Data shows that in March of this year, the daily average number of token calls in China has exceeded 140 trillion, an increase of over a thousand times in two years. The sharp increase in the use of word elements not only confirms the deepening of artificial intelligence application scenarios, but also gives rise to new opportunities - related enterprises are accelerating their layout around the "word element economy" industry chain, such as high-performance computing power supply, word element management services, and high-quality datasets. The so-called morpheme is the smallest unit of information processed by a large model. For example, every time an AI is asked a question or used to generate content in daily work and life, it is called a morpheme. Especially since the beginning of this year, the explosion of intelligent agents such as "lobster" has significantly increased the consumption of verb elements. Since the end of January this year, some model companies have set a performance record of 20 days of revenue exceeding the annual revenue of 2025. Behind this set of numbers is a new business logic based on word element billing that is rapidly evolving. ”Liu Liehong, Director of the National Data Administration, recently stated at the 2026 Annual Meeting of the China Development Forum that a new value system is rapidly evolving around the use, distribution, and settlement of keywords, and has become an important path for the potential monetization of the artificial intelligence industry. The sharp increase in the use of lexical elements is due to the consumption of computing power, the operation of algorithms, and the investment in electricity, which also puts higher demands on computing power. Recently, many enterprises are accelerating their innovative layout of high-performance computing power, among which architecture innovation represented by supernodes is an important path to improve computing efficiency. For example, Zhongke Shuguang recently launched its first wireless cable box type super node, scaleX40. In typical inference scenarios, the inference throughput performance of the same card size is improved by more than 4 times, which can significantly enhance the word element output capacity per unit of computing power; ZTE Corporation has launched supernode technology, which integrates dozens to hundreds of multi manufacturer GPU logic into a unified computing unit by restructuring the computing power interconnection system, achieving system level optimization of computing power. In the eyes of industry insiders, the core innovation of the computing power industry will revolve around the reduction of effective keyword costs, with a focus on deep collaboration between computing power and applications, full stack optimization of software and hardware, and ecological collaboration across the entire industry chain. Tu Jiashun, Chief Expert of Strategy and Ecology at ZTE, said that the implementation and promotion of supernode technology will promote the evolution of intelligent computing infrastructure towards higher efficiency, greener, and more open directions. The surge in lexical elements has put forward higher requirements for computing density, memory access bandwidth, communication efficiency, etc., which has prompted computing power to shift from pursuing single card peak to systematic 'memory bandwidth interconnection' collaborative optimization. ”Lu Feng, the director of Beijing Frontier Future Technology Industry Development Research Institute, told reporters that the "word element economy" will drive the development of underlying computing infrastructure towards higher energy efficiency. Given the characteristics of measurability, pricing, and trading, the term element is also a "settlement unit" that connects technology supply and commercial demand. Recently, the three major operators have expressed their intention to further explore "token management services". In the era of intelligence, to create a new form of intelligent economy, it is necessary to accelerate the transformation and upgrading from 'traffic management' to 'token management'. ”Liu Guiqing, General Manager of China Telecom, introduced that China Telecom's word element management has already had preliminary practice. Taking a certain enterprise's AI privatization deployment on the China Telecom Information Infrastructure platform as an example, 73 intelligent agents were customized and developed, driving an annual consumption of 1.2 trillion word elements. In addition, China Unicom will accelerate the construction of a computing power business model of "intelligent agent+Token+AI cloud"; China Mobile stated that by integrating high-quality AI models and creating trustworthy inference services, it has connected the complete service chain of "intelligent agents using tokens and tokens to drive computing power", and the keyword market is rapidly opening up. The significant increase in the daily usage of lexical elements signifies a substantial increase in the supply of datasets, and the empowerment of data elements in the innovation and development of artificial intelligence has entered a stage of positive interaction. The reporter learned from the National Data Administration that by the end of 2025, there will be over 100000 high-quality datasets built nationwide, with a total volume of over 890PB (petabytes, computer storage capacity units), equivalent to about 310 times the total digital resources of the National Library of China. The data is decomposed into lexical elements and processed and applied by AI, providing solid support for model iteration and application implementation. ”Lu Feng stated that as word elements become standardized pricing and circulation units for AI capabilities, the value of the industrial chain is also redistributed to each optimization node in the "word element lifecycle". In the data service sector, high-quality word element supply will become a premium resource and is expected to drive the development of independent commercial products such as prompt word engineering, word element compression, and vertical word element libraries. It is reported that in the next step, the National Data Administration will collaborate with all parties to implement a new round of action plan for high-quality dataset construction. With scenario requirements as the driving force, it will create AI Ready high-quality datasets that are technically feasible, practical, convenient, and have quality assurance, achieving a quantitative and qualitative improvement in the supply of high-quality datasets. (New Society)