DeepSeek Is Secretly Building Its Own AI Inference Chip

Reuters reports that DeepSeek, the Chinese AI model company, has been secretly developing its own AI inference chip. The project started about a year ago and focuses on running models rather than training them. In recent months, the company has been quietly recruiting chip design engineers.

Right now, DeepSeek relies heavily on Nvidia and Huawei chips to run and train its models. A successful custom inference chip could reduce that dependency and give DeepSeek more cost-effective hardware control. But the effort is still in its early stages, and US export controls are making it tough to manufacture the chip and secure enough memory.