American AI startups have increasingly shifted their app traffic towards Chinese large language models (LLMs), with Chinese models becoming the main growth engine for traffic in early 2026, according to OpenRouter data. This change follows a period in 2025 when U.S. models dominated weekly token consumption, illustrating a notable transition in the AI model market, which is increasingly driven by raw utility and performance rather than brand loyalty.

Thompson: Nathaniel Thompson, known online as @nxthompson, serves as CEO of The Atlantic and hosts the podcast “The Most Interesting Thing in AI.” He publicly highlighted the striking shift in model usage by American AI startups toward Chinese options since the beginning of 2026 based on OpenRouter data.
OpenRouter: OpenRouter operates as a unified API platform and marketplace that connects developers to hundreds of AI models from multiple providers through a single endpoint with automatic fallbacks and routing. It serves as the data source for tracking weekly token consumption patterns across global models. The platform’s usage metrics underpin the reported shift in traffic from American AI startups toward Chinese LLMs starting in early 2026.

Market Trend: AI model selection among developers is becoming driven more by raw utility and performance than by brand loyalty.
Platform Insight: OpenRouter provides the infrastructure for measuring and routing app traffic to a wide range of global large language models.