It was revealed for the first time that AI business revenue surged by 50%, and Baidu-SW (09888) accelerated value fission under the dual flywheel effect

Buffett, who has been committed to value investing his whole life, made a big decision on the eve of his retirement: to open a position with Google for the first time and include it in Berkshire Hathaway's top ten stocks. This move has sparked huge repercussions in the investment community and is seen as an important sign of the evolution of the value investment paradigm — even the most conservative value groups have begun to face up to and acknowledge the long-term value of AI.

Behind this decision is the continuation of Buffett's investment philosophy in the AI era. Google has built a full-stack AI ecosystem with self-developed TPU chips, Gemini models, and massive application scenarios such as search, YouTube, and cloud services. This wide and deep “moat” is the core element that Buffett values the most in the value investment system.

In addition to Google, Baidu-SW (09888) is the company with the most complete full-stack AI technology layout and the most significant synergy effects in China and the world. It has independent core technology at the four key levels of chip, framework, model, and application, and has achieved large-scale industrialization, and the “moat” is also broad.

The third quarter earnings report is a true picture of Baidu's strong competitiveness and leading commercialization capabilities in the AI industry. On November 18, Baidu released its financial results for the third quarter of 2025. The company revealed two key achievements of its AI strategic layout for the first time in this financial report. First, AI business revenue surged 50% year on year. Among them, AI cloud revenue increased 33% year over year; AI application revenue reached 2.6 billion yuan; and AI native marketing service revenue increased 262% year over year to 2.8 billion yuan.

Second, Baidu announced in a performance conference call that in the 10 quarters since the generative AI wave swept the world in March 2023, its total investment in AI has exceeded 100 billion yuan, and will continue to increase investment in R&D in the future.

What this financial report shows is Baidu, which is more determined about AI. While continuing to increase R&D investment to ensure the company's leading technological advantage in the AI industry, it is also making every effort to promote AI's restructuring of the company's full range of products, making AI the core driving force for the company's performance growth.

But this is just the beginning. As Baidu continues to advance at the two levels of AI technology and AI commercialization, while Baidu continues to lead the development of China's AI industry, its entire ecosystem will also usher in an accelerated fission of intrinsic value driven by the “flywheel effect.” At that time, Baidu's valuation in the capital market is also expected to usher in a new round of large-scale cyclical rise.

Restructuring business lines to drive qualitative changes in revenue structure and move towards a new stage of AI-driven growth

Currently, there is no shortage of voices on the “AI bubble theory” in the market. This is mainly because American AI companies represented by OPNAI continue to invest huge amounts of capital under the “vigorous miracle” strategy, but the output side is relatively limited, causing the market to doubt the sustainability of their business model.

Compared to that, Baidu, which has been deeply involved in the field of AI for more than ten years, has chosen a more pragmatic path. Baidu has long recognized the limitations of a simple “volume model” parameter scale, and instead anchored its core strategy on large-scale application of AI technology and deep integration with the industry.

As an enterprise with a full-stack self-development layout in China and around the world on the four levels of “chip-framework-model-application”, Baidu is the first company in the world to quickly launch Wenxin, after ChatGPT set off a wave, and has taken the lead in fully injecting big model capabilities into the company's business line, making it the most thorough domestic enterprise that uses AI to restructure existing businesses.

On the C-side, search is Baidu's first restructured product, and it is also the most aggressive search engine for AI transformation in the world. In July 2025, Baidu completed the largest revision of the search business in 10 years, completely innovating everything from search boxes and results pages to the entire content ecosystem. The results of this revision are remarkable. As of October 2025, 70% of Baidu search's mobile search results pages already include AI-produced content. Furthermore, on October 15, Baidu Search announced a full upgrade of Wenxin Assistant's AIGC creative capabilities. Up to now, Baidu search users have generated more than 10 million AIGC content per day.

Thanks to strong AI capabilities to restructure search products, Baidu AI search is still in a clear leading position in the industry. According to the “AI Application Industry Report for the Third Quarter of 2025” released by QuestMobile on October 28, as of September 2025, Baidu's monthly active AI search users increased 18.63% quarterly to 382 million, topping the domestic AI search industry monthly activity list for three consecutive quarters. Furthermore, according to the “AI Search Product Evaluation, 2025” released by IDC on July 25, Baidu AI search ranked first in the overall score of general-purpose AI search products in China, with two important indicators of user data and technical ability ranking first.

