Sequoia Warns U.S. Firms Dependent on Chinese AI Models; Qwen Usage Surges to 69% as of February 2026

According to Sequoia partners Dean Meyer and Konstantine Buhler, U.S. enterprises are increasingly relying on Chinese open-weight models like Qwen, Kimi, and DeepSeek for product foundations and training data. ATOM data shows Qwen's share among newly fine-tuned and adapted models jumped from 1% in January 2024 to 69% as of February 2026.

The partners attribute this trend to asymmetric rules around model distillation. While OpenAI and Anthropic restrict customers from using their model outputs to train competing products, U.S. firms can legally learn from Chinese open models. If the trend continues, the U.S. may retain leadership in closed-source models but cede the open-model foundation to China.