Baidu's Total Revenue Reaches 129.1 Billion Yuan in 2025, Full-Stack Capabilities Complete Commercial Model Loop

On February 26, Baidu released its 2025 financial report, achieving total revenue of 129.1 billion yuan that year. In the report, Baidu disclosed for the first time that AI-driven business revenue reached 40 billion yuan, a 48% year-over-year increase. This indicates that under its full-stack AI capabilities, AI is shifting from a “cost center” to a “profit center.”

Substantial Growth in AI Business

From Cost Center to Profit Center

As a leading domestic company with full-stack self-developed AI “core/cloud/model/body,” Baidu controls the entire chain from underlying chips to top-level applications, forming a core competitive barrier.

Among them, the computing power layer’s self-developed Kunlun chip architecture is specially designed for AI, capable of providing large-scale, stable, high-performance AI computing, compatible with mainstream models and frameworks, enabling enterprises to deploy faster and at lower integration costs. Baidu Intelligent Cloud, as a full-stack integrated platform bridging computing power and applications, connects the “computing power-model-application” end-to-end chain, offering private deployment, hybrid cloud intelligent computing, and other one-stop services, serving as the core engine of AI commercialization. The Qianfan large model platform is an enterprise-level AI development and operation base, and the core of Baidu Intelligent Cloud’s Agent Infra. PaddlePaddle, a leading open-source deep learning framework in China, covers scenarios in education, industry, healthcare, supporting the entire process of large model training and inference. Wenxin 5.0, with 2.4 trillion parameters, can perform native multimodal (text/image/audio/video/3D) unified modeling, ranking first domestically in the latest LMArena rankings. Famoso, based on Wenxin large models, is a self-evolving super-intelligent agent capable of autonomously finding optimal solutions, providing core support for applications, and serving as a benchmark product of Baidu’s intelligent agent technology. On the application layer, from B-end to C-end to Robotaxi, Baidu realizes deployment and monetization through intelligent agents and industry solutions.

Since the second half of 2025, the narrative around AI industry in capital markets has fundamentally changed, shifting focus from “computing power arms race” to “commercialization realization + full-stack deployment.” Capital markets are moving from attention on financing and investment to revenue and paid conversion, seeking a rebalancing between investment and income, indicating a return to mature business models for the AI industry. To control computing costs and expand application revenue, AI giants are extending into upstream and downstream of the industry chain.

Against this backdrop, Baidu’s full-stack capabilities provide a significant competitive advantage. From the massive data accumulated through search (603138), to integrated hardware-software vertical integration and full-stack self-development, Baidu possesses a unique full-chain autonomous control technology domestically. The financial report shows that in 2025, AI business revenue accounted for 39% of Baidu’s general business income. Driven by full-stack capabilities, AI is generating revenue across multiple businesses, mainly from enterprise and user paid services. The “technology-business” closed loop has been validated, and AI is transforming from a “cost center” to a “profit center.”

The disclosure of AI business revenue share indicates that Baidu’s AI capabilities, which were previously more “investment-oriented,” are now a sustainable revenue segment. This reflects a clearer AI strategy, with substantial growth in AI-related businesses and a continued solidification of its core business status.

AI Reshaping Baidu’s Growth Model

Bringing High-Quality, Sustainable Revenue

From B-end to C-end to Robotaxi, Baidu leverages AI capabilities across scenarios, using technological investment to drive multi-scenario revenue growth and continuous cost reduction.

To address the difficulty of AI implementation on the B-end, Baidu uses full-stack capabilities. In 2025, Baidu AI Cloud’s revenue grew 34% year-over-year.

In November 2025, Baidu Intelligent Cloud launched new Kunlun chips and the Tianchi super-node, announcing that Kunlun chips will be launched annually for the next five years, combined with Baidu Bacto AI computing platform, to continuously build more powerful AI infrastructure, solving domestic enterprise computing supply issues. Kunlun chips, with Wenxin large models and PaddlePaddle deep soft/hardware synergy, outperform general-purpose chips in inference efficiency and cost. They have been deployed at scale domestically and are part of a self-developed 30,000-card cluster built on Baidu Intelligent Cloud. Commercially, Baidu has secured a billion-yuan order from China Mobile (600941) and achieved large-scale deployment with hundreds of enterprises in banking, power grids, automotive industries, etc.

The full-stack solution lowers the threshold for enterprises to adopt AI. Through the Qianfan platform, models, toolchains, and deployment capabilities are packaged together, allowing enterprises to go live with ready-made capabilities. This advantage helped Baidu Intelligent Cloud win both the highest number and largest amount of contracts in 2025. On January 14, 2026, the international consulting firm Strategy Analytics released the “2025 China GPU Cloud Market Research Report,” showing that Baidu Intelligent Cloud holds 40.4% of the GPU cloud market share in China, leading in GPU cloud products, computing scale, and key industry deployments.

In November 2025, Baidu officially launched the world-leading evolvable super-intelligent agent Famoso. Over 2,000 enterprises have applied for trials, covering logistics, manufacturing, AI for Service (AI4S), and other fields, helping clients explore optimal solutions globally.

On the C-end, Baidu has achieved more than just MAU growth; users are also paying for real AI productivity. In 2025, AI application revenue exceeded 10 billion yuan.

