China could make AI cheaper.
That does not mean Western investors win.

Verified evidence indicates China is expanding domestic AI capacity across models, cloud, chips and data centres. Alibaba’s AI cloud revenue growth and funding raise show commercial and capital commitment; SMIC and several domestic accelerator developers show demand-led growth. Chinese open-weight model releases reinforce price-performance competition, although direct like-for-like frontier benchmarks and sustainable profitability remain less clear.

Inference: cheaper capable Chinese models could lower inference costs and accelerate global usage, increasing compute demand and adoption. They could also compress proprietary-model pricing and valuations, reduce the share of rents captured by US model providers, and redirect infrastructure profits toward Chinese clouds, chips and applications. This is not automatically bearish because lower prices can enlarge the addressable market.

• Whether Alibaba’s new equity capital translates into profitable cloud and model deployment rather than another capex-and-margin trade-off.
• SMIC capacity expansion and whether domestic accelerators can improve performance, software compatibility and supply reliability under export controls.
• Adoption and pricing of Qwen and other Chinese open-weight models outside China.
• Evidence on Chinese AI-chip makers’ cash conversion, receivables and profitability.