Chinese AI startup DeepSeek is developing its own custom chip designed for inference rather than model training, according to three people familiar with the matter. The company, which currently relies on Nvidia and Huawei chips to train and run its models, began the chip project roughly a year ago and has intensified hiring of chip-design engineers. Nvidia shares fell about 1.6% in pre-market trading on the reports.
The move follows a broader industry trend, with OpenAI unveiling its first custom inference chip last month and Anthropic also exploring in-house chip development. DeepSeek is seeking to raise $7 billion in a funding round that would value it between $52 billion and $59 billion.
In markets, Nasdaq, S&P 500, and Dow Jones all opened lower, while oil prices climbed after reports of Iranian attacks on commercial vessels in the Strait of Hormuz. Samsung reported record second-quarter results with operating profit up 19-fold.