Baidu closed 2025 with a stark reminder that the shift from search to artificial intelligence is neither smooth nor cheap. The company reported full-year revenue of RMB 129.1 billion, a 3% decline from 2024, and attributable net profit of RMB 5.6 billion, down 76% year-on-year after a one-off RMB 16.2 billion write-down of long-lived infrastructure assets.
The numbers reflect a painful transition. Baidu’s legacy online-marketing business—the company’s historical cash cow—has been shrinking for successive quarters, with year-on-year declines steepening from 6% in Q1 to 18% in Q3. User behaviour is fragmenting: Chinese consumers increasingly obtain information through social platforms and AI-powered apps rather than traditional search, and market-research firm QuestMobile now ranks Baidu ninth on its media index behind rivals such as Douyin, Taobao and WeChat.
Baidu’s response has been to double down on AI. Since the company declared an “All in AI” strategy in 2023 it has poured more than RMB 100 billion into the effort, reorganised its reporting to present a dedicated AI-new-business segment, and pushed AI-generated content into the search experience—about 70% of mobile search result pages contained AI-generated content by October 2025.
The AI business is growing but not yet filling the hole left by advertising. AI new-business revenue reached RMB 40.0 billion in 2025, up 48% year-on-year, and contributed 43% of core revenue in Q4. But the segment remains heavily weighted toward infrastructure and nascent applications: Q4 infrastructure revenue was RMB 5.8 billion, AI applications RMB 2.7 billion and AI-native marketing RMB 2.7 billion, figures that underscore the early-stage economics and the absence of disclosed margins.
Management made a sharp accounting decision that hit results but clarified the strategy. The RMB 16.2 billion impairment was chiefly a revaluation of legacy compute infrastructure deemed inadequate for modern AI workloads. Baidu said it wrote off outdated capacity to free itself from ongoing depreciation and mobilise capital for frontier compute—a move that produced a single-quarter loss of RMB 11.2 billion and became the primary driver of the annual profit collapse.
Cost control has accompanied the technical pivot. Baidu cut about 3,100 employees from its core business through 2025, reducing headcount to roughly 29,000 and paying RMB 700 million in severance. Paradoxically, research-and-development spending fell 8% to RMB 20.4 billion in 2025 even as the company vows to maintain heavy investment in AI going forward.
Investors have reacted ambivalently. Baidu’s U.S.-listed shares tumbled more than 7% on the night of the results and its market capitalisation stood near USD 42.8 billion at the end of February. Management sought to reassure analysts by changing disclosure formats—grouping activities under a newly named “Baidu general business” structure of AI new business, traditional business and others—and by signalling possible value unlocked through a planned listing of Kunlun, its chip unit.
Competitors are intensifying the fight for AI attention and users. ByteDance, Alibaba and Tencent have accelerated model and consumer-app launches—Seedance 2.0, Doubao 2.0 and Qwen 3.5 among them—while a crop of AI-native consumer apps has climbed the download charts. Baidu’s Wenxin assistant reported 202 million monthly active users in December and a fourfold spike in activity since the Spring Festival promotion drive, but that traction has not translated into dominance of consumer app rankings or into ad-replacement revenue at scale.
For China’s tech ecosystem the episode is consequential. Baidu’s write-down and its pivot encapsulate wider truths about the industry: AI requires fresh capital, new infrastructure and different monetisation models; early leadership in models does not guarantee long-term platform advantage; and competition for users and compute is now a capital-intensive, winner-take-most contest. Whether Baidu’s bet produces a renewed growth engine or becomes an expensive strategic drift will depend on its ability to convert AI engagement into durable, high-margin revenue streams.