Chinese AI models, led by companies like DeepSeek, Z.ai, and Alibaba, are rapidly gaining market share on global developer platforms such as OpenRouter and Vercel, largely due to their competitive pricing and strong performance in tasks like coding. This surge in adoption, particularly among companies in the U.S., Europe, and the Global South, has intensified concerns in Washington regarding technology competition, national security, and China’s growing global influence.
Artificial intelligence emerged as a central topic during the recent meeting between U.S. President Donald Trump and Chinese President Xi Jinping. While leading American AI models continue to hold an edge in many benchmarks, Chinese powerhouses like DeepSeek, Z.ai, and Alibaba have introduced new models demonstrating significant performance enhancements, particularly in complex tasks such as coding.
New data obtained by CNBC reveals a substantial increase in the global corporate adoption of Chinese AI models throughout 2026. These models have shifted from representing a minor fraction of usage to becoming the majority on two key developer platforms that facilitate company access to various AI providers. On OpenRouter, Chinese models accounted for an impressive 57%-67% of tokens used in the week of September 14, a dramatic rise from just 6%-13% in February. Similarly, on Vercel, their share surged to 55% in August from 11% in January.
OpenRouter’s data encompassed companies in the U.S., Europe, and what it defines as the "Global South" – a collective of 82 nations spanning Central and South America, Africa, and Asia. Vercel, however, did not provide a specific geographic breakdown for its data.
This rapid shift is generating significant apprehension in Washington, where two U.S. House Committees have launched investigations into the growing embrace of Chinese AI models. The U.S. has been actively working to maintain its technological lead in AI, implementing export controls aimed at preventing Chinese AI companies from acquiring the most advanced chips.
Further concerns in Washington include the potential for Chinese AI firms to bypass these restrictions by remotely accessing Nvidia chips through offshore data centers, as well as their progress using "distillation" techniques where newer models learn to mimic the performance of older, more established ones.
Daniel Remler, a senior fellow in the technology and national security program at the Center for a New American Security (CNAS), articulated the gravity of the situation, stating, “Chinese AI represents real economic and security risks for the United States.” He further elaborated to CNBC, “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.”
Peter Walker, head of insights at OpenRouter, told CNBC that Chinese open-source models released this year are now "credibly perform[ing] in advanced agentic use cases, especially in regards to coding, in a way that was just not true in late 2025." He also highlighted their significant cost-effectiveness compared to most models from American research labs. In response to this competitive pressure, OpenAI and Anthropic recently unveiled new, more affordable models. Dianne Penn, head of product management, research and labs at Anthropic, informed CNBC that her company is striving to make its models’ responses "more efficient, so it uses less tokens depending on your effort setting."
Harpreet Arora, head of agentic infrastructure at Vercel, confirmed to CNBC that pricing is a crucial factor in the surging 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. However, he noted that companies still prefer frontier U.S. models for certain highly complex tasks.
Businesses within OpenRouter’s "Global South" definition have emerged as the primary users of Chinese AI models on its platform in recent weeks, accounting for over two-thirds (67%) of the tokens these companies utilize. In contrast, approximately half of the tokens on OpenRouter are used by companies located in the U.S.
CNAS’ Remler predicted significant uptake of Chinese AI models in regions like Southeast Asia, citing "close economic and cultural linkages [with China] plus growing digital infrastructure." He concluded, "Anywhere from Lagos to São Paulo to Jakarta where entrepreneurs and governments are looking for cheap, open models, will look first to Chinese AI."
