Ant Group Taps Chinese Chips to Cut AI Training Costs by 20%

Ant Group, backed by Jack Ma, is leveraging Chinese-made semiconductors to reduce the cost of training artificial intelligence models by up to 20%, according to a Bloomberg report. The breakthrough, achieved using domestic chips from Ant affiliate Alibaba Group (NYSE:BABA) and Huawei Technologies, highlights China’s growing self-reliance in AI development amid U.S. export restrictions.

Despite still using NVIDIA (NASDAQ:NVDA) chips like the H800, Ant is increasingly experimenting with alternatives from AMD (NASDAQ:AMD) and local manufacturers. Sources familiar with the matter said Ant’s new techniques delivered results on par with those produced using high-end Nvidia hardware.

This development aligns with a broader trend among Chinese tech firms to shift toward local chip solutions after the U.S. imposed strict curbs on AI chip exports, blocking companies like Nvidia from selling their most advanced products to China. The restrictions were expected to hamper China's AI progress, but recent advancements have proven otherwise.

In January, the release of DeepSeek R1—a Chinese AI model—demonstrated performance comparable to OpenAI’s ChatGPT, despite being trained with older, less expensive hardware. The model's success spurred similar AI launches from Chinese giants such as Baidu (NASDAQ:BIDU), Alibaba, and Tencent.

These milestones have fueled investor confidence, contributing to a sustained rally in Chinese tech stocks throughout 2025. The progress underscores China’s determination to compete in the global AI race by overcoming geopolitical and technological hurdles with domestic innovation.