China's DeepSeek reportedly developing its own AI chip amid Chinese firms’ shift to local suppliers

Chinese startup DeepSeek is reportedly developing its own artificial intelligence (AI) chip designed for inference, the stage where a trained AI model generates responses for users rather than training new models, three anonymous sources told Reuters.

News of the startup's move came after a Bloomberg Intelligence survey released on Tuesday found that Chinese companies are increasingly turning to domestic AI products instead of Nvidia's, with nearly half of executives saying they plan to allocate their AI accelerator budgets to local suppliers over the next 12 months.

DeepSeek, which relied on Nvidia and Huawei chips, gained global attention last year after releasing two open-source models that cost significantly less and required less computing power than rivals such as OpenAI's ChatGPT. 

One of the three sources said that DeepSeek began exploring the development of its own AI chip about a year ago, reaching out to external partners and holding discussions with chip-design, foundry, and memory companies.

Meanwhile, two of the sources added that the company has quietly expanded its chip-design team in recent months, recruiting engineers privately instead of job postings on public hiring platforms.

All three sources declined to be identified because the information is not public, while DeepSeek has not responded to a request for comment.

Reports that the Hangzhou-based startup is developing its own AI chip also sent Nvidia shares down 1.6% in premarket trading on Tuesday, Investor’s Hub reported. /TISG

Read also: Sam Altman signals OpenAI’s desire to partner with China after DeepSeek disrupts the AI market