Chinese AI Models Now Handle a Majority of Tokens on Two Major Platforms

Chinese AI models accounted for 57% to 67% of all tokens used on OpenRouter in the week of September 14 — up from a range of 6% to 13% in February, according to usage data the company shared with CNBC. On Vercel, a second major gateway that lets companies access models from different providers, the Chinese share of usage climbed to 55% in August from 11% in January.

The numbers mark one of the fastest adoption shifts documented in the commercial AI market, and they land at a moment when AI sat near the top of the agenda as U.S. President Donald Trump and Chinese President Xi Jinping met this week.

From niche to majority in seven months

The most advanced U.S. models still lead most benchmarks. But Chinese companies including DeepSeek, Z.ai and Alibaba have released new models this year with major performance gains in tasks such as coding. Global adoption of those models by companies has increased substantially in 2026, according to the usage data.

OpenRouter's figures cover companies in the U.S., Europe and what the platform defines as the "Global South" — 82 countries across Central and South America, Africa, and Asia. Vercel did not specify a geographical breakdown for its data.

Peter Walker, head of insights at OpenRouter, told CNBC that 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." They are also "incredibly cost-effective compared to most models from American labs," he said.

Washington takes notice

The shift has triggered a response in Washington. Two U.S. House Committees are now investigating the impact of rising adoption of Chinese models.

The U.S. has spent years trying to preserve its AI lead by restricting Chinese AI companies from buying the most advanced chips through export controls. Officials now worry about two workarounds: Chinese firms accessing Nvidia chips remotely through overseas data centers, and "distillation" — the technique of training new models to mimic older, more established ones.

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 (CNAS), 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.

The price mechanism

The economic driver behind the shift is straightforward: cost per token. Harpreet Arora, head of agentic infrastructure at Vercel, told CNBC that price is the key factor in the rising adoption of Chinese models. "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," he said.

But Arora added a qualifier that matters for how the market splits: companies still want frontier U.S. models for more complicated tasks. The pattern emerging from the data is a two-tier market, with Chinese models absorbing routine, high-volume work like coding agents while American frontier models retain the most demanding workloads.

U.S. labs are responding on price. Earlier this week, both OpenAI and Anthropic announced new, cheaper models. Dianne Penn, head of product management, research and labs at Anthropic, told CNBC the company was trying to make its models' answers "more efficient, so it uses less tokens depending on your effort setting."

The Global South is the battleground

The geography of adoption is uneven, and it maps onto geopolitical lines. Businesses in what OpenRouter defines as the "Global South" have been the biggest users of Chinese AI models on the platform in recent weeks. More than two-thirds — 67% — of the tokens these companies use run on Chinese models. About half the tokens on OpenRouter overall are used by companies in the U.S.

Remler sees that concentration as strategically significant. "Southeast Asia in particular may see significant uptake of Chinese AI models given the close economic and cultural linkages [with China] plus growing digital infrastructure," he said. "Anywhere from Lagos to São Paulo to Jakarta where entrepreneurs and governments are looking for cheap, open models, will look first to Chinese AI."

Why it matters

Export controls were designed to slow Chinese AI at the hardware layer. The OpenRouter and Vercel data suggest the competition has already moved past that chokepoint at the application layer, where open-source distribution and aggressive pricing — not chip access — determine which models companies route their tokens to.

With two House Committees investigating and U.S. labs cutting prices in response, the question now is whether Washington's next move targets the adoption side of the equation — how companies in the U.S. and allied markets choose their models — rather than only the supply of chips.