones is mainly driven by three aspects: firstly, the significant improvement of model capabilities, such as the bean bun video generation model Seedance 2.0, which consumes over 1 million keywords to generate a 1-minute video; The second is the upgrading of application forms, shifting from simple Q&A to the era of intelligent agents, where task planning, tool calling, and other contextual interactions significantly increase the consumption of lexical elements; The third is the acceleration of commercialization landing, with thousands of industries accelerating AI applications and driving exponential growth in keyword usage. The sharp increase in the number of word element calls combined with the rise in hardware costs and other factors has also pushed Alibaba Cloud, Baidu AI Cloud, Tencent Cloud and other cloud manufacturers to raise the prices of AI computing power and other products recently, with a general increase of about 30%. The demand for computing power in the market is strong, and the supply price naturally rises. "Pan Helin, a member of the Information and Communication Economy Expert Committee of the Ministry of Industry and Information Technology, believes that cloud vendors have improved their ability to price keywords by providing intelligent services, gradually getting rid of their dependence on free traffic models, and promoting the commercialization of artificial intelligence applications into a more sustainable new stage. The economic value is gradually highlighted by the popularity of AI intelligent agents, allowing more people to intuitively feel the economic value of lexical elements. If the traditional Internet is a traffic economy and attention economy, then the lexical economy in the AI era is a new model with productivity as the core and pay as you go billing. The consumption of keywords, unit output efficiency, and cost control ability are becoming key indicators for digital transformation and refined operation of enterprises. Traffic is gradually becoming ineffective, and lexical elements are the core of the future. ”Zhou Hongyi, founder of 360 Group, believes that the traffic economy of the traditional Internet is based on the scale effect of diminishing marginal cost: the capacity of optical fiber is almost unlimited, and the platform transmits data packets. The larger the number of users, the lower the unit cost. The cost is mainly related to the length of time and bandwidth occupation. But in the era of AI, the essence of AI is not data handling, but the deep coupling of computing power consumption, information processing, and "intellectual cost". The term element is the core unit of measurement for measuring AI's consumption of intelligence and computing power. The more it consumes, the higher the marginal cost, and the business logic will be completely changed accordingly. The shift of the big model industry from "traffic management" to "word element management" is a qualitative change in the underlying business logic. In the first two years, the 'Hundred Model Wars' focused on training and computing power competition, where everyone mainly used AI to chat, create, and verify model capabilities. Nowadays, big models have entered the stage of industrial implementation, and they need to enter factories and offices to become real 'digital employees' who work hard. ”Zhou Hongyi believes that the core scenario for large-scale consumption of computing power will be the comprehensive popularization of enterprise level intelligent agents. At present, the amount of lexical metaphone calls is mainly concentrated in the fields of the Internet, consumer electronics, finance, new retail and business services. The application scenarios are mainly unstructured information processing, education and teaching, content creation, search and recommendation. Bai Runxuan believes that in the future, scenarios such as software development, in-depth research, and personal intelligent assistants will become important growth points for word element calling. The term economy will reconstruct the value release path of data elements, reshape the AI industry business ecosystem, connect data elements with the real economy, and accelerate the intelligent upgrading of traditional industries such as manufacturing, agriculture, healthcare, and finance. According to the latest data from OpenRouter, a global API aggregation platform that balances efficiency and security, from April 6th to 12th, the total call volume of AI models worldwide was 21 trillion tokens. Among the top five global call volume rankings, three Chinese AI models were included, with Alibaba Qwen3.6-Plus ranking first with a weekly call volume of 1.66 trillion tokens. It is worth noting that the weekly usage of AI big models in China has surpassed that of the United States for six consecutive weeks. Chinese AI has truly been put into use. The call volume reflects the actual usage scale, rather than the high or low model parameters or the strength of the score. ”Zhou Hongyi stated that global developers tend to choose Chinese large models, mainly because the gap in model capabilities between China and the United States has narrowed to less than three months, and the price of Chinese models is only one tenth or even lower than that of top overseas models. Behind the low-cost advantage is the rapid maturity of the open source ecosystem, which has led to the formation of a universal development pattern similar to the Android system in China's big model. Currently, global AI competition has shifted from model parameter competition to application landing capability competition. China has achieved a leapfrog development in the field of AI by relying on its super large market, improving its power computing infrastructure, and advantages in high cost-effective industries. But it should also be noted that the demand for computing power is showing explosive growth, but the supply has not fully caught up, forming a temporary supply-demand imbalance. One is the insufficient supply of high-end chips, the tight production capacity of overseas high-end GPUs, and the extended delivery cycle. Secondly, there is a lack of data quality and standardization, and the construction of high-quality datasets still needs to be strengthened. ”Bai Runxuan said. What needs to be more vigilant is that both word element transactions and AI service billing are completed online, with deep binding of data, permissions, and fund links, resulting in a significant increase in security risks. Qi Xiangdong, Chairman of Qianxin, believes that criminals can steal keywords through vulnerability exploitation, terminal hijacking, malicious plugins, prompt word injection, and other means, and then illegally access sensitive data, call AI services beyond their authority, steal billing quotas, and even engage in identity forgery, network fraud, and other behaviors, causing chain risks such as data leakage, loss of control of permissions, and property damage. "For ordinary users, when using AI applications, they do not enter sensitive information such as ID number, bank account, home address, face photos, trade secrets, etc. When using AI in the workplace, it is important to conceal or blur real information, and try to use locally deployed or enterprise level AI services that have been certified for security. ”Qi Xiangdong suggested establishing a risk classification and management system for word element security, as well as a safety standard specification for the entire word element process, clearly dividing the main responsibilities of each link. With the acceleration of artificial intelligence empowering thousands of industries, industry insiders generally believe that the next step should focus on key links such as high-quality data supply and standardization of word element measurement and pricing, accelerate the promotion of computing power as a convenient and accessible public basic resource like water and electricity, and develop it in a standardized and regulated manner, so that the word element economy can truly become an important force driving the deep implementation of the "AI+" action and promoting the high-quality development of the intelligent economy. (New Society)