Chinese technology giants Tencent and Baidu have revealed their strategies for maintaining competitiveness in the global artificial intelligence (AI) race despite the United States tightening restrictions on key semiconductors.
Tencent’s president, Martin Lau, reported during the company’s earnings conference call that they have been stockpiling graphics processing units (GPUs), the semiconductors crucial for training large AI models. Lau stated that Tencent has a “pretty strong stockpile” of GPUs previously purchased, which will be sufficient for training models for “a few more generations going forward.”
To optimize their AI capabilities, Tencent is focusing on making their AI models more efficient rather than simply expanding their GPU clusters. Lau explained that the company is achieving good training results with a smaller group of GPUs by implementing “software optimization” to improve efficiency. This approach allows them to deploy the same amount of GPUs to execute specific functions more effectively.
Furthermore, Tencent is exploring the use of smaller AI models that require less computing power and investigating the potential of custom-designed chips and semiconductors available within China. Lau emphasized the need to explore various avenues and invest more time in software development rather than relying solely on purchasing additional GPUs.
Baidu, China’s leading search engine company, highlighted its “full-stack” AI capabilities during its earnings call. Dou Shen, president of Baidu’s AI cloud business, stated that the company’s combination of cloud computing infrastructure, AI models, and applications enables it to build robust applications and deliver significant value even without access to the most advanced chips.
Baidu is also focusing on software optimization to reduce the cost of running its AI models. Shen noted that owning much of the technology in their stack allows for greater efficiency. The company is leveraging its ability to effectively utilize its existing GPUs and manage large-scale GPU clusters as a competitive advantage.
Shen further emphasized the progress made by Chinese technology firms in developing AI semiconductors, which will help mitigate the impact of US chip restrictions. He stated that domestically developed chips, combined with an increasingly efficient homegrown software stack, will form a strong foundation for long-term innovation in China’s AI ecosystem.
The US has been tightening export restrictions on semiconductors, particularly those from Nvidia and AMD, used for AI applications. While the Biden administration scrapped one controversial Trump-era chip rule, it maintained and expanded other restrictions in April.
Industry analysts note that while China remains behind the US in GPU and AI chip technology, it has been making consistent progress in developing its domestic semiconductor ecosystem. Gaurav Gupta, a semiconductor analyst at Gartner, observed that Chinese companies are dealing with export restrictions by stockpiling chips and advancing their semiconductor technology.
As the US-China tech competition intensifies, major US executives have been urging the government to reconsider export restrictions. Nvidia CEO Jensen Huang recently described the curbs as a “failure,” arguing that they harm American businesses more than Chinese ones.
