Chinese artificial intelligence models are surging in global popularity, and the speed of their rise is sharpening anxieties in Washington about the future of the AI race, the economics of enterprise software and the geopolitical reach of Chinese technology.
The issue was already high on the agenda this week as U.S. President Donald Trump and Chinese President Xi Jinping met, underscoring how central AI has become to the broader strategic competition between the two powers. While the most advanced U.S. models still lead on many benchmarks, Chinese companies including DeepSeek, Z.ai and Alibaba have released new systems with major performance gains, especially in coding and other developer-heavy tasks.
Usage data shared with CNBC suggests those improvements are translating into real-world adoption. Chinese AI models have moved from a relatively small share of usage to a majority on two major developer platforms that act as gateways for companies to access models from different providers. On OpenRouter, Chinese models accounted for 57% to 67% of tokens used in the week of Sept. 14, up sharply from 6% to 13% in February. On Vercel, their share rose to 55% in August from 11% in January.
The data points to a notable shift in how businesses are choosing AI tools: not necessarily by brand prestige or geopolitical alignment, but by cost, capability and speed. OpenRouter's figures cover companies in the U.S., Europe and what it defines as the "Global South" — 82 countries across Central and South America, Africa and Asia. Vercel did not specify the geographic breakdown of its data.
The trend is drawing close scrutiny in Washington, where two U.S. House committees are investigating the implications of rising adoption of Chinese models. U.S. policymakers have spent years trying to preserve America's AI edge by restricting Chinese firms' access to the most advanced chips through export controls. But officials are increasingly worried that Chinese companies may still be able to access Nvidia chips remotely through overseas data centers, and that they can accelerate development through "distillation," a technique in which newer models learn from older, more established systems.
For some analysts, the concern is not merely commercial competition but strategic dependence. Chinese AI represents "real economic and security risks for the United States," said Daniel Remler, a senior fellow in the technology and national security program at the Center for a New American Security, a think tank. "The ultimate concern is that the integration of Chinese AI models pulls countries into a Chinese technology sphere of influence that hardens into geopolitical alignment," he told CNBC.
That warning reflects a broader fear in Washington that AI adoption abroad could shape not only software procurement, but the digital architecture of entire regions. If companies, governments and startups in Asia, Africa and Latin America build on Chinese models, the long-term result could be a deeper reliance on Chinese technical standards, cloud infrastructure and developer ecosystems.
Peter Walker, head of insights at OpenRouter, said Chinese open-source models released this year "can credibly perform in advanced agentic use cases, especially in regards to coding, in a way that was just not true in late 2025." He added that they are "incredibly cost-effective compared to most models from American labs."
That price advantage is proving decisive. Harpreet Arora, head of agentic infrastructure at Vercel, told CNBC that "Chinese models are becoming capable enough for more tasks at a much lower cost. Once a model meets the quality bar for the job, that price difference becomes compelling." Still, he said companies continue to rely on frontier U.S. models for some of the most complex tasks.
The competitive response from U.S. firms is already underway. Earlier this week, OpenAI and Anthropic both announced new, cheaper models, signaling that pricing pressure is becoming a central battleground in the AI market. Dianne Penn, head of product management, research and labs at Anthropic, said the company was trying to make its models answer "more efficient, so it uses less tokens depending on your effort setting."
The strongest uptake of Chinese models appears to be in the Global South, where cost sensitivity is especially acute and where Chinese commercial and infrastructure ties are often already deep. More than two-thirds — 67% — of the tokens used by companies in OpenRouter's Global South category are on Chinese models. About half the tokens on OpenRouter overall are used by companies in the U.S.
Remler said Southeast Asia in particular could see significant adoption because of "the close economic and cultural linkages [with China] plus growing digital infrastructure." He added that the pattern could extend from "Lagos to São Paulo to Jakarta" as entrepreneurs and governments look for cheaper AI tools that can still deliver useful performance.
For Washington, that is the core dilemma: even if U.S. models remain technically superior at the frontier, Chinese systems are becoming good enough — and cheap enough — to win market share. If that trend continues, the contest over AI may be decided not only in benchmark charts and chip fabs, but in the everyday procurement decisions of businesses and governments around the world.