Meanwhile, during the reporting period, Baidu accelerated the development of AI native marketing services, including smart devices and digital people. These two major services have been recognized by the market due to their significant advantages. More and more customers seeking to use AI to solve problems and are willing to pay for technology have achieved double efficiency improvements in productivity and marketing effectiveness through this service. This led to a 262% year-on-year increase in revenue from AI native marketing services to 2.8 billion yuan during the reporting period.

Baidu Library/Baidu Netdisk is also a C-end product that Baidu is focusing on restructuring. Baidu Library launched the one-stop AI learning and office platform “Oreate” in September of this year. Up to now, “Oreate” has served more than 200 countries and regions around the world, the number of overseas users has exceeded one million, and has attracted widespread attention on overseas social media platforms. The multi-modal AI features launched by Baidu Netdisk in China, such as AI notes and AI cameras, have been loved by many users. Currently, Baidu Netdisk's AI MAU has surpassed 80 million.

Furthermore, at the Baidu World Conference held in November 2025, Baidu Library and Baidu Netdisk jointly released GenFlow 3.0. It has more than 20 million active users, making it one of the world's leading general-purpose smart devices. Currently, GenFlow 3.0 is fully integrated into library and network disk products to help users improve efficiency in their daily work and study and become “super individuals”.

On the B-side level, Baidu released “Baidu Fumou,” an intelligent body that can evolve on its own, on November 13. As the world's first commercially-usable self-evolving superintelligence, it strived to reach the top of MLE-Bench, the authoritative benchmark for OpenAI machine learning engineering, winning SOTA results, and surpassing the AIDE system displayed by Microsoft R&D Agent and OpenAI. The product has already attracted more than 1,000 companies to apply for access testing within a few hours of launch, covering transportation, energy, finance, logistics and other industries. Currently, test places are still open.

The continuous innovation and upgrading of digital employees is also a prominent highlight of Baidu at the B-side commercialization level. At the beginning of August this year, Baidu Smart Cloud released the world's first batch of AI digital employees, covering core business functions in seven scenarios, including marketing and recruitment. Baidu then teamed up with Ashi Education at the 2025 Baidu Yunzhi Conference to jointly develop a new digital employee, “Wu Yanzu Digital English Coach”, and officially released the “Digital Content Operation Employee” at Baidu World 2025, and cooperated with China Eastern Airlines to create a “East East” for digital employees.

At present, Baidu Smart Cloud digital employees have been widely serving more than 10 industries and thousands of enterprises, including finance, consumption, automobiles, cultural tourism, etc., helping enterprises increase efficiency and revenue in multi-dimensional scenarios such as marketing growth, service efficiency improvement, and strategy optimization.

Furthermore, as Baidu's business segment with the most technical and leading edge in the B-side, Radish Express also achieved outstanding performance during the reporting period. According to the data, the number of global travel services during the reporting period of Radish Express reached 3.1 million, an increase of 212% over the previous year, and continued to accelerate compared with the 148% growth rate in the second quarter. Currently, the number of unmanned orders for Radish Express has exceeded 250,000 per week, and the number of travel services worldwide has exceeded 17 million, making it number one in the world.

At the same time, the globalization of Radish Express is once again accelerating. While it reached strategic cooperation with Uber and Lyft within the third quarter to promote large-scale implementation of driverless cars, Radish Express was granted Dubai 001 test permits and 50 autonomous driving test licenses, and has unlocked new developments in Switzerland and Abu Dhabi. It has reached a strategic cooperation with PostBus (PostBus), a subsidiary of Swiss Post, to launch autonomous travel services in Switzerland, and has obtained the first batch of fully unmanned commercial operation licenses in Abu Dhabi.

So far, the footprint of Radish Run has covered 22 cities around the world. The total driverless mileage has exceeded 140 million kilometers, and the total mileage of autonomous driving has exceeded 240 million kilometers, ranking first in the world. According to Frost & Sullivan's forecast, the global Robotaxi market will grow from US$290 million in 2025 to US$66.6 billion in 2030, and reach US$352.6 billion by 2035, and the entire market is showing an explosive growth trend. With its leading first-mover advantage, Radish Express is expected to become the biggest beneficiary of the rapid development of the global driverless industry.

Judging from the results, Baidu's development in the direction of large-scale application of AI technology and deep industry integration is certainly forward-looking. After reconstructing a full range of products with AI capabilities, