I’ve recently noticed the rapid expansion of China’s AI industry. The head of Huawei’s ICT division mentioned at the recent Mobile World Congress that, over the past two years, the global token consumption has increased rapidly by approximately 300 times. The main reason is that AI applications such as text-to-video conversion and smart shopping are being put into practical use one after another.

At present, more than 30 million AI agents are working together across the world, and the number of API calls to large language models developed in China is also growing rapidly. Supporting this trend is China’s overwhelming power-supply capability. In 2024, China’s electricity generation reached about 10 trillion kilowatt-hours, surpassing the combined total of the US and EU—an astonishing figure.

What’s interesting is China’s strategy. Even amid US tariff pressure, China continues to supply the global market through low-cost token production. Seeing products such as Zhipu’s GLM coding model and Kimi K2.5 rapidly gaining popularity makes it clear that this strategy is actually working.

It reminds me of past global expansion efforts in low-wage labor and the textile industry. But this time, low-cost tokens are flowing into Silicon Valley’s application layer, effectively subsidizing development there in practice. I can feel the overall landscape of the AI competition in the market is about to change significantly.