Chinese AI models now eat up to 46% of US token usage
US firms are quietly routing nearly half their AI traffic to cheaper Chinese open models, up from 11% a year ago.
US companies are quietly shifting serious workload to Chinese open-weight models. The share of tokens American firms route through OpenRouter to Chinese models has stayed above 30% every week since February, spiking as high as 46% — up from a 12-month average of just 11%, per CNBC.
The driver is boring but brutal: cost. Open Chinese models like DeepSeek run 60–90% cheaper than frontier offerings from OpenAI and Anthropic, whose token prices keep climbing. By June, the platform's top Chinese models were processing ~18 trillion tokens/week against ~5.5T for US models.
Engineers aren't making a geopolitical statement — they're following the invoice. The most capable cheap model increasingly ships from China, and for high-volume backend jobs, capability-per-dollar wins over brand.
Why it matters: the US frontier-lab lead means little if developers vote with their wallets and route the actual work elsewhere.
Sources
Independent coverage
Written by an AI pipeline from the sources above. Methodology · Report an error
Feed, daily deep-dive and bytes — readable offline, with push alerts for the topics you follow.