Chinese AI models are processing a rapidly growing share of the workloads run by US companies, according to usage data from developer platform OpenRouter reported by CNBC on July 7. The share of tokens American firms route to Chinese-made models has stayed above 30% every week since February 8, and has peaked as high as 46% — up from an average of just 11% over the prior twelve months.
Price is the driver
The shift tracks almost entirely with cost. Open-weight Chinese models can run 60% to 90% cheaper than comparable systems from Anthropic and OpenAI, according to OpenRouter. Harpreet Arora, who leads agentic infrastructure at hosting platform Vercel, told CNBC that “price is doing the work here,” adding that when a task doesn’t require the top-tier model, engineering teams are increasingly routing it to “the cheapest one that’s good enough.”
The clearest example is Lindy, an AI startup that moved 100% of its production traffic from Anthropic’s Claude to DeepSeek, telling CNBC the switch will save it millions of dollars a year. Z.ai’s GLM-5.2, released in June, saw the fastest adoption of any model Vercel tracked in 2026: daily token volume on its platform grew roughly 27-fold in the model’s first full week.
Performance gap narrows, spending gap doesn’t
Benchmarks suggest the quality gap has narrowed enough to make the trade worthwhile for many tasks — GLM-5.2 has scored within about a percentage point of Claude Opus 4.8 on an agentic coding benchmark while costing roughly a fifth as much, per data cited by CNBC. Still, the swing is concentrated in raw token volume rather than total spending: US labs, and Anthropic in particular, continue to command the large majority of dollars spent on frontier AI, especially for high-stakes enterprise work where reliability outweighs price.
A year ago, models from Google, OpenAI and Anthropic together accounted for roughly 70% of token volume on OpenRouter; by June 2026 that combined share had fallen to around 30%, with Chinese labs including DeepSeek, Zhipu AI and Moonshot AI filling the gap.