China’s AI Chatbot Price War Escalates as DeepSeek Cuts API Rates by 75%

China’s generative AI race is heating up — not just in model development but in aggressive price wars aimed at attracting developers and enterprise users. On April 8, emerging AI startup DeepSeek slashed the API usage rates of its large language model “DeepSeek-R1” by up to 75%, undercutting competitors like Tencent, Baidu, and iFlytek.

DeepSeek’s new pricing puts its model inference costs significantly below market averages, with tokens priced at less than ¥0.0005 ($0.00007) per call, according to public documentation. This follows Baidu’s move on April 1 to make its flagship ERNIE Bot (文心一言) completely free to the public — including previously paid premium tiers — in a bid to boost its user base and developer community.

Major players are now competing not only on model capabilities but also on cost, developer experience, and ecosystem integration. Tencent Cloud recently introduced lower pricing for its “Hunyuan” model APIs, while iFlytek has bundled its AI model services with education and healthcare platforms to drive adoption.

This price-cutting wave mirrors similar trends seen in cloud computing over the past decade, where hyperscalers like AWS and Alibaba Cloud lowered prices to scale usage and lock in long-term users. In China’s AI space, it reflects a strategic shift from hype to monetization, as firms seek to convert massive R&D investments into sticky enterprise customers.

However, analysts warn the pricing war could squeeze profit margins for smaller players while favoring those with cloud infrastructure advantages or corporate backing. DeepSeek, which spun out of a top academic lab, has received backing from several Chinese tech VCs and is betting that scale will bring sustainability.

As more enterprises experiment with LLM-based services, affordability may become the next major differentiator — especially in markets like China, where cost-efficiency often drives adoption more than brand prestige.