The artificial intelligence (AI) industry has been shaken lately after DeepSeek, the Chinese AI startup, claimed that they only spent less than $6 million on equipment (including the chips) to train DeepSeek R1’s predecessor, DeepSeek-V3.
Silicon Valley and other big US tech companies felt the news of the cost-effective Chinese AI model more strongly, as the open-source chatbot leaped to the top of the App Store in 51 countries.
While the R1 LLM was trained using the industry standard Nvidia H800 (a modified version of H100 exported to China), it’s been revealed that the Chinese AI model is using a Huawei AI chip, Ascend 910C (an alternative to H100), for inference, which is the process of using AI models to make predictions or solve problems using live data.
This recent information comes from an X post from user @Dorialexander (Alexander Doria),
“I feel this should be a much bigger story: DeepSeek has trained on Nvidia H800 but is running inference on the new home Chinese chips made by Huawei, the 910C,” said Doria.
This new update indicates that AI computing might be switching away from expensive Nvidia chips in exchange for cheaper Huawei chips from China. This can also be a significant factor in the AI race, which heavily relies on these AI chips, which used to be dominated by Nvidia.
The thread also indicated a potential shift to Huawei’s upcoming 920C chip, which is said to be comparable with Nvidia B200’s (current flagship) performance, will be used in training DeepSeek V4.
Only time will tell whether Huawei can catch up with Nvidia, but the stock market responded with the recent massive sell-down on NVD stocks.
This article, DeepSeek AI is powered by Huawei chips, was originally published at NoypiGeeks | Philippines Technology News, Reviews and How to's.
Comments
Post a Comment