Chinese startup DeepSeek is working on its own AI chip to lessen its reliance on Nvidia and Huawei chips, which are currently essential for training and running its popular AI models. This new chip will focus on inference, the part of AI that generates user responses rather than training new models. If DeepSeek succeeds in creating this chip, it would signify a significant shift for a company recognized as a leader in AI within China and could challenge Huawei’s market position.
DeepSeek gained international recognition over a year ago with two AI models that became very popular, surprising many in Silicon Valley and Washington. The company has traditionally focused more on AI model advancements rather than commercializing its technology. Despite Huawei’s presence in the market, it only offers chips that lag behind Nvidia’s top products. A U. S. export ban on advanced chips has allowed Huawei to secure approximately half of the $50 billion domestic AI chip market, serving DeepSeek and other leading firms. However, Huawei’s market strength is weakening as competitors like Alibaba and Baidu develop their chips.
DeepSeek’s efforts in chip development are still in the early stages. The company has been seeking partnerships and holding talks with various chip design and manufacturing firms over the past year. Additionally, DeepSeek has been quietly hiring chip design engineers without public job postings. Though a prominent figure in China’s AI ambitions, DeepSeek has maintained a low profile and has not responded to requests for comments.
DeepSeek’s potential in-house chip is part of a trend where global AI developers aim to gain more control over their hardware and decrease reliance on Nvidia. Other companies, like OpenAI and Anthropic, are also working on their AI chips due to similar motivations. DeepSeek founder Liang Wenfeng acknowledged that U. S. export controls pose challenges for the company. Historically, DeepSeek has used both Nvidia and Huawei technologies for its models.
The proposed inference chip targets the rapidly expanding segment of AI computing that focuses on using models rather than training them. However, designing a competitive AI chip is complex and costly, and manufacturing poses additional difficulties given U. S. restrictions on Chinese access to advanced foundries and critical memory components. Furthermore, DeepSeek plans to raise $7 billion in funding, reflecting a shift from its previous strategy of rejecting outside investment.
With information from Reuters
