The concept of "token" continues to ferment, and the A-share computing power sector is strengthening. On March 25th, several stocks including Shenzhen Tefa Information Co., Ltd., 263 Network Communication Co., Ltd., and Shanghai Tianji Technology Co., Ltd. rose by the daily limit, while several stocks including Beijing Halo New Network Technology Co., Ltd. and Guangdong Aofei Data Technology Co., Ltd. saw the highest gains. Lexicon is the smallest unit of information processing in large models, with three core attributes of measurability, pricing, and trading. It is the basic unit for AI service billing and settlement. If the big model is an intelligent machine, the word element is the "electricity per kilowatt hour" consumed by the machine during operation, and it is also the basic unit of measurement for "billing by kilowatt hour" when the machine provides services to the outside world. According to data from the National Data Bureau, the daily average usage of keywords and tones in China has shown explosive growth, with an average of 100 billion yuan per day at the beginning of 2024, soaring to 100 trillion yuan per day by the end of 2025, and exceeding 140 trillion yuan per day by March 2026, a growth of over a thousand times in two years. The 'lexical economy' is expected to reshape the business logic of AI, shifting large model enterprises from 'selling power' to 'selling quantity'. ”Yuan Shuai, Deputy Secretary General of Zhongguancun Internet of Things Industry Alliance, told reporters that under this model, enterprise revenue no longer depends on long-term project delivery, but on real-time concurrent interaction scale. This transformation can drive the AI industry to shift from pure technology driven to refined operation driven, enabling high-frequency and high viscosity application scenarios to quickly transform into explosive performance growth. Faced with the potential business changes brought about by the "lexical economy", global tech giants are accelerating their strategic layout. NVIDIA founder and CEO Huang Renxun proposed "Token Economics" at the 2026 GTC conference, believing that the future data center will no longer be a warehouse for storing files, but a "factory" for producing lexical elements. At the same time, Nvidia also released a full stack hardware system to reduce the production cost of lexical elements. On March 16th, Alibaba Group Holdings Limited officially established the Alibaba Token Hub (ATH) business group, with the core goal of "creating tokens, delivering tokens, and applying tokens". The group is directly managed by CEO Wu Yongming. The token economy will bring huge development opportunities. ”Li Peng, Senior Vice President of Huawei Technologies Co., Ltd., predicts that by 2030, the scale of China's AI related industry will exceed 10 trillion yuan. Around the production, invocation, distribution, and settlement of lexical elements, domestic listed companies are accelerating the layout of the entire industry chain with upstream computing infrastructure as the core, midstream platform services as the link, and downstream applications as the lever, resulting in a continuous increase in industrial concentration. The upstream computing infrastructure is the core foundation of lexical element production and has become the most directly beneficial link. For example, Xiechuang Data Technology Co., Ltd. (hereinafter referred to as "Xiechuang Data") achieved a significant year-on-year increase in both revenue and net profit in 2025. The company plans to invest 21.2 billion yuan last year to purchase high computing power servers and deploy large-scale computing power cluster construction. Geng Kangming, Chairman of Xiechuang Data, told reporters that current AI industry giants are generally optimistic about the sustained growth of computing power. OpenAI、Anthropic、 Microsoft, Google, and others have successively laid out intelligent agents, with reasoning models represented by DeepSeek pushing agents into the landing stage. The midstream platform is a hub for scheduling and settlement, responsible for key functions such as keyword distribution, billing, and computing power scheduling. Taking Zhongbei Communication Group Co., Ltd. as an example, it has the advantage of large-scale operation and currently operates a computing power scale of over 17000P, providing intelligent computing services to clients such as Kingsoft Cloud Holdings Co., Ltd. In the downstream application field, listed companies rely on industry advantages to promote the integration of lexical element technology and vertical scenarios, accelerate commercial monetization, and mainstream AI application manufacturers such as large models and robots belong to this field. For example, iFlytek Co., Ltd. relies on the Spark model to achieve large-scale monetization in education, healthcare, office and other scenarios; Shenzhen Youbisxuan Technology Co., Ltd. relies on large models to empower robots to support the implementation of intelligent interaction and embodied intelligence. (New Society)