DeepSeek and Zhipu AI Reportedly Developing In-House Chips

新智元 · wechat · 2026-07-08

Scoops from Reuters and The Information reveal that DeepSeek and Zhipu AI are secretly developing custom AI chips tailored for inference scenarios. The goal is to reduce long-term inference costs and decrease reliance on external computing power from providers like Nvidia. The article also mentions the computing pressure brought by the recent business growth of both companies and the rapidly increasing share of inference computing in overall model lifecycle costs.

Related event: DeepSeek Reportedly Developing Own AI Inference Chip to Reduce External Reliance(16 posts)→

Original post →

More from Companies & People

Companies & People channel →