Robin Li, CEO of Baidu, emphasized the need for ongoing investment in cloud infrastructure and data centers to develop competitive AI models.

At the World Government Summit in Dubai, he noted that “in order to develop models that outperform others, you have to use more compute,” highlighting the importance of sufficient computational resources for training and operating AI systems.

Li’s comments come after the rise of DeepSeek, a Chinese startup creating AI systems similar to OpenAI’s GPT but with lower computational power requirements.

Despite this, Li underscored that Baidu, one of the first Chinese companies to adopt AI models like Ernie, will continue to invest heavily in AI infrastructure.

He said compute constraints have driven Chinese companies to innovate to reduce costs, while also discussing the benefits of open-source AI models for spreading technology faster.

Update: (February 12, 1:02 p.m. SGT): This article was updated to remove any reference to Li saying that US chip sanctions brought challenges. He referred to computing constraints instead.