Chinese Fintech Giant Ant Group Reduces AI Costs Using Domestic Chips

In a significant development, Chinese fintech giant Ant Group has successfully reduced its artificial intelligence (AI) training costs by 20% using domestically produced chips. We were able to reach this milestone by utilizing chips manufactured by Alibaba and Huawei. These chips only slightly underperformed the in-demand Nvidia chips. Nvidia’s newest Blackwell chip, for example, continues to be in high demand in China despite even the most stringent U.S. export controls.

Production of AI chips has made Nvidia, perhaps better than any other company, the most profitable company in history. However, earlier this year, its stock experienced a brief but harrowing crash. This came largely due to speculation that DeepSeek’s models would be able to run with fewer Nvidia chips. This terrible situation highlighted the critical importance of Nvidia’s chips and the promise of new alternatives to them in emerging architectures.

Ant Group, whose founder is Alibaba founder Jack Ma, is leading the technological innovation currently sweeping the global fintech sector. The company’s recent deep loss, via experiments by TSMC and Chinese chips from Alibaba and Huawei, underscore a momentous turnaround. China is making a push for greater self-reliance in technology. In testing, these homegrown chips demonstrated performance capability on par with Nvidia’s offerings. This only increases their attractiveness as a potential alternative for AI use cases.

Ant Group’s strategic decision to use domestically made chips contributes to the national push to accelerate technological self-sufficiency. Besides improving the quality and accountability of our tech investments, this decision will reduce supply chain disruption and geopolitical risk associated with technology that we import from abroad.

Despite repeated questions, Nvidia did not respond immediately with comment on these surprising moves. Likewise, Ant Group has been silent on all platforms about any further clarifying reply or local responses with their strategic use of Chinese chips. The absence of authenticated high-level guidance opens the door for speculation regarding longer-term trends in AI chip usage and broader market directions.