Particle.news
Download on the App Store
6 ARTICLES
·
4w ago

Ant Group Cuts AI Costs by 20% Using Chinese and U.S. Chips

The Alibaba-affiliate achieves cost efficiency with a dual-chip strategy and Mixture of Experts technique, advancing AI healthcare solutions in Chinese hospitals.

Ant Group headquarters
Image
Image
Image

Overview

  • Ant Group has reduced AI training costs by 20% through a dual-chip strategy combining Chinese-made chips from Alibaba and Huawei with U.S.-made AMD chips.
  • The company employs the Mixture of Experts (MoE) technique to maximize efficiency and reduce reliance on Nvidia's advanced semiconductors.
  • Ant's AI healthcare solutions, built on models like DeepSeek's R1 and V3, Alibaba's Qwen, and its proprietary BaiLing, are now operational in seven major Chinese hospitals.
  • The healthcare AI models can answer medical questions and improve patient services, showcasing practical applications of its cost-efficient AI advancements.
  • Ant's innovations highlight China's growing ability to adapt to U.S. semiconductor export restrictions while maintaining competitive AI performance.