Baidu ended 2025 with a stark financial silhouette: revenue slid to ¥1,291bn (about ¥129.1 billion) and net profit plunged 76% year‑on‑year to ¥5.6bn, leaving the company at a strategic inflection point. The plunge was aggravated by a one‑time ¥16.2bn impairment charge taken in Q3 as the company retired legacy infrastructure deemed unsuited to contemporary AI compute needs. Management framed the move as a painful but necessary reset to finance a full tilt into next‑generation AI.
The traditional search‑advertising engine that made Baidu wealthy is shrinking. Online marketing revenues declined sharply through the year, accelerating from a 6% drop in Q1 to an 18% decline by Q3, and the company even omitted a standalone disclosure of Q4 online marketing revenue for the first time since listing. User behaviour is shifting: Baidu has fallen in media rankings and now trails short‑form platforms and e‑commerce apps as a user gateway, reflecting a structural erosion of its historic distribution of attention and ad inventory.
Baidu’s strategic counterpunch is to rebuild itself around AI. Management has reorganised the core business into three pillars — AI new business, traditional business and other — and now reports AI revenue separately. The AI new business generated ¥40.0bn in 2025, up 48% year‑on‑year, and accounted for roughly 43% of core revenue in Q4; yet that growth has not matched the scale or the margin of the legacy advertising engine.
The write‑down and headcount reduction underline the urgency of the pivot. Baidu took a major impairment on ageing data‑centre and compute assets it judged incapable of efficient AI workloads and cut about 3,100 roles in the mobile‑ad organisation, paying roughly ¥700m in severance. Management says the cleansing will lighten depreciation burdens and allow refreshed investment in frontier compute for 2026, but the short‑term hit to profitability and investor confidence was sharp: Baidu’s US‑listed shares tumbled over 7% on the results day.
Competition for AI attention and utility is fiercer than ever. Domestic rivals and start‑ups have flooded the market with consumer‑facing AI apps — from ByteDance’s and Alibaba’s models to a raft of newer tools — and Baidu’s own Wenxin assistant, despite reported growth in monthly active users, has struggled to dominate app store charts. The increasingly AI‑driven search experience, which inserts generative answers and rich media into result pages, also risks cannibalising the ad placements that historically monetised search queries.
Financially the story is mixed. Q4 marked a small recovery: quarterly revenue of ¥327.4bn returned the company to profitability, operating and free cash flow turned positive, and management says AI investments are beginning to generate cash. At the same time R&D spending fell 8% to ¥20.4bn in 2025, a puzzling retrenchment at a moment when product velocity matters, and Baidu has not disclosed gross or net margins for its AI business, leaving questions about the unit economics unanswered.
Founder Li Yanhong has personally reasserted control of model development, restructuring teams and creating direct reporting lines to accelerate decision‑making. Baidu is also positioning a potential spin‑off of its Kunlun chip unit as a value‑unlocking move, emphasising a full‑stack, end‑to‑end AI narrative that ranges from infrastructure to consumer apps. The gamble is clear: Baidu must convert AI research and platform reach into scalable, monetisable products faster than its rivals can outflank it.
What happens next matters beyond one company. Baidu’s transition is a bellwether for how legacy internet incumbents monetise generative AI while managing the destruction of legacy ad inventory and the capital intensity of the compute race. If Baidu can rebuild margin‑bearing enterprise AI and cloud services around its models, it can replace lost ad cashflows; if it fails to keep pace in product iteration or monetisation, the company risks being overtaken at both the consumer and enterprise layer.