The statement that "Baidu cannot merely be a training ground for AI talent" has become a central point of discussion.

In May of this year, Baidu founder Robin Li announced a new metric for the AI industry at the Create 2026 Baidu AI Developer Conference: DAA, or Daily Active Agents.

He explained that tokens represent cost, not necessarily the final outcome or revenue. To gauge the prosperity of a platform and ecosystem, one should look at DAA—how many AI agents are actively working for humans and delivering results.

Reactions from developers and investors at the event were mixed. Some viewed it as Baidu's new ticket to the future, while others believed a new concept alone was insufficient to define the company's path forward. Behind this debate lies an undeniable fact reflected in the financials: Baidu's two growth trajectories are converging.

On May 17th, Baidu reported its Q1 2026 results. General business revenue reached 26 billion yuan, a 2% year-over-year increase. Within this, AI business revenue was 13.6 billion yuan, accounting for 52% of the general business revenue and marking its first time exceeding the halfway point, following several quarters of growth. In contrast, traditional online marketing revenue was 12.6 billion yuan, a 22% year-over-year decline.

This turning point comes at a cost. During the period, Baidu's net profit attributable to shareholders was 3.4 billion yuan, a drop of over 55% compared to the previous year.

A senior executive at an AI company expressed puzzlement over Baidu's perceived lag in the large language model wave. Baidu was an early investor in AI, with many leaders at ByteDance's Seed and Alibaba's Tongyi Qianwen hailing from Baidu. Baidu alumni also constitute a significant portion of leadership in autonomous driving companies. Search and AIGC have natural synergies in consumer applications, yet Baidu has not secured a spot in the top tier.

QuestMobile data shows that as of March 2026, monthly active users of AI-native apps in China reached 440 million. ByteDance's Doubao, Alibaba's Qianwen, and DeepSeek led with 345 million, 166 million, and 127 million MAUs respectively, occupying the top three spots. Baidu's "Wenxin" app, however, has fallen out of the top ten.

On one side is a structural breakthrough with AI taking the lead; on the other is short-term profit pressure. Baidu needs a decisive moment to shift gears: using profits from the old track to pave the way for growth on the new one.

It is understood that this year, Baidu has undertaken a series of product and organizational changes within its Mobile Ecosystem Group (MEG), clearly focusing on user-centric product integration. The standalone Wenxin app is being more tightly linked with internal entry points like Baidu Search and Baidu Wenku.

Correspondingly, Baidu has also made a series of organizational and personnel adjustments. The test for Baidu now extends beyond technology to the strategic resolve and organizational resilience of a tech giant.

Challenges and Growing Pains

For over two decades, search has been Baidu's foundation. However, in the AI era, the way users access information has been reshaped. The shift from "search-click-read" to "ask-AI gives direct answer" seems similar to search. The real challenge lies in Baidu's more "conservative" strategic choices despite its AI technical accumulation, and the internal conflict between AI applications and its "pay-for-ranking" cash cow business model.

In late April, an anonymous Baidu employee admitted that for daily research, they "use Doubao and DeepSeek more often." To this day, "Baidu's capability in consumer productization is not on the same level as ByteDance's."

A lead researcher for a vertical large model at a major company analyzed from an industry perspective: "Robin Li was clear from the start about developing large models: no open-sourcing. This choice has its pros and cons."

Open-source versus closed-source fundamentally reflects different companies' judgments on technological paths and business models. In the view of the aforementioned source, open-source strategy has been key to the rapid iteration of the AI industry over the past two years, while Baidu initially chose a more closed path. Closed-source can protect core technology and build differentiated barriers, but it does come at the cost of ecosystem expansion speed.

Deeper pressure lies in organizational and ecological synergy. An industry insider close to Baidu noted that Baidu's internal product planning is overly complex, with a plethora of products like Wenxin App, Wen Xiaoyan, and Wenxin Yuge, whose names are confusing. Each department seems to have its own roadmap, often leading to internal competition, severe product homogenization, dispersed resources, and a fragmented user experience.

