Chinese chipmakers and cloud providers are swiftly aligning with DeepSeek's AI models, marking a pivotal industry development. On February 3, Moore Threads and Hygon Information Technology, notable for AI chip production and direct competitors to Nvidia, announced their support for DeepSeek's R1 and V3 models across their computing infrastructures. Moore Threads expressed admiration for DeepSeek, suggesting that these advancements might revolutionize China’s AI sector using local graphic processing units.
Similarly, on February 1, Huawei Technologies revealed a collaboration with SiliconFlow, aiming to enhance the availability of DeepSeek’s models on its Ascend cloud service. Huawei asserted that the performance of these models on its chips matches those run on top-tier international hardware. This integration signifies a major breakthrough, described by analysts from Bernstein as a "watershed moment," showcasing China's capability to host competitive large language models on domestically produced chips, thus reducing dependency on advanced US technology.
Alibaba, Baidu, and Tencent, China’s largest internet entities, also confirmed the integration of DeepSeek’s AI into their cloud offerings. DeepSeek, earlier in January, introduced a cost-efficient AI assistant that quickly surpassed ChatGPT in Apple’s App Store downloads, leading to a global tech stock sell-off. Moreover, DeepSeek gained further acclaim after revealing that the training of its V3 model on Nvidia’s H800 chips cost significantly less than the billions typically spent by tech giants like Meta and Microsoft.
This rapid adoption has turned the Hangzhou-based startup and its founder, Liang Wenfeng, into cultural icons, even as global giants like Microsoft and Amazon begin hosting DeepSeek’s models. However, the growing popularity also faces scrutiny over privacy concerns in various nations, including Australia, Italy, and the Netherlands.
