Baidu closed 2025 with a bruising set of results that underscore how perilous its pivot to artificial intelligence has become. Full-year revenue slipped to ¥129.1 billion, down 3% year-on-year, while attributable net profit plunged 76% to ¥5.6 billion after a one-off ¥16.2 billion impairment tied to outdated compute infrastructure. Management framed the write-down as a necessary purge to refocus the company on frontier AI capacity, but the move exposed how quickly the old cash engine — search advertising — is eroding.
The company’s once-reliable online marketing business weakened sharply throughout the year, with sequentially worsening declines in advertising revenue (from -6% in Q1 to -18% in Q3). Baidu even declined to disclose a standalone figure for Q4 online marketing revenue — the first such omission in its listed history — signaling how sensitive the ad trajectory has become to the structural shifts in user behaviour.
Search itself is under structural assault. Increasingly, Chinese users turn to social platforms and large-language-model-powered services for answers rather than classic blue-link results. Market researcher QuestMobile ranks Baidu’s media influence at ninth place as of June 2025, trailing the likes of Douyin, Taobao, WeChat, Kuaishou and Xiaohongshu, a stark illustration of how traffic and attention have migrated.
Baidu’s response has been to “all in” on AI. Founder Robin Li has publicly framed search as the company’s first product to be rebuilt around generative AI: by October 2025 roughly 70% of Baidu’s mobile search result pages included AI-generated content, offering users rich-media answers but also reducing the inventory and visibility for traditional search ads. Analysts warned that an increasing share of AI-generated results will compress search ad exposure and therefore hard-to-replace revenue.
The AI business is growing but has not yet filled the gap. Baidu’s AI-new-business segment generated ¥40.0 billion in 2025, up 48% year-on-year, and accounted for about 43% of core revenue in Q4. Still, the fourth-quarter breakdown — approximately ¥5.8 billion from cloud infrastructure, ¥2.7 billion from AI applications and ¥2.7 billion from AI-native marketing — shows that consumer-facing AI applications are not yet a reliable “household earner.” Meanwhile a flood of rivals, from ByteDance and Alibaba to niche AI app makers like Qianwen, Doubao, Kimi and Yuanbao, are aggressively contesting user attention.
Investors reacted coolly. Baidu’s American depositary receipts tumbled as much as 7.1% on the day results were released, though the stock retraced some losses in subsequent trading; market capitalisation stood at about $42.8 billion at the end of February. There were, however, glimmers of operational improvement: Q4 revenue rose to ¥32.74 billion and the quarter returned to profitability, with operating and free cash flow both turning positive (free cash flow of ¥637 million and H2 operating cash flow of ¥3.9 billion).
The single largest drag on 2025’s results was a decisive asset clearance. In Q3 Baidu took a ¥16.2 billion impairment on legacy infrastructure judged unable to meet modern AI compute efficiency, producing a quarterly loss of ¥11.2 billion and sharply depressing annual profit. Management presented the charge as a strategic reset: shedding obsolete assets to “lighten the load” and reallocate capital to AI-native compute.
Cost rationalisation extended to staff. Baidu’s core business headcount fell to about 29,000 by year-end, a reduction of roughly 3,100 employees since Q3, with severance payments of about ¥700 million. At the same time, the company disclosed that cumulative AI investment since 2023 exceeded ¥100 billion and pledged to sustain that investment intensity even as it trimmed research-and-development expenditure — R&D spending declined 8% in 2025 to ¥20.4 billion.
The strategic stakes are high. Robin Li has tightened his grip on AI development, reorganising the technical platform to create two model research divisions that report directly to him and signalling that he will personally steer the company’s AI push. Yet Baidu still does not disclose key profitability metrics for its AI units — such as gross or net margins — leaving investors unsure whether the new growth engines can ever match the cash generation of the old search-ad business. The next year will be a test of whether a major incumbent can trade legacy scale for AI relevance without surrendering its commercial foundation.