China's AI cost war is moving beyond covering token cost cuts as companies increasingly vie on the disbursal of getting synthetic intellect systems to complete tasks, according to a Bank of America study published Thursday.
The change follows assured cost cuts following low-cost Chinese AI models specified as DeepSeek intensified competition throughout the industry.
The economics are now changing, BofA said.
"China's LLM cost war is evolving from covering cuts to tiered pricing," wrote BofA analysts led by Winnie Wu, the bank's caput of Asia equity scheme and co-head of China equity research.
Instead, companies are increasingly competing on how efficiently their AI systems complete tasks fairly than merely how cheaply they procedure tokens.
Models alongside stronger reasoning, longer environment windows, multimodal capabilities, and agentic workflows can motionless command premium prices, according to the BofA analysts.
"Basic conclusion is increasingly commoditized, during frontier capabilities remain differentiated," the analysts wrote.
"The key financial metric is shifting from cost per token to disbursal per completed task," they added.
The change additionally comes as years of US-led export controls have accelerated China's shove for family AI technology. The restrictions expanded the market for local suppliers during intensifying competition as additional companies entered the sector.
The changing economics could additionally reshape anywhere profits are made throughout China's AI industry, according to the bank.
The strongest long-term economics are apt to remain alongside AI chips, semiconductor equipment, memory, and ample haze platforms, fairly than standalone AI example developers, its analysts wrote.
The trend isn't distinctive to China. Businesses are increasingly choosing distinct AI models for distinct tasks according to disbursal and performance, alternatively of relying on a sole model, Business Insider reported in July.
