U.S. Developers Adopt Chinese AI Models to Cut Costs

Rest of World reports U.S.-based developers and small companies are adopting Chinese open models to reduce costs. Stu Clott, a San Diego developer, told Rest of World that an hourlong coding session costs about $10 on Claude and less than 50 cents on DeepSeek. Lindy, a San Francisco startup that builds AI work assistants, recently switched from Anthropic to DeepSeek. Founder Flo Crivello posted on X that the switch saved the firm millions and said: You don't need God to write your email.

Rest of World reports that on OpenRouter, a platform that routes AI tasks to different models, DeepSeek, Tencent, Minimax, and Xiaomi are now the four most popular models by usage. Vercel said DeepSeek's share of its token usage jumped from under 1% to 17% in May 2026, though revenue share stayed near 1%. Dallas developer Ruben Garcia Jr. told Rest of World he pays $500 per month for Claude and ChatGPT for complex tasks and $200 per month for Minimax, Kimi, and Xiaomi MiMo to handle 90% of routine coding and voice recognition work.

Rest of World attributes the pricing advantage to lower infrastructure and salary costs in China, combined with open-model releases and subsidised plans. DeepSeek, Xiaomi's MiMo, and Minimax rank among the most cost-efficient models on the Artificial Analysis performance-vs-cost index.

U.S. lawmakers launched investigations into Airbnb and Anysphere (owner of Cursor) after both disclosed using Chinese open models (Qwen and Kimi) in their AI infrastructure. Businesses that access Chinese models through U.S. cloud providers avoid direct data exposure to China but limit Chinese firms' ability to build long-term client relationships. Poe Zhao, founder of the Hello China Tech newsletter, told Rest of World: Chinese models are becoming part of the global AI infrastructure layer. But usage is only the first step. The Wall Street Journal separately reports OpenAI is considering significant price cuts as it competes with Anthropic, which may narrow the cost gap over time.

Kyle Chan, a Brookings Institution fellow, told Rest of World that AI adoption at large companies is fairly saturated and the growth market for Chinese providers will be mid-sized businesses starting to use AI but wary of costs. For practitioners and procurement teams, the story illustrates a real cost-performance trade-off that is already driving model selection decisions at budget-constrained startups.