Alibaba, Baidu Shift from Nvidia to Domestic AI Chips Amid US Export Limits

Chinese technology giants Alibaba and Baidu have begun transitioning to domestically produced chips for training their AI models, aiming to reduce reliance on U.S. suppliers like Nvidia amid tightening export restrictions. Both companies are developing and deploying homegrown processors such as Alibaba’s Zhenwu unit and Baidu’s Kunlun P800 chip, reflecting Beijing’s strategic push for semiconductor self-sufficiency backed by significant state support. Despite this progress, Chinese chips still lag behind Nvidia’s most advanced offerings, which remain largely inaccessible due to U.S. export controls. Top Chinese government advisers, including Wei Shaojun of Tsinghua University, have publicly urged China and other Asian nations to move away from Nvidia GPUs, warning that continued dependence on U.S. general-purpose graphics accelerators could jeopardize regional technological autonomy and AI development. Wei advocates for the creation of specialized processors tailored specifically for large language model training rather than repurposed gaming GPUs, emphasizing that Asia’s AI strategy must diverge from the American model to avoid long-term strategic risks. Meanwhile, emerging Chinese AI firms like DeepSeek are demonstrating notable algorithmic advances despite hardware constraints, signaling a competitive domestic AI ecosystem evolving under geopolitical pressures.