US tech companies are turning to Chinese AI models
According to a MarsBit report, a number of US technology companies are shifting their AI deployment strategies because of cost considerations. The companies mentioned include Coinbase, Lindy, and Snowflake, while the Chinese models cited include GLM-5.2 and Kimi. The core takeaway is that enterprise buyers are placing greater emphasis on cost efficiency when selecting model providers, especially as AI usage scales and inference expenses become a larger operational burden.
Cost reduction goes beyond simply switching models
The report notes that these companies are not relying solely on a change in model vendor. They are also combining the adoption of lower-cost models with engineering strategies such as intelligent routing, cache optimization, and context engineering. In practice, that means assigning different requests to different models based on complexity, reducing unnecessary repeat computations through caching, and optimizing prompts or context windows to cut token usage. Together, these measures can materially reduce AI spending.
Implications for pricing and competitive dynamics
MarsBit describes the trend as one that could trigger broader changes in global AI pricing models and reshape the industry landscape. If major US firms increasingly choose providers based on cost-performance trade-offs rather than brand alone, competition in the model market may shift toward pricing efficiency, deployment performance, and practical enterprise value. For crypto-adjacent technology companies and infrastructure providers that rely on large-model capabilities, this kind of pricing reset could also influence future budgeting and vendor selection decisions.