The global AI large model token consumption has recently shown fluctuations, and the market landscape has undergone significant shifts.

According to the latest industry monitoring data, after ten consecutive weeks of growth, the total token usage of global AI large models has declined for two consecutive weeks. During the period from April 13 to 19, the total token usage remained at 20.6 trillion tokens. Notably, the weekly token usage of Chinese AI large models, which had previously performed strongly, dropped by 23.77%环比, reaching 4.44 trillion tokens.

In sharp contrast, the U.S. AI large model market experienced a 20.62% increase in token usage during the same period, reaching 4.91 trillion tokens. This is the first time in the past two months that U.S.-based large models have surpassed their counterparts in activity level.

Industry experts point out that this "one down, one up" reversal is closely related to the global surge in computing power costs. Since March this year, major cloud service providers in China have adjusted their pricing strategies, with internet giants such as Alibaba Cloud, Tencent, and Baidu successively increasing the fees for large model-related services. Meanwhile, Zhipu, a domestic AI unicorn, has also completed its third price adjustment of the year. In overseas markets, AI giant Anthropic has also updated the pricing of its models.

Although the token usage has temporarily declined due to price fluctuations and market adjustments, the industry remains optimistic. Most analysts believe that with the further deepening of AI application scenarios, the total global token consumption will continue to grow in the long term, and the current fluctuations are seen as a market self-regulation under the backdrop of reshaping computing power costs.