DeepSeek, the China-based AI start-up, is reportedly developing its own AI chip to reduce dependence on processors from Nvidia and Huawei. The chip is designed to tackle inference workloads, a stage of AI computing where trained models generate responses for users. The company has been working on the project for about a year and is in talks with external partners including chip-design firms, foundries and memory suppliers, while also increasing recruitment of chip design engineers privately without public job listings. Developing its own silicon could alleviate growing pressure from U.S. export restrictions limiting Chinese companies’ access to advanced AI chips. DeepSeek has previously relied on Nvidia and Huawei hardware, including Nvidia’s H800 chip to train the foundation model behind its R1 reasoning model. However, developing in-house AI silicon could prove challenging as China’s access to critical minerals, advanced manufacturing facilities and high-bandwidth memory components remains constrained by U.S. restrictions. Domestic rivals including Alibaba and Baidu have previously developed their own chips, increasing competition in China’s AI semiconductor space. Globally, AI companies are seeking greater control over hardware; last month, OpenAI unveiled Jalapeno, its first custom inference chip developed with Broadcom, while US AI player Anthropic has reportedly weighed developing its own AI chips.