Companies from Silicon Valley to Europe are turning to Chinese AI models to cut costs and reduce dependence on US labs. Door Dash, Siemens, and Airbnb have adopted tools built in China, drawn by cheaper, increasingly capable models that run on their own infrastructure. Chinese models from Deep Seek and Z.ai have overtaken US rivals in token consumption this year, according to Open Router. The shift is largely driven by cost, as companies face ballooning AI bills; in Europe it has a geopolitical edge after US export controls on Anthropic’s models. Eugene Cheah of Featherless AI said enterprises now realize they can use faster, cheaper models. Door Dash’s Andy Fang now delegates “lower‑level work” to Kimi K2.6 from Moonshot AI, reserving Anthropic’s Fable for the hardest tasks. Siemens uses a mix of Deep Seek, Z.ai, Anthropic, Nvidia and French Mistral. Some firms, like Lindy, have switched entirely to Chinese models, saving millions. Open‑weight models are 10–60 times cheaper than proprietary ones, giving companies control and customization.

Open‑weight models’ popularity rose after the US ban on Anthropic’s Fable, especially in Europe, where firms seek sovereign AI solutions. Critics note that while Chinese models improve, geopolitical risks still exist. The trend reflects a broader move toward diversified AI sourcing to avoid single‑vendor dependence.