The rise of Chinese AI models is accelerating: in the first week of July, systems developed in China accounted for 63% of token usage by U.S. companies on the OpenRouter marketplace, up from under 10% a year earlier. The shift comes as President Xi Jinping urges China to take a larger role in shaping global rules for artificial intelligence.
Chinese AI models: rapid adoption on OpenRouter
The change in market share on OpenRouter highlights a quick pivot toward open, lower‑cost alternatives. DeepSeek recently became the most popular Chinese model among U.S. firms on the platform, and Lindy AI reported that migrating models from Anthropic to DeepSeek cut its inference costs by 90%. Those figures point to strong commercial interest in open‑weight models that can materially reduce operational spending.
New competitors and market reactions
Competition has intensified with new entrants. Chinese startup Moonshot unveiled the Kimi K3 model shortly before Xi’s speech, claiming it can compete with top offerings from OpenAI and Anthropic. The announcement put pressure on tech and semiconductor stocks, with investors likening the response to last year’s “DeepSeek moment.”
Political stakes and influence efforts
At the World AI Conference in Shanghai, Xi emphasized that AI must remain controllable by humans and called for international cooperation on safety, technical standards and access for developing economies. He also promoted the World AI Cooperation Organization, a new body that brings together nearly 30 countries and could expand Beijing’s influence over global AI governance.
Regulatory risks and cross‑border tensions
The rapid spread of Chinese AI models also creates regulatory risks. In China, officials have raised security concerns about Alibaba—developer of the Qwen model family—and are reportedly considering restrictions on foreign access to the most capable systems, though those discussions remain preliminary and no enforcement measures have been announced. In the U.S., probes have been opened into the use of Chinese models by companies such as Airbnb and Anysphere (publisher of the Cursor platform), and Palantir has described these open models as an economic threat. These developments could fragment global research networks or limit access to lower‑cost technologies.
The recent trajectory shows Chinese AI models are no longer confined to domestic markets: their adoption by U.S. companies highlights immediate economic choices and raises medium‑term strategic questions for private firms and regulators on both sides of the Pacific.
