U.S. Companies Shift 30%+ of AI Usage to Chinese Models

Cost pressures drive adoption of Chinese AI alternatives

U.S. companies are rapidly increasing their use of Chinese-built AI models, with usage climbing from 4.5% in the first half of 2025 to more than 30% weekly throughout 2026, according to data from OpenRouter, a platform that provides access to multiple AI models. The shift peaked at 46% during certain weeks, representing a dramatic reversal in enterprise AI procurement patterns.

The surge reflects mounting cost pressures as American AI labs raise token prices for their most advanced systems. Chinese models from companies including DeepSeek and Z.ai now cost 60% to 90% less than leading offerings from OpenAI and Anthropic, according to Justin Summerville, who works on data and analytics at OpenRouter.

Performance gap narrows as Chinese labs advance

The adoption wave is fueled not just by price but by improving capability. Chinese models now trail top U.S. systems by an estimated six to nine months, operating "close to the top American frontier models," according to Kyle Chan, a fellow at the Brookings Institution's John L. Thornton China Center.

Z.ai's GLM 5.2, released in June, saw the fastest adoption of any model tracked by Vercel in 2026. Harpreet Arora, head of agentic infrastructure at Vercel, reported that daily token volume for GLM 5.2 grew approximately 27 times in its first full week, while customer adoption increased roughly 80 times. The model scored within a percentage point of Anthropic's Opus 4.8 on a closely watched agentic benchmark while costing about one-fifth as much.

AI startup Lindy moved 100% of its traffic from Anthropic's Claude models to DeepSeek in June. CEO Flo Crivello said the switch will save the company millions of dollars within months while actually improving performance on many core use cases.

Why it matters

This trend poses strategic challenges for U.S. AI leadership. American companies are demonstrating they will prioritize cost efficiency over national origin when performance differences narrow. The shift occurs as the U.S. government attempts to regulate powerful AI models and considers restricting overseas alternatives—creating tension between commercial realities and policy objectives. If Chinese open-source models continue closing the capability gap while maintaining massive price advantages, U.S. labs may face sustained market share erosion in enterprise deployments.

Open models gain enterprise traction

Most leading Chinese models use open-source or open-weight architectures, making code and model weights available for developers to inspect and sometimes modify. This contrasts with closed systems from OpenAI, Anthropic, and Google, where inner workings remain proprietary.

"We're seeing companies increasingly motivated to turn to cheaper AI stacks they can control and adapt themselves, and given the state of open-source and open-weight models that often means leveraging Chinese options," Yacine Jernite, head of machine learning at Hugging Face, told CNBC.

Cien Solon, CEO of LaunchLemonade, an AI agent platform for regulated industries, noted that while Claude and ChatGPT still dominate usage, GLM 5.2 has entered the platform's top five models. "Businesses with more mature AI strategies are increasingly willing to use them where they make technical or commercial sense," Solon said.

These details were first reported by CNBC.