Chinese technology companies Tencent and Baidu have disclosed their strategies for remaining competitive in the global artificial intelligence (AI) race despite the United States tightening restrictions on key semiconductors.
The companies’ approaches include stockpiling chips, enhancing the efficiency of their AI models, and utilizing domestically developed semiconductors. The U.S. administration, under President Joe Biden, has continued to restrict China’s access to advanced chips from companies such as Nvidia and AMD, which are crucial for AI development.
Tencent’s Strategy
Tencent’s president, Martin Lau, revealed during the company’s earnings conference call that they have a “pretty strong stockpile” of graphics processing units (GPUs), the type of semiconductor that has become essential for training large AI models. Lau explained that contrary to the belief that GPU clusters need to be expanded to create more advanced AI, Tencent can achieve satisfactory training results with a smaller group of such chips.
“That actually sort of helped us to look at our existing inventory of high-end chips and say, we should have enough high-end chips to continue our training of models for a few more generations going forward,” Lau stated. For inferencing – the process of executing AI tasks – Tencent is employing “software optimization” to enhance efficiency, allowing them to deploy the same number of GPUs for specific functions. The company is also exploring the use of smaller models that require less computing power and is investigating the potential of custom-designed chips available in China.
Baidu’s Approach
Baidu, China’s largest search engine company, highlighted its “full-stack” capabilities, which combine cloud computing infrastructure, AI models, and applications based on these models, such as its ERNIE chatbot. Dou Shen, president of Baidu’s AI cloud business, emphasized that even without access to the most advanced chips, their unique full-stack AI capabilities enable them to build robust applications and deliver significant value.
Baidu also focused on software optimization and reducing the cost of running its models, as it owns much of the technology in its stack. The company’s management discussed efficiencies that allow them to maximize the utility of their GPUs. Shen noted that the ability to build and manage large-scale GPU clusters and utilize GPUs effectively has become a key competitive advantage.
China’s Domestic Chip Development
China has been intensifying its development of domestically designed and manufactured chips over the past few years. While experts agree that China remains behind the U.S. in GPU and AI chip technology, there have been notable advancements. Gaurav Gupta, an analyst covering semiconductors at Gartner, observed that stockpiling is one strategy Chinese companies are using to cope with export restrictions. Additionally, there has been progress in China’s semiconductor technology, although it still lags behind the U.S.
“China has also been developing its own domestic semiconductor ecosystem, from materials to equipment to chips and packaging. Different segments have made varying levels of progress, but China has been surprisingly consistent and ambitious in this goal, and one must admit that they have achieved decent success,” Gupta commented.
The development of China’s domestic chip industry is seen as crucial for mitigating the impact of U.S. chip curbs. As Dou Shen stated, “Domestically developed self-sufficient chips, along with an increasingly efficient home-grown software stack, will jointly form a strong foundation for long-term innovation in China’s AI ecosystem.”
