Alibaba-Antient Ant uses Chinese, American chips to cut costs on II

Beijing – Alibaba-Afille ants uses both Chinese and American semiconductors to create more efficient artificial intelligence models, a source familiar with this issue reports.

The combination of chips not only reduces the time and cost of learning AI models, but also limits the dependence on a single supplier, such as Nvidia, noting that in the branch tendency of pressing multiple networks known as a mixture of experts – techniques that allows models that allows models undergo training with much less calculated.

The company said earlier this month that it was able to use the equipment with a lower cost to effectively train your own Moe models, Reduced computing cost by 20%.

Ant works Alipay, one of two major mobile payments applications in China. Jack Ma founded the company and her partner, Alibaba.

Bloomberg reported Monday with reference to sourcesThis ants used chips from Alibaba and Huawei to train AI models. Ant also used NVIDIA chips, but now more counts on alternatives from advanced micro -devices and Chinese chips, Bloomberg said.

Ant refused CNBC’s request for comment.

On Monday, the company announced the “basic modernization” of its healthcare decisions, which, she said, used seven major hospitals and medical institutions in Beijing, Shanghai, Hangzhou and Ninbo.

The AI ​​Healthcare model is built on R1 and V3 Deepseek, Qwen and Ant Alibaba Qwen and Ant. The Ant, characteristic of healthcare, can answer questions on medical topics and can also help improve patients, the company said.

The United States sought to limit the development of AI China, limiting Chinese business access to the most advanced semiconductors used for training models. Nvidia can still sell its lower chips to China.