Baidu’s latest earnings update reinforced a message it has been sharpening for several quarters: investors should judge the company less as a legacy search and advertising platform, and more as a deep tech provider whose growth is increasingly defined by cloud infrastructure, artificial intelligence, and autonomous driving.
That shift is more visible in internal revenue mix metrics than in consolidated results. Baidu reported fourth-quarter revenue of RMB 32.7 billion (USD 4.6 billion), up 5% quarter-on-quarter but lower than a year earlier. Full-year 2025 revenue declined 3% to RMB 129.1 billion (USD 18.1 billion), which management attributed to weakness in the legacy business, partially offset by growth in AI-related segments.
AI is moving to the center of Baidu’s business
Baidu used the quarter to introduce a revised disclosure framework it calls “general business,” grouping operations into AI-powered business, legacy business, and other revenue. Under this framework, AI-powered business revenue exceeded RMB 11 billion (USD 1.5 billion) in the fourth quarter, accounting for 43% of general business revenue. Management said this proportion has been rising rapidly.
The reclassification is partly cosmetic, but it provides clarity. Consolidated figures still reflect advertising softness and cyclical pressure that can offset AI momentum. The internal mix, however, suggests AI-driven lines are becoming a material share of the revenue base.
For the full year, AI-powered business revenue reached RMB 40 billion (USD 5.6 billion), up 48% year-on-year (YoY). This included RMB 19.8 billion (USD 2.8 billion) from AI cloud infrastructure, RMB 10.2 billion (USD 1.4 billion) from AI applications, and RMB 9.8 billion (USD 1.4 billion) from AI-native marketing services. The three segments grew 34%, 5%, and 301%, respectively.
Within the AI portfolio, AI cloud infrastructure showed the clearest signs of acceleration. Baidu is advancing a narrative that enterprise adoption is shifting from pilot projects to production deployment, with a greater share of revenue becoming recurring.
The company reported RMB 5.8 billion (USD 812 million) in AI cloud infrastructure revenue in the fourth quarter. Subscription-based revenue from AI accelerator infrastructure grew 143% YoY, accelerating from 128% growth in the previous quarter. Management described the shift toward subscriptions as “structurally healthier,” arguing it reflects deeper enterprise integration rather than one-off project work.
On the earnings call, Baidu sought to differentiate its cloud strategy by stressing its “full-stack” architecture and heterogeneous computing approach, combining domestic and international chips. It positioned its self-developed Kunlun chips as a cost and compatibility advantage for inference workloads, the stage when trained AI models generate outputs. Ultimately, cloud growth will depend not only on demand, but also on Baidu’s ability to supply computing power economically and retain customer workloads as they scale.
Impairments reshape the annual picture
Baidu’s Q4 profitability improved on a non-GAAP basis, even as management signaled deliberate cost controls. Non-GAAP operating income was RMB 3 billion (USD 420 million) in the fourth quarter, representing a 9% operating margin. Operating expenses rose quarter over quarter, driven by higher expected credit losses and a one-time employee severance charge intended to improve efficiency.
The more consequential accounting development appeared in the full-year results. Baidu recorded RMB 16.2 billion (USD 2.3 billion) in impairment of long-lived assets in 2025, contributing to a GAAP operating loss of RMB 5.8 billion for the year. By contrast, non-GAAP operating income was RMB 15 billion (USD 2.1 billion).
In effect, Baidu is asking investors to separate a one-time reset of asset values from its current operating trajectory. The key question is what these impairments imply about the durability of parts of the legacy business, and whether additional resets could follow if the AI transition takes longer than management anticipates.
AI engagement grows, monetization remains gradual
In consumer AI, Baidu reported 679 million monthly active users, or MAUs, for its core app in December, flat YoY. Ernie Assistant, integrated into the main app, reported 202 million MAUs, or roughly 30% of the app’s user base.
Rather than emphasizing user growth alone, management framed its strategy around building task-oriented, reliable AI within the app by integrating service capabilities and vertical tools. That approach aligns with Baidu’s monetization model. If search shifts from displaying links to completing actions, the objective is to protect, and potentially expand, monetizable engagement while maintaining low error rates in sensitive categories such as healthcare.
Management suggested monetization would follow as products mature, but offered limited quantitative guidance on timing or revenue contribution.
RELATED ARTICLE
Baidu continues to position Apollo Go as its most visible “physical AI” application and a long-term value driver. The company reported 3.4 million fully driverless rides in the fourth quarter, with total rides up more than 200% YoY. Cumulative rides surpassed 20 million as of February across 26 cities.
Management also highlighted progress on unit economics, including prior statements about breakeven in Wuhan and plans to reach breakeven in additional cities over time. Investors are likely to focus on whether improvements in per-ride economics can keep pace with geographic expansion, particularly as Apollo Go enters higher-cost international markets with different regulatory and operational requirements.
Alongside operating updates, Baidu outlined steps to return capital and potentially unlock asset value. The company announced a new share repurchase program of up to USD 5 billion through the end of 2028, adopted its first dividend policy with an initial payment potentially by year’s end 2026, and reiterated plans for a proposed Kunlunxin spinoff and separate listing.
With RMB 294.1 billion (USD 41.2 billion) in total cash and investments at the end of 2025, Baidu is signaling it can sustain significant AI investment while returning capital. The company said it has invested more than RMB 100 billion (USD 14 billion) in AI since launching Ernie in 2023 and expects to maintain a similar level of investment intensity.
What to watch in 2026
The quarter strengthens the argument that AI is becoming a larger contributor to Baidu’s growth, particularly through cloud infrastructure subscriptions and AI-native marketing services. The central question is whether these segments can consistently outpace the drag from the legacy advertising base at the consolidated level.
Three issues merit attention:
- First, whether AI cloud infrastructure growth remains durable as enterprise deployments deepen and computing costs fluctuate, especially given Baidu’s emphasis on accelerator subscriptions as a recurring driver.
- Second, whether improvements in Baidu’s consumer AI products translate into clearer monetization within its mobile ecosystem.
- Third, whether Apollo Go can scale while demonstrating sustained progress in unit economics, making the business more than a long-dated option.
For now, Baidu’s message is that its AI transition isn’t speculative. The durability of that thesis will, however, depend on whether its AI revenue compounds quickly enough to outweigh the legacy decline the company is still managing.