Chinese AI models surge in global popularity — and Washington is worried
- ID
- 28863
- Status
- summarized
- Published
- 26 Sep 2026, 1:00 PM
- Fetched
- 26 Sep 2026, 2:06 PM
- Provider
- CNBC Technology
- Category
- technology
- Original URL
- https://www.cnbc.com/2026/09/26/china-ai-global-adoption.html
- Source URL
- https://www.cnbc.com/id/19854910/device/rss/rss.html
Summary
- Score
- 7.5
- Created
- 26 Sep 2026, 2:06 PM
- Tags
- Audience
- developersai_ml_learnersai_agent_userssaas_founders
What happened
Chinese AI models went from a small minority to a majority of tokens on two major model-gateway platforms in 2026: on OpenRouter they were 57%-67% of tokens in the week of Sept. 14, up from 6%-13% in February, and on Vercel they hit 55% in August, up from 11% in January. DeepSeek, Z.ai and Alibaba released models with large gains on coding and other agentic tasks, and lower prices are driving the shift, though U.S. frontier models still attract more overall spending. The trend is now under scrutiny in Washington, with lawmakers investigating Chinese AI use, and AI featured in this week's Trump-Xi meeting.
Why it matters
If you route through OpenRouter or Vercel, the default economic choice has flipped: the majority of tokens flowing through those gateways are now Chinese models, so pricing benchmarks for coding and agentic workloads should be re-checked against DeepSeek/Z.ai/Alibaba rather than assumed from US frontier pricing. The split matters more than the headline — Chinese models win token volume, US frontier models still win spending, which suggests teams are using cheap Chinese models for high-volume agentic loops and paying frontier prices only for the hardest tasks. The Washington investigation is the practical risk: if you sell into US government, defence, or regulated enterprise, model provenance and data routing may become a procurement question, so know which provider your gateway actually calls.
Discussion angle
The usage data covers 82 countries across Central and South America, Africa and Asia — if Chinese models are winning on price for agentic and coding tasks while US models still take more spend, where would you actually switch in your own stack, and what would stop you (compliance, tooling, evals, latency)?