DeepSeek has disrupted the AI industry by offering an unprecedented price-to-performance ratio. So far, the AI-focused company has utilized high-performance graphics processing chips from the American tech giant NVIDIA to train its AI models. However, DeepSeek has now joined hands with the cloud-computing unit of Huawei Technologies. DeepSeek is now providing its artificial intelligence models through cloud servers powered by Huawei chips.
DeepSeek is offering its V3 and R1 AI models on cloud servers powered by Huawei chips
According to the South China Morning Post, Huawei Technologies’ cloud-computing subsidiary has collaborated with Beijing-based AI infrastructure startup SiliconFlow to provide DeepSeek’s AI models available to end users via its cloud servers. Huawei’s cloud-computing unit has worked overtime with the local company during the Lunar New Year holidays to make DeepSeek’s V3 and R1 AI models available to users efficiently and affordably.
DeepSeek’s V3 and R1 AI models are available to end users through Huawei’s Ascend cloud service. It offers performance matching “DeepSeek models run on global premium graphics processing units”, Huawei cloud said in a statement on Saturday.
DeepSeek’s AI models on Huawei’s cloud servers are available for less than one US dollar
SiliconFlow stated that it discounted the charge for DeepSeek’s V3 AI model on its platform to 1 Yuan. It is roughly about $0.14 and provides users with 1 million input tokens. Users can also purchase 1 million output tokens by paying 2 Yuan or $0.28. SiliconFlow is charging 4 Yuan (~$0.55) and 16 Yuan ($2.20) for the R1 model for 1 million inputs and 1 million outputs, respectively.
Huawei and SiliconFlow’s latest move comes as the open-source DeepSeek models are gaining popularity in both the US and China. OpenAI investor Microsoft recently added support for DeepSeek’s R1 model on its Azure cloud-computing platform and GitHub. In addition, the e-commerce giant Amazon has also enabled developers to create apps using the “powerful, cost-efficient” DeepSeek R1 AI model through Amazon Web Services.
Despite several states and companies banning it, an overwhelming server demand exists for DeepSeek AI models in the US.
