I just saw some pretty impressive numbers about global token consumption. In the last two years, daily consumption has multiplied almost 300 times. It's a scale that's hard to visualize, but think about what it means: AI applications generating video content, intelligent purchasing systems running nonstop, all of that requires massive processing.

The interesting part is who is leading this. According to data from Huawei presented at the Mobile World Congress, more than 30 million AI agents are operating collaboratively worldwide. But here’s the key point: the large models developed in China are experiencing rapid growth in global API calls. Projects like Zhipu’s GLM Coding Plan and Kimi K2.5 are gaining traction quickly.

The reason is simple: low-cost tokens. While other markets are in a slump seeking efficiency, China has chosen a different strategy: producing at a massive scale with reduced costs. It’s basically the same model that worked with manufacturing and textiles years ago, but now applied to the AI application layer.

And here’s the data that explains it all: China’s electricity generation in 2024 reached approximately 10 trillion kilowatt-hours. That surpasses the total combined output of the United States and the European Union. It’s no coincidence. All this growth in AI processing requires brutal energy infrastructure, and China has it.

Despite U.S. tariff blockades, this low-cost production continues to supply global markets. It’s like an indirect subsidy for the entire layer of applications that Silicon Valley is building. The reality is that the global AI economic model is being shaped by China’s ability to produce cheap tokens at scale. That’s what’s really driving this 300-fold growth.