OpenAI and DeepSeek Diverge on Pricing as U.S.-China AI Business Model Battle Intensifies

Competition in generative AI has expanded beyond model performance to pricing power and infrastructure. OpenAI is using closed-source models and premium API pricing to target high-end enterprises and complex agentic tasks, while Hangzhou-based DeepSeek is pursuing broader adoption through open-source models and low prices in a bid to turn AI into a widely accessible service. The two approaches affect developer costs and ecosystem alignment, while reflecting different U.S. and Chinese business choices amid computing constraints.

On April 24, 2026, OpenAI launched GPT-5.5, with the standard version priced at $5 per million input tokens and $30 per million output tokens. DeepSeek released a preview of V4 the same day, pricing V4-Flash at just 0.2 yuan for cached input and 2 yuan for output. V4 has shifted to Huawei Ascend chips, and DeepSeek signaled further price cuts after mass production of Ascend 950 supernodes begins in the second half of the year.