Tokens now have a Chinese name, "word units," and the concept continues to be popular. Tianji Technology hits the 20cm daily limit.

（Source: Caixin）

          AI is shifting from “storytelling” to “getting deployments across the finish line,” and the surge in demand directly forces upstream compute supply to speed up capacity expansion, creating strong market pull.

On March 25, the “token” concept continued to surge, with Tianji Technology (300245.SZ) hitting the 20cm daily limit. Previously, Aured (600666.SH) and 263 (002467.SZ) also hit the daily limit; Huanhuan New Network (300383.SZ), Tongniu Information (300895.SZ), Aofei Data (300738.SZ), Mili Cloud (000815.SZ), Wangsu Technology (300017.SZ), uknowme etc. all rushed sharply higher. In Hong Kong stocks, the “token leading stock” xunce (03317.HK) once rose more than 10%.

In terms of news, earlier, Liu Liehong, director of the National Data Administration, said that the official “token” is identified as a value anchor and the connecting technology for the intelligent era—“the settlement unit” between supply of technical services and commercial demand—providing a quantifiable foundation for implementing business models. From the official perspective, Token has been given a translation: “token.” Data show that by the end of 2025, China had built more than 100,000 high-quality datasets; by this March, the average daily Token (token) usage in China exceeded 140 trillion, up more than 40% from the end of 2025, and the increase versus the 100 billion at the beginning of 2024 was over 1,000 times.

OpenRouter, the latest release from the world’s largest AI model API aggregation platform, shows that as of March 15, the number of daily calls to China’s AI large models reached 4.69 trillion tokens, surpassing the United States for the second consecutive week. The top three positions in global usage are even entirely taken by Chinese models. The domestic large model MiniMax M2.5 has dominated the global large-model call volume rankings for five consecutive weeks.

JPMorgan Chase predicts that China’s AI inference token consumption will grow from about 10 trillion in 2025 to about 390 trillion in 2030—about a 370x increase over five years. When token consumption continues to climb, it means that downstream application-side willingness and ability to pay have significantly strengthened, and AI is moving from “storytelling” to “getting deployments across the finish line.” The demand surge directly forces upstream compute supply to accelerate capacity expansion, creating strong market pull.

Endless news, precise interpretation—now in the Sina Finance app