China AI Price War Shifts to Tiered Task Pricing

China's artificial intelligence price war is evolving from blanket token price cuts toward tiered pricing based on the cost of completing specific tasks. Bank of America reports that basic inference has become commoditized, forcing companies to compete on efficiency and the ability of AI systems to finish tasks rather than simple token processing.

Premium pricing remains viable for models that offer stronger reasoning, multimodal capabilities, longer context windows, and agentic workflows. This shift follows a period of aggressive price reductions triggered by low-cost models from developers like DeepSeek.

Analysts suggest that long-term profitability will likely concentrate in AI chips, semiconductor equipment, memory, and large cloud platforms rather than standalone model developers. This trend is partly driven by US-led export controls, which accelerated the development of domestic Chinese AI technology and expanded the local supplier market.