Assist, launched by Baidu App, is an AI smart assistant that relies on models and Baidu Orion AI engine, achieving a deep reconstruction of search and AI, integrating deep thinking, multimodal interaction, and full-scenario services. As of 2021, the number of active users exceeded 200 million, forming three major domestic AI super apps with Baidu Baike and Qianwen. Since the Spring Festival red envelope campaign, Assist’s MAU has quadrupled, with multiple core functions exploding, indicating user shift towards frequent, real-world scenario usage.

Seconds Dada 2.0 reduces development barriers to “just speak to create AI applications,” enabling “full-stack deployment” and “one-stop development and distribution,” helping ordinary users create enterprise-grade applications, making “one-person companies” possible. By December 2025, Seconds Dada reported that over 50,000 commercial applications had been launched across 8 platforms, with a growth rate of over 150%. Half of these applications include backend functions, covering education, commerce, content creation, enterprise services, and more than 200 scenarios, generating over 5 billion yuan in economic and efficiency value. Among Seconds Dada users, 81% are programmers, and the platform has served over 10 million users.

In November 2025, the GenFlow 3.0 intelligent agent was launched across Baidu Library and Baidu Cloud, transforming these into super-intelligent entities providing personalized, autonomous, and universal capabilities. By replacing repetitive work with AI, Library and Cloud become super-individuals capable of expression and versatile operation. GenFlow 3.0 has achieved 20 million active users, demonstrating genuine demand for time-saving subscriptions.

With the releases of Famoso, Seconds Dada, and GenFlow 3.0, Baidu’s full-stack AI applications are accelerating deployment.

While the performance of pure-text large language models (LLMs) continues to improve, doubts about whether this path can reach general artificial intelligence persist. Leading AI scientists Yang Likun and Li Feifei see world models as the key to surpassing pure-text LLMs and approaching human-like intelligence. Autonomous driving is one of the most mature, deeply deployed, and clearly commercialized applications of world models today.

Baidu’s Luobo Kuaipao, as the most direct physical AI deployment, is opening new growth space. In Q4 2025, Luobo Kuaipao’s travel service trips reached 3.4 million, a year-over-year increase of over 200%, with weekly trips reaching 300,000.

The global deployment of this business is accelerating. By February 2026, it had provided over 20 million trips worldwide, covering 26 cities.

In November 2025, Luobo Kuaipao launched full commercial operation in Abu Dhabi, planning to build the largest local fleet by 2026; in December 2025, Luobo Kuaipao partnered with Uber and Lyft to begin driver testing and services in London in 2026, entering another right-hand-drive market outside China; on January 16, 2026, Luobo Kuaipao became the first company approved for full testing in Dubai.

During the 2026 World Government Summit (WGS), Dubai Crown Prince Sheikh Hamdan rode Luobo Kuaipao to attend the event, completing the “Crown Prince’s first ride.” After experiencing Luobo Kuaipao, the Crown Prince posted on overseas social media, “Free your hands, the future is here!” His praise for this trip became a landmark event in China’s autonomous driving globalization process.

Major Shareholder Return Initiatives Announced

Kunlun Chip Spin-off Raises Target Price

As AI revenue continues to grow and the “technology-business” closed loop is realized, Baidu also announced major shareholder return measures.

On February 5, 2026, Baidu issued announcements on stock repurchase and the first dividend distribution, further enhancing shareholder value. First, from now until December 31, 2028, Baidu plans to repurchase up to $5 billion of its common stock. Second, the board is expected to approve and adopt a dividend policy, including regular and special dividends, with the first dividend expected to be announced and paid in 2026.

Compared to cash-rich companies like Apple and Google, which typically repurchase 2-4% of their market cap, and Meta at around 5%, Baidu’s repurchase of 10% of its market value exceeds expectations. This unconventional move suggests the company believes it is undervalued, implying potential for stock price recovery.

Historically, many tech giants have entered long-term bull markets after initiating dividends and buybacks. For example, after Apple restarted dividends in 2012, it paid over $600 billion in dividends over ten years, with a 700% stock price increase, proving stable dividends can support long-term value. Microsoft, after its first dividend in 2003, combined dividends and buybacks to grow its market cap from $250 billion to $3 trillion, becoming a benchmark for tech stock revaluation.

Baidu’s “combo return” strategy signals to the market that the company maintains healthy cash flow and that its AI commercialization is on track, showing confidence in higher AI revenue in the future and embodying a balanced growth and shareholder return philosophy.

In January 2026, Baidu announced a proposal to spin off Kunlun Chip for listing. The proposal states that this aligns with the company’s and shareholders’ overall interests, highlighting Kunlun Chip’s independent value, attracting vertical industry investors, and enabling long-term development for both parties.

Under the narrative that “computing power is the infrastructure of the AI industry,” Kunlun Chip, as a leading domestic computing power enterprise, is expected to unlock significant value through an independent listing.

Following the announcement, dozens of institutions raised Baidu’s target price based on segment valuation.

Benchmark raised Baidu’s target price by $215, believing that these catalysts will help unlock undervalued assets and facilitate a systematic valuation reassessment via segment valuation methods (SoTP). Benchmark considers Baidu one of the most promising stocks for exceeding expectations in fiscal year 2026.

JPMorgan Chase also increased Baidu’s target price by $200 using segment valuation, noting that by Q4 2025, Baidu’s AI-driven transformation is becoming clearer. Early signs show increased AI user penetration, integration of AI products with existing mobile internet services, and improved operational efficiency. Baidu is beginning to convert AI investments into more sustainable growth drivers.