Redefining three major business boundaries.

Author丨Wei Bing

Editor丨Zhao Lei

LatePost has learned that Baidu Intelligent Cloud recently completed a round of organizational adjustments in its product research and development. The original Platform Product Division has been renamed as the Intelligent Body Division, under which the MaaS business, Qianfan brand, and products related to MaaS have been transferred to the Infrastructure Division. The Data Platform Department has been renamed as the Data Intelligence Department, which has moved to the newly renamed Intelligent Application Division originating from the original Application Product Division.

After the adjustment, the three divisions of Baidu Intelligent Cloud will be led by Hou Zhenyu, Yin Dawei, and Ruan Yu respectively: Hou Zhenyu will be responsible for the Infrastructure Division, managing AI Infra and Agent Infra; Yin Dawei will be responsible for the Intelligent Body Division, focusing on Agent products such as Baidu Douzi, Miaoda, and Famou; Ruan Yu will be responsible for the Intelligent Application Division, managing Baidu Yijian, Hogee, Shengsu, as well as applications in industries like transportation and finance.

All three will report directly to Shen Dou, Executive Vice President of Baidu Group and President of the Intelligent Cloud Business Group.

Sources close to Baidu told LatePost that this adjustment has been brewing since the second quarter of this year, with a plan formed in the third quarter and officially implemented in August. The MaaS and data platform teams will migrate with the structure, and their reporting relationships will change simultaneously; the reorganized Intelligent Body Division will also receive incremental resources.

The core of this adjustment is to separate MaaS, data, and general intelligent body that were originally placed in the same division. According to the aforementioned sources, the adjustment aims to clarify the strategic boundaries of several business segments: "To scale mature businesses (Infra), and to make emerging businesses (Agent, Intelligent Applications) agile." Overall, teams directly related to model inference, hosting, and distribution will enter the Infrastructure Division, general intelligent body capabilities will belong to Yin Dawei, and industry applications and data capabilities will belong to Ruan Yu.

In the past, MaaS, data platform, and general intelligent body were all under the Platform Product Division. This arrangement had benefited Baidu in quickly establishing a comprehensive product chain from models to applications, helping general intelligent bodies evolve from zero to one. However, as business expanded, these three types of products began to require different resources and management methods, and multiple business lines being under one division resulted in insufficient strategic focus, requiring the products to share resources and management attention.

MaaS is increasingly approaching an infrastructure business. Model invocation requires Qianfan to provide inference, hosting, and other model services, and also relies on computing power, chips, and cloud resources. Operational results depend on resource utilization, inference costs, and client deployment scales. A person familiar with this adjustment said, "MaaS is essentially Infra."

The most noticeable change in this adjustment is to categorize MaaS under the Infrastructure Division. Model services and underlying computing power and cloud resources are placed on the same business line, connecting the chain from Kunlun chip computing scheduling to Qianfan model inference, aiming to reduce cross-team coordination and improve resource utilization and delivery efficiency. In the past, clients might have needed to interface with multiple teams; after the adjustment, these capabilities can be delivered by the same division, reducing costs for large clients who were simultaneously interfacing with computation and model teams.

The Data Platform Department did not enter the Infrastructure Division with MaaS but instead transitioned to the Intelligent Application Division. It is understood that this arrangement aims to bring the data team closer to industry projects, supporting Yijian, Hogee, Shengsu, and applications in transportation, finance, and others to enter clients' business processes, reducing inter-division collaboration between data processing and final business applications. The Intelligent Application Division has listed the number of industry customers and the reuse rate of solutions as key metrics, focusing on implementing FDE-style delivery.

The general intelligent body faces a different set of problems. Baidu Douzi, Miaoda, Famou, and other general intelligent bodies are still in a rapid iteration phase, requiring quicker responses to user demands, and focusing on user retention, task completion rates, and monetization. The Intelligent Body Division will no longer bear the responsibilities of MaaS and data businesses simultaneously, and resources will further concentrate on Agent, increasing strategic priority.

It is understood that after this adjustment, Yin Dawei will be responsible for user growth and product experience of the general intelligent body, exploring commercialization methods including subscription models. The team will focus on DAU, MAU, retention rates, paying customers, and end-to-end task completion rates.

After the adjustment, Hou Zhenyu, Yin Dawei, and Ruan Yu will be responsible at different levels while sharing goals. Intelligent body products will still depend on the Infrastructure Division for providing model inference and computing power, and industry applications will also need to leverage general agent capabilities. Use cases that can be realized may become capability components of the general intelligent body. The resource coordination that originally occurred within the Platform Product Division will, from now on, take place more frequently among the three peer divisions.

Over the past year, several business lines at Baidu Intelligent Cloud have begun to change. By 2025, model services, data, tools, and agents will still be integrated into a set of Qianfan platform systems. In May of this year, Baidu began to emphasize the reconstruction of AI cloud infrastructure with Agent Native, merging AI Infra and Agent Infra. At the same time, the data platform launched "Shengsu", which is also beginning to transition from data management to specific business scenarios.

In the second quarter financial report this year, Baidu AI Cloud Infra generated 7.3 billion yuan in revenue, a year-on-year increase of 50%, with GPU Cloud growing 283%; AI Applications revenue was 2.5 billion yuan, a year-on-year increase of only 3%. The revenue sources and demand signals for infrastructure are relatively clear, but intelligent bodies and industry applications are still in the product iteration phase, working through solutions.

Sources close to Baidu Intelligent Cloud stated that the Infrastructure Division already has clear billing items and growth targets. The next step is to scale up and improve resource efficiency; the Intelligent Application Division mainly generates revenue through industry solutions and data intelligence services, and needs to increase project reuse rates; the Intelligent Body Division will continue to explore commercialization, and as the user scale reaches a certain level, the path to commercialization will gradually become clear.

Agent is one of the most competitive directions among large enterprises this year. Since July, Alibaba has merged QoderWork, Wukong, and MuleRun into Qianwen Office, while Tencent has transferred QClaw-related businesses and teams into the department of WorkBuddy; ByteDance has also further integrated product teams related to Feishu and Doubao.

Baidu's adjustment has similarly raised the organizational priority of the general intelligent body but has not yet consolidated multiple office agent entrances within the group. Baidu Douzi, at the public release on August 27, cited data from the AI product rankings stating that its monthly active users on the PC side were about 6.74 million in July, with a month-on-month growth of 1063.79%.

People close to Baidu indicated that Baidu's current assessment of office agents is that the market is still in a rapid development phase. Apart from scale, there is a greater focus on whether office intelligent bodies can genuinely enter professional groups and enterprises' production processes and deliver results. The teams mainly examine end-to-end task completion rates, while the group level pays more attention to DAA (Daily Active Agents).

An organization can provide a new business with resources, positions, and a shorter decision-making chain. However, it cannot replace the product in answering several fundamental questions: Are users continuously using it? Are they willing to pay? And can revenue cover the machine learning models and computing power costs required to complete complex tasks? For Baidu Intelligent Cloud, after the new organization is implemented, the real test is yet to come.

Cover image source: "Turning Goals into Gold"

- FIN -

免责声明：本文章仅代表作者个人观点，不代表本平台的立场和观点。本文章仅供信息分享，不构成对任何人的任何投资建议。用户与作者之间的任何争议，与本平台无关。如网页中刊载的文章或图片涉及侵权，请提供相关的权利证明和身份证明发送邮件到support@aicoin.com，本平台相关工作人员将会进行核查。