China's AI Strategy: Mass Adoption Through Affordable, Open-Weight Models

Are the US’ AI models better than China’s? That may be beside the point

Rather than fight US frontier AI control, China is thinning the margins of premium AI with cheap-enough, good-enough models for mass adoption

3 -MIN READ3 -MIN

Jeffrey Wu is a director at MindWorks Capital, a leading Hong Kong-headquartered venture capital firm specialising in technology investment across Greater China and Southeast Asia.

Published: 4:30pm, 8 Jul 2026

As another heatwave rolled across Europe, the warehouses emptied before the politics could catch up. Air conditioners and fans sold out across Spain, Italy and Germany, most of them Chinese. Fan sales in Spain alone on the retail platform of Alibaba Group Holding (which owns the South China Morning Post) nearly doubled last month while Midea Group’s air-conditioner sales in western Europe surged by over 70 per cent in the first six months of the year.

A decade from now, such heatwaves may become ordinary. Air-conditioning units in the European Union are expected to more than double from 2019 levels to 275 million by 2050. The question will not be whether Europeans approve of air conditioning or whether their governments are comfortable depending on Chinese supply chains. It will be who can deliver affordable units fast enough when demand becomes desperate.

Increasingly, the answer is China – because its air conditioners are available, affordable and good enough exactly when it matters. The same logic may be reshaping something far less visible: artificial intelligence.

At first glance, Chinese AI looks like an extension of China’s manufacturing playbook: take a complex technology, compress the cost, iterate quickly, accept thinner margins and make it cheap enough for mass adoption. China did this with solar panels, batteries and electric vehicles. Now, through open-weight models, it appears to be doing something similar with AI.

The analogy works until it reaches the strategic bottleneck. In manufacturing, China’s advantage became durable because the supply chain stayed in China. Once built, it was brutally hard to relocate. Open-weight AI moves differently.

Once released, a model can live on someone else’s graphics processing units (GPUs), be fine-tuned with someone else’s data and create value inside someone else’s systems. The feedback loop shifts too: in manufacturing, the learning compounds inside the Chinese system. In open-weight AI, much of it accrues to whoever adapts the model.