.png){kind=logo}
Baidu CEO weighs in on the latest DeepSeek debate, claiming that although cheaper models like DeepSeek AI have risen recently, continued investment in AI infrastructure is still needed. Speaking at the World Government Summit in Dubai, Robin Li emphasized that cloud computing and data center expansion should be a priority.
“The investment in cloud infrastructure is still very much required. To come up with models that are smarter than everyone else, you have to use more computing,” said Li while sharing his thoughts with the attendees.
Chinese AI startup DeepSeek has gained attention for developing a language model that matches OpenAI’s GPT. The Chinese AI model has been made using less computing power. This has sparked debate over large AI firms like Baidu’s need to maintain high infrastructure costs.
Baidu continues to invest heavily in AI despite the cost issues. The company was among the first in China to release AI models following OpenAI’s ChatGPT debut in 2022. The company launched Ernie 4, claiming that it matches GPT-4’s capabilities. However, the model has struggled to gain widespread user adoption.
Li recognized that DeepSeek’s rapid rise highlights the unpredictable nature of AI breakthroughs. “You just don’t know when and where innovations come from,” he admitted. He also noted that compute constraints in China have forced AI companies to innovate in cost-effective ways.
Li, who previously advocated for closed-source AI models, now sees merit in open-source approaches. “If you open things up, a lot of people will be curious enough to try it. This will help spread the technology much faster,” he said.
Baidu is set to launch Ernie 5 in the second half of 2025. According to sources, the new model will include multimodal capabilities, enabling it to process text, video, images, and audio. However, Baidu faces stiff competition from ByteDance’s Doubao chatbot and DeepSeek, which have gained stronger user traction in China.
As competition intensifies in China's AI sector, Baidu remains committed to scaling computing power to maintain an edge.
