Chinese technology giants Alibaba and Baidu are increasingly leveraging their in-house designed chips to power and accelerate their AI model training initiatives. This move signifies a strategic shift away from heavy reliance on US-based Nvidia, driven by a combination of factors including increasing U.S. export restrictions and a push for technological self-sufficiency within China.
Driving Forces Behind the Shift
The U.S. government's tightening of export controls on advanced AI chips to China has significantly impacted the access of Chinese companies to cutting-edge technology. These restrictions, aimed at limiting China's technological advancement, have spurred the nation to develop its own domestic alternatives. Simultaneously, the Chinese government has been encouraging domestic companies to prioritize domestically produced processors, further fueling the transition. This push for self-reliance aligns with national objectives to bolster domestic innovation and achieve greater autonomy in critical technological sectors.
Alibaba's AI Chip Initiatives
Alibaba has reportedly begun utilizing its custom-designed processors, including the Zhenwu processing unit, for training smaller AI models earlier in 2025. According to reports, the most advanced Zhenwu chip performs moderately better than Nvidia's A100, which was released in 2020. Some company engineers even suggest that Alibaba's AI chips now rival the H20 in training efficiency for certain tasks. Alibaba has reportedly collaborated with Semiconductor Manufacturing International Corporation (SMIC), a partially state-owned foundry in China, to produce its Zhenwu chips.
Baidu's AI Chip Efforts
Baidu is piloting its latest Kunlun P800 chip in the development of newer versions of its Ernie large language model. The Kunlun P800 chip has primarily been used in the post-training, or refining stage, of AI models. Baidu's chip-design unit, Kunlunxin, also secured a $139 million AI chip deal with China Mobile for its AI projects. Earlier in 2025, Kunlunxin stated that it had begun operating a cluster of 30,000 of its third-generation P800 Kunlun chips for large-scale AI model training.
Impact on Nvidia and the Global Market
The increasing adoption of domestic AI chips by Alibaba and Baidu is expected to have significant implications for Nvidia, which has historically dominated China's AI accelerator market. Nvidia's revenue from China has already decreased by more than half over the past year. Despite the growing capabilities of domestic chips, both Alibaba and Baidu are expected to continue utilizing Nvidia's technology for their most advanced AI systems, balancing innovation with access to proven solutions. An Nvidia spokesperson stated that the company will continue focusing on earning the trust of developers worldwide. Nvidia is also in discussions with the U.S. government to allow limited chip sales to China.
Broader Implications and Future Outlook
The move by Alibaba and Baidu to leverage in-house chips for AI model training reflects a broader trend in China's technology industry towards self-sufficiency and reducing reliance on foreign technology. This trend is likely to continue as China invests heavily in its domestic semiconductor industry and seeks to become a global leader in AI. The development and deployment of these in-house chips also signify a growing competition in the AI chip market, potentially leading to further innovation and diversification in the industry.