Within ByteDance's ecosystem, Doubao, Douyin, and Jimeng have formed tight synergy from tools to platforms.

For instance, if one wants to create an image with AI, many examples can be found on Douyin. Users can also @Doubao in comments for feedback. It truly operates on the principle of "from the users, to the users." The source added that this year, they clearly feel Doubao is evolving from a chat tool into a search-like entry point, building user stickiness. "Sometimes it feels a bit clumsy, but it's fun to use, so it's acceptable."

In contrast to the consumer side, Baidu AI Cloud performs relatively well in the enterprise (B2B) sector.

According to third-party reports, in Q1 2026, Baidu held a 40.4% share in the self-developed GPU cloud market in China, ranking first. In China's AI application public cloud service market, its share rose to 30.7%. A significant portion of AI demand from automakers, banks, and state-owned enterprises lands on Baidu's platform. Particularly in automotive and finance, clients highly recognize Baidu's technical implementation capabilities.

However, within the 13.6 billion yuan AI revenue, intelligent cloud contributed nearly 65%, while application-side revenue saw a decline. This indicates Baidu's current positioning in the AI industry chain is more that of an infrastructure provider rather than a "gold digger." From another perspective, this is also Baidu's strategic choice to "build the road first, then run the cars": solidify the computing power foundation before naturally extending to the application layer.

The aforementioned lead researcher emphasized that industry competition is shifting from competing on computing power to competing on data engineering. The companies performing well in large models are those doing more solid work at the code level, which fundamentally comes down to investment—how much is invested in patents and how much effort is devoted to data processing.

In terms of technical accumulation, Baidu is not behind. Public data shows that by the end of 2024, Baidu had publicly filed over 27,000 AI patent applications globally, covering the full stack of AI fields including deep learning, natural language processing, and computer vision.

The development of its self-developed "Kunlun Chip" also demonstrates Baidu's commitment to long-term investment. The third-generation Kunlun Chip is stably operating on ten-thousand-card clusters, providing an autonomous and controllable computing foundation for the training and inference of the Wenxin large model, which will also optimize Baidu's cost advantage in the long run.

However, the upcoming test will be whether high growth in enterprise orders can sustain the entire narrative of Baidu AI.

Organizational Restructuring for Faster Decisions

External market changes have also catalyzed a significant reorganization within Baidu.

According to media reports, by the end of 2025, Baidu initiated its "largest-ever" adjustment, reducing total employee count from 41,300 in 2022 to 35,900. Positions in AI Cloud and intelligent driving were prioritized, with resources further倾斜 toward AI.

Entering 2026, top-down organizational and business restructuring continues to accelerate. In January, Baidu Wenku and Wangpan were merged into the Personal Super Intelligence Group (PSIG), with Group Vice President Wang Ying reporting directly to Robin Li. In late April, Baidu announced the elimination of its long-used alphabetical job level labels, unifying them into a numerical system aimed at "breaking down barriers between professional and management tracks."

According to incomplete statistics, over the past year, several core executives have departed, including Vice President Zhao Shiqi, Search General Manager, and Chen Yingmei, head of Search AI products and Robin Li's first management trainee.

Most notably, the newly established Model Committee (BMC) in May this year drew attention. Previously, in November 2025, Baidu had already set up two parallel departments: the Basic Model Unit (BMU) and the Applied Model Unit (AMU), both under Robin Li's direct oversight.

According to public information, the BMC consists of young researchers with a deep understanding of large models. It will coordinate the R&D work of Baidu's basic and applied large models, reporting directly to Robin Li. The BMC adds a coordination layer above the BMU and AMU, marking the formalization of this organizational structure established last year.

These moves point to the same goal: shortening the decision-making chain, allowing frontline technical judgments to reach the top decision-makers faster.

The aforementioned lead researcher expressed approval for this design: "Ultimately, all problems come back to 'people.' By having young people report directly to him, Robin Li hopes to find capable and motivated individuals to do something different. He is under great pressure; Baidu cannot just be a training academy for AI talent."

Compared to startups, the most difficult balance for established giants in innovation often lies in the "historical baggage" at the organizational level.

A former executive with over a decade of team-leading experience at a major company, now at a large model firm, admitted that companies performing well in large models generally have younger employees. Retaining them requires moving away from traditional, step-by-step, assembly-line management styles. The recent departure of Lin Junyan from Alibaba was partly due to an inability to tolerate such management models.

Alibaba's "Damo Academy" also faced similar structural dilemmas, eventually dispersing "like stars across the sky." Its early positioning was innovation, but later, influenced by cost and profitability, it faced requirements for self-sufficiency. Ultimately, commercial logic drove the outcome.

Therefore, the final driving force and results still point to the determination of the top decision-maker to make tough choices.

"The dual pressures of business targets and leadership responsibilities create too many uncertainties. Sometimes you have to control innovation; with uncertain timelines and business goals pushing backwards, technical solutions can only be tightened. In the end, the organization can only focus on what the top boss cares about most," the source said.

AI Agents: Baidu's Next Play

At the Create 2026 conference, Robin Li launched four AI agent products in one go: the general-purpose agent DuMate, the coding agent Miaoda 3.0, the digital human agent Baidu Yijing, and the decision-making agent Famou 2.0.

Among them, DuMate has achieved State-of-the-Art (SOTA) levels in multiple international authoritative Agent Benchmark evaluations, capable of operating software, processing files, and connecting business systems. Following the release of Miaoda 3.0, its App and enterprise version were simultaneously launched, allowing users to generate applications through natural language descriptions. Baidu Yijing is positioned as the world's first full-scenario, multi-agent digital human platform.

While introducing the DAA concept, Robin Li emphasized the importance of "applications." He believes the AI industry is shifting from a "model-centric" to an "application-centric" approach, with intelligent agents becoming the new entry point.

From establishing the Institute of Deep Learning in 2013, to proclaiming "All in AI" in 2017, to the comprehensive deployment of large models and agents today, Robin Li's commitment to AI investment spans over a decade, unwavering. Such long-termism is rare among Chinese internet entrepreneurs.

"The key to how far AI can develop lies in whether the boss clearly sees the product's upper limit and dares to invest. I believe in many Chinese companies, many executives haven't seen it clearly yet," the lead researcher said. Robin Li's willingness to continuously bet on the agent direction reflects his judgment on the technological endgame. "But compared to US companies, the pressure on the Chinese team remains immense."

However, Baidu still has time. This year, market education for the agent ecosystem has just been completed. The path to commercialization is shifting from consumer token consumption to coding/development付费, with answers still being explored.

"Truly high-frequency commercial growth points haven't been found by anyone yet. Baidu's pressure lies in others抢先 occupying entry points," an employee with many years at Baidu stated. Just as ByteDance secured the Douyin entry point, Doubao gained continuous iterative data and user feedback from it.

Baidu' opportunity lies more in combining its search entry point with agent products. For example, embedding DuMate directly into the Baidu App, covering hundreds of millions of daily active users. Furthermore, Baidu possesses two decades of积累 in Chinese data processing, search logs, and knowledge graphs, which is still sufficient to build unique competitive barriers.

"Qianwen has done the best in open-source, with very fine data granularity, able to拆解 items like PDFs and web pages in detail. Zhipu's underlying data work is also good," the lead researcher pointed out. "The core is still about strategic choices; mere distillation肯定 offers limited value." With Alibaba adjusting its open-source strategy and Doubao preparing to implement付费 plans, the degree of openness and commercialization strategies of domestic large model companies are also on the agenda.

But judging from the financial reports, the inflection point in Baidu's revenue structure has already emerged, and the time窗口 is becoming具体而迫切. Baidu needs to clarify not just its technological and business direction, but also what kind of organization, culture, and利益 distribution structure will execute this strategic direction.

In the long marathon of AI, how will Baidu realize its long-held belief in technology? As Robin Li stated in his Create conference speech: "We are not conducting an experiment; we are paving a road."

But for Baidu standing at the crossroads, the most difficult pass has never been in the laboratory.