Chinese AI startup DeepSeek is developing its own chip for artificial intelligence inference. The move aims to reduce the company's reliance on chips from Nvidia and Huawei.
The chip is designed for inference, the process where a trained AI model generates responses. This marks a strategic shift for the Hangzhou-based company, which has been hailed as China's AI champion.
DeepSeek gained global attention with highly efficient AI models. The effort is in its early stages, with the company engaging external partners and privately hiring chip-design engineers.
The push aligns with a broader industry trend. OpenAI recently unveiled a custom inference chip, while Anthropic is also exploring similar development.
The strategy carries geopolitical weight. U.S. export controls block Chinese firms from buying Nvidia's most advanced chips, pushing Beijing to champion domestic alternatives.
DeepSeek has used both Nvidia and Huawei chips. Its R1 model was trained on Nvidia's H800. More recently, it has leaned on Huawei's Ascend chips.
An in-house inference chip targets a rapidly growing segment of AI computing. However, designing a competitive chip takes years and significant capital, and U.S. manufacturing restrictions pose major hurdles.
This development comes as DeepSeek prepares for its first external funding round, reportedly seeking $7 billion at a valuation exceeding $50 billion.