China's Cheap AI Models Encroach on Market, Pressuring Anthropic and OpenAI's Developer Ecosystem
Claire Weston
DeepSeek charges roughly $0.87 per million output tokens; OpenAI charges about $30, Anthropic about $25. That 30-fold price gap is pushing U.S. startups to switch en masse — and eating into the two leaders' developer market share.
How wide is the price gap?
DeepSeek's latest flagship model costs about $0.87 per million output tokens. OpenAI charges roughly $30; Anthropic roughly $25.
This means → for the same workload, DeepSeek costs about one-thirtieth of OpenAI. For a cash-burning startup, this is not "a bit cheaper" — it is an entire order of magnitude off the bill.
Who has already switched?
San Francisco AI-assistant startup Lindy (a 25-person team) has moved from Anthropic's Claude to DeepSeek. CEO Flo Crivello says the switch saved "millions of dollars."
The key detail: Lindy's AI costs had already exceeded its payroll. In plain terms = a 25-person company was spending more on API calls than on salaries — the incentive to switch is self-evident.
Coding tasks, however, still run on Anthropic, because its Max plan carries a heavy subsidy. This reflects Anthropic using discounts to defend its highest-value use case.
What does the platform data show?
Vercel, a cloud platform tracking developer AI traffic, reports DeepSeek's share jumped from under 1% in May to 17%.
OpenRouter, an AI-model routing platform, shows DeepSeek usage doubled between January and June 2026, making it the most popular model among its users.
This means → the shift is not one company's decision — it is a trend-level migration across the developer community.
Where is the lost share coming from?
Token share for Chinese open-source models — Xiaomi, MiniMax, Tencent — has been rising steadily over the past six months.
Those gains came mainly at the expense of Google and OpenAI. In plain terms = the users Chinese models are winning are not new demand appearing from nowhere — they are being pried away from U.S. incumbents.
This signals something deeper: once the capability gap narrows to "good enough," price becomes the deciding factor.
Content is for reference only, not financial advice.