🚀 Ant Group has developed AI model training techniques using Chinese-made semiconductors, potentially reducing costs by 20%.
📊 The company achieved results similar to Nvidia’s H800 chip using a mixture of experts approach in machine learning.
🇨🇳 Ant Group is increasingly relying on domestic chip alternatives, including those from its affiliate Alibaba and Huawei, for its latest AI models.
@bernardodegarcia:
“Jack Ma’s Ant Group, which you know at the time wanted to do an IPO, and then the Chinese government told them no, and did it behind Alibaba’s back, and that’s when the decline of Chinese stocks began, and now they are resurrecting. Well, they used Chinese-made semiconductors to develop AI model training techniques that would reduce costs by 20%, according to people familiar with the matter. These chips are the closest thing you can get to an Nvidia H800. Ant used domestic chips, including those from its Alibaba subsidiary and Huawei Technologies, to train models using the so-called mixture of experts machine learning approach. According to the sources, it obtained results similar to Nvidia’s chips, such as the H800, asking not to be identified since the information is not public. According to the people who were revealing this information, Ant, which is based in Hun, like Alibaba, continues to use Nvidia for AI development but is now mainly relying on alternatives, including AMD chips and Chinese chips, for its latest models, according to one of the sources.”
Watch the exact part of the video where @bernardodegarcia talks about Ant Group here:
View the video on YouTube.
Read more articles by the world’s top 100 analysts on Ant Group at the following link. BLOG.
