DeepSeek develops its own AI inference chip

4 min read

Chinese startup DeepSeek is developing a custom AI chip focused on inference, Reuters reported, a strategy aimed at reducing dependence on Nvidia and Huawei processors that have powered its globally popular models. The effort began about a year ago and includes private hiring of chip-design engineers and outreach to chip-design, foundry and memory partners, while the company prepares a planned $7 billion maiden funding round. The push mirrors global trends of AI developers seeking hardware control but faces major hurdles from U.S. export controls and restricted access to advanced foundries and high-bandwidth memory.

DeepSeek is building an internal inference chip to reduce hardware reliance