KuCoin Report: US Companies' Chinese AI Model Token Usage Surges to 46%

Cryptocurrency exchange KuCoin has released a research report showing that US companies' usage of Chinese AI models is growing rapidly, with Chinese AI models' token consumption share soaring to 46%. The data is based on KuCoin's analysis of on-chain data and AI service usage patterns.

The report indicates that an increasing number of US enterprises are integrating Chinese-developed AI models into their operations, particularly in cost-sensitive and high-volume inference scenarios. Chinese models, with their competitive pricing and rapidly improving performance, are reshaping the global AI model market landscape.

The 46% token share figure is striking — it means that nearly half of all inference tokens consumed by US enterprises are now processed by Chinese AI models. If this trend continues, Chinese AI models' penetration in the global enterprise market will keep climbing.

Multiple drivers are fueling this shift: the rapid performance catch-up of Chinese AI models (such as the DeepSeek and Qwen series), highly competitive API pricing strategies, and US companies' demand for supplier diversification under cost pressure.

Notably, this trend is unfolding against a backdrop of ongoing US-China tech competition. Despite policy uncertainties, enterprise-level technology choices appear to be driven more by cost-benefit and performance metrics.

KuCoin, as a global crypto trading platform, offers a unique data lens for observing the AI industry. However, the full methodology and sample scope of the report have not been disclosed, and the 46% figure may be influenced by specific sampling periods or customer segments.

This trend carries profound implications for the global AI supply chain: if US enterprises continue to adopt Chinese AI models at scale, it could reshape the geographic distribution of model training and inference workloads, and potentially influence AI regulatory policy directions worldwide.