Ant Group is leveraging a combination of Chinese and U.S.-made semiconductors to train more efficient artificial intelligence models, reducing costs and dependency on any single supplier. A source familiar with the strategy said the approach aligns with the industry’s shift toward Mixture of Experts (MoE) techniques, enabling significant compute savings.
Earlier this month, Ant reported a 20% reduction in AI training costs using lower-cost hardware. The company also announced major upgrades to its AI healthcare solutions, now deployed across hospitals in Beijing, Shanghai, Hangzhou, and Ningbo. Built on models including DeepSeek, Alibaba’s Qwen, and Ant’s own BaiLing, the system enhances patient services and supports clinical decision-making through AI.
