However, the same restrictions have pushed Chinese companies to another play: Compute consolidation, efficiency optimization, and open weight model release. For example, only 2.6 million graphics hours were spent on the DeepSeek-V3 training run, which is much lower than its American counterparts. But Alibaba's Qwen models are now among the most downloaded open weight models in the world, and companies like Zhipu and MiniMax are creating competitive multimodal and video models.
China's industrial policies mean that new models can move quickly from the laboratory to implementation. Local governments and large enterprises are already implementing reasoning models in management, logistics and finance.
Education is another advantage. Largest Chinese Universities implement literacy programs in the field of artificial intelligence into their curricula, introducing skills before the labor market demands them. The Ministry of Education has also announced plans to integrate AI learning for children of all school ages. I'm not sure the phrase “engineering state” fully captures China's relationship with new technologies, but decades of infrastructure building and top-down coordination have made the system unusually effective at promoting large-scale adoption, often with far less social resistance than you might see elsewhere. Large-scale use naturally allows for faster iterative improvements.
Meanwhile, Stanford HAI AI Index 2025 found that Chinese respondents are the most optimistic in the world about the future of artificial intelligence – much more optimistic than the US or UK population. This is astonishing given that China's economy has slowed since the pandemic for the first time in more than two decades. Many in government and industry now view AI as a much-needed spark. Optimism can be a powerful fuel, but whether it can sustain itself in the face of slowing economic growth is still an open question.
Social control remains part of the picture, but a different kind of ambition is taking shape. This new generation of Chinese AI founders are the most globally minded I've ever seen, moving seamlessly between hackathons in Silicon Valley and pitch meetings in Dubai. Many are fluent in English and understand the rhythms of global venture capital. Having watched the last generation struggle with the burden of a Chinese label, they are now creating companies that are multinational from the start.
The US may still lead the way in speed and experimentation, but China can influence how AI becomes part of everyday life both at home and abroad. Speed matters, but speed is not the same as excellence.
John Thornhill answers:
You're right Caiwei, speed is not the same as superiority (and “killing” is perhaps too strong a word). And you are also right to highlight China's strength in open weight models and the US preference for proprietary models. This is not just a struggle between the economic models of two different countries, but also between two different ways of applying technology.
