On the evening of November 16,
At the same time as the earnings disclosure, Baidu continued to deepen synergy between underlying computing power and upper-layer applications. Within AI cloud infrastructure, GPU cloud, Kunlun chips, and MaaS contributed more than half of revenue, while Qianfan MaaS revenue from external customers' token invocations surged 1,159% year over year. To address the trust challenge enterprises face in continuously using AI, Baidu officially launched the "Token Exchange" model supply-and-demand marketplace.
During its trial operation, the platform listed more than 40 mainstream models, including ERNIE, DeepSeek, Kimi, Zhipu GLM, and MiniMax, brought together more than 100 onboarded supplier resources, and exceeded 50 million average daily invocations. On the first day of its official launch, more than 100 new suppliers joined and more than 1,000 new customers were onboarded. Meanwhile, Baidu's internal daily active agents (DAA) grew from 5,000 to 37,000, weekly completed valid tasks exceeded 1 million, the AI automated audit execution rate in internal audit processes reached 40%, and through agent tools such as "Famou" and "Dazi," AI capabilities were deeply embedded into real business scenarios such as industrial manufacturing and sales operations.
The fact that Baidu's AI business revenue has accounted for more than half for consecutive quarters, together with the rapid growth of its infrastructure, indicates that it is advancing deeper from computing power procurement to the model ecosystem and transformation of business processes. By building a trusted industrial supply network through platforms such as the "Token Exchange" and promoting the deep deployment of agents in enterprise operations scenarios, Baidu is accelerating its transformation into an intelligent organization, providing important infrastructure and organizational capability support for the long-term, large-scale operation of AI technology in industry.
Join Now