US Developers Quietly Shift to Chinese Open-Weight AI Models to Cut Costs

American developers and enterprises are increasingly adopting open-weight AI models from Chinese labs like DeepSeek and Alibaba's Qwen, moving away from costly proprietary APIs such as OpenAI's GPT-4 and Anthropic's Claude. The shift is driven by economics and flexibility rather than ideology, as Chinese models now rival frontier performance at a fraction of the token cost. US export controls on advanced chips paradoxically pushed Chinese labs to develop highly efficient software architectures, including Mixture-of-Experts frameworks and Multi-head Latent Attention, which reduce hardware requirements significantly. Mozilla CTO Raffi Krikorian is among those who have publicly acknowledged integrating Chinese-origin models into developer workflows. However, analysts caution that total cost of ownership for self-hosted open-weight models still includes infrastructure, engineering integration, and compliance expenses that enterprises must carefully evaluate.