Chinese AI Models Grab Up to 46% of US Enterprise API Tokens
In one line: Chinese-origin AI models have held above 30% of weekly enterprise API token volume on OpenRouter every week since February 8, 2026 — peaking at 46% — as price gaps widen against US rivals.
Key points
- Share surged from a 4.5% average in H1 2025 to a sustained 30%+ since February 8, 2026, with a 46% peak.
- Week of Feb 9–15: Chinese models processed 4.12 trillion tokens, surpassing US models for the first time.
- Price gap is decisive: DeepSeek V4 Flash costs $0.14/M input tokens; OpenAI GPT-5.5 costs $5.00/M — a 36× difference.
- DeepSeek ranked as the No. 1 trending software vendor on the Ramp corporate spending index, signaling formal budget adoption, not just developer sandboxing.
Why it matters
The shift from experimental usage to formal corporate procurement means cost pressure on OpenAI and Anthropic is structural, not cyclical. The trend puts enterprise AI buyers squarely in the middle of a national-security-versus-cost-efficiency debate with no easy resolution.