Newsletter N°81 - October 2025 

Chinese technology companies Alibaba and Baidu are increasingly adopting domestically developed processors to train their artificial intelligence models, signaling a gradual reduction in reliance on US chipmaker Nvidia. Alibaba’s Zhenwu PPU is reported to offer performance comparable to Nvidia’s China-specific H20 GPU, while Baidu employs its Kunlun chip to support training of the Ernie AI model.

​

The move is driven largely by US export restrictions, which limit China’s access to advanced Western semiconductors. While neither company has fully abandoned Nvidia chips—particularly for their most complex and demanding AI models—the trend reflects a broader push toward technological self-reliance. Analysts note that this shift could influence the regional market for high-performance AI chips and reshape supplier dynamics.

​

This development illustrates a rapidly evolving AI chip ecosystem in China, where domestic innovation is accelerating in response to geopolitical pressures. The adoption of in-house processors also underscores China’s strategic emphasis on reducing dependency on foreign technology, especially in critical sectors such as AI and data processing.

​

In parallel, Nvidia is reportedly exploring the development of new chips specifically for the Chinese market, although such plans require approval from the US government due to export regulations. The company’s potential adjustments indicate the complexity of balancing commercial opportunities with compliance to international trade restrictions.

​

Experts suggest that the trend may lead to increased competition between domestic and foreign chipmakers, driving further innovation in AI hardware. It also highlights the broader challenge for global technology companies operating in China, as local alternatives become more viable and geopolitical considerations increasingly influence procurement decisions.

​

​​

​

​

​

​

​

​

​​

​

​

​

​​