Chinese AI startup DeepSeek is making waves as domestic tech giants, from chipmakers to cloud providers, rally to support its AI models. Companies like Moore Threads and Hygon Information Technology, aiming to compete with Nvidia (NASDAQ:NVDA), announced their AI accelerators now support DeepSeek's R1 and V3 models.
Huawei Technologies has also integrated DeepSeek’s models with its Ascend cloud service, in partnership with SiliconFlow, marking what Bernstein analysts call a “watershed moment” for China’s AI industry. This development shows that competitive large language models (LLMs) can run efficiently on China’s homegrown AI chips, reducing dependence on advanced U.S. hardware like Nvidia's H800. Cambricon and Hygon are also set to release AI chips to further boost China’s capabilities.
Tech giants Alibaba (NYSE:BABA), Baidu (NASDAQ:BIDU), and Tencent have made DeepSeek’s AI accessible on their cloud platforms. The startup’s AI assistant, launched last month, quickly surpassed OpenAI’s ChatGPT in Apple (NASDAQ:AAPL)'s App Store downloads, leading to a global selloff in tech stocks.
DeepSeek first gained global recognition with a research paper in December, revealing that its V3 model was trained on less than $6 million worth of Nvidia hardware—far less than the billions spent by Microsoft (NASDAQ:MSFT) and Meta (NASDAQ:META). The success of the Hangzhou-based company and its founder, Liang Wenfeng, has made them rising stars in China’s tech landscape.
Despite its rapid growth, DeepSeek faces regulatory scrutiny, with countries like Italy and the Netherlands investigating its AI app over privacy concerns. Meanwhile, Microsoft and Amazon (NASDAQ:AMZN) have also begun offering DeepSeek’s models, solidifying its position in the AI industry.