Chinese AI models are rapidly gaining ground with American businesses. Data from OpenRouter shows they have captured over 30% of token usage by US companies each week since February 8, peaking at 46%. This marks a dramatic surge from just 11% over the previous 12 months.
The shift reflects a narrowing performance gap between Chinese and US AI systems. As state-backed and private Chinese developers improve their large language models, cost advantages are drawing budget-conscious firms. The trend challenges the assumption that American AI dominance in the enterprise segment is secure.
Token usage—a proxy for actual model adoption—reveals the scale of this pivot. OpenRouter, a platform that routes queries to various models, recorded the peak 46% share during the period. The data underscores a real-time competitive shift rather than just market sentiment or investment flows.
For US AI companies like OpenAI and Anthropic, the rise of Chinese alternatives introduces pricing pressure and erodes an assumed lead in quality. Enterprise clients now have viable substitutes, potentially reshaping procurement strategies and margins across the sector. Geopolitical tensions could complicate further adoption, however.
The longer-term implications depend on model improvement rates and export controls. If Chinese models continue closing the capability gap, their token share may climb even higher, forcing US incumbents to compete more aggressively on both performance and price.