Companies Adopt Chinese AI Models to Cut Costs and Reduce US Dependence

A growing number of global firms are moving away from U.S. artificial‑intelligence models and incorporating Chinese alternatives to lower operating expenses and lessen geopolitical risk. DoorDash, for example, now uses Moonshot AI’s Kimi K2.6 for routine tasks while retaining Anthropic’s Fable for complex work. Siemens blends Chinese providers such as DeepSeek and Z.ai with Nvidia, Anthropic and France’s Mistral to diversify its AI stack. U.S. startup Lindy has fully replaced Anthropic models with DeepSeek V4, reporting savings of millions of dollars. Airbnb runs a limited set of Chinese models through approved U.S. cloud services to protect data.

According to the Financial Times and token‑usage data from OpenRouter, Chinese‑origin models are projected to exceed U.S. models in token consumption by 2026, reflecting rapid adoption driven by lower per‑token pricing after recent pricing changes by OpenAI and Anthropic. Many Chinese models are offered as open‑weight solutions, allowing companies to host them on‑premises and retain full data control. The shift is also motivated by recent U.S. export restrictions on advanced AI models, heightening concerns about reliance on American technology providers.