AsianFin -- Ant Group Co., backed by billionaire Jack Ma, has developed innovative techniques for training artificial intelligence (AI) models using domestically produced semiconductors, achieving a 20% reduction in costs, according to sources familiar with the matter.

The financial technology giant utilized Chinese-made chips, including those from affiliate Alibaba Group Holding Ltd. and Huawei Technologies Co., to train AI models using the advanced "Mixture of Experts" machine learning approach. The results were comparable to those achieved with Nvidia Corp.'s H800 chips, the sources revealed, requesting anonymity as the information is not publicly disclosed. While Ant continues to use Nvidia chips for some AI development, it has increasingly shifted to alternatives, including processors from Advanced Micro Devices Inc. (AMD) and other Chinese suppliers, for its latest models.

This breakthrough highlights Ant Group's entry into the intensifying competition between Chinese and U.S. companies in AI development. The race has gained momentum since DeepSeek demonstrated that highly capable AI models can be trained at a fraction of the costs incurred by industry leaders like OpenAI and Alphabet Inc.'s Google. Ant's achievement also underscores the broader efforts by Chinese tech firms to rely on local alternatives to advanced Nvidia semiconductors, particularly in light of U.S. export restrictions.

Although the H800 is not Nvidia's most advanced chip, it remains a powerful processor and is currently prohibited from being exported to China under U.S. regulations. Ant's success in leveraging Chinese chips to achieve similar results signals a significant step forward in reducing dependency on foreign technology and advancing China's domestic semiconductor capabilities.

This development reflects the growing emphasis on self-reliance in China's tech sector, as companies strive to innovate and compete globally while navigating geopolitical challenges and supply chain constraints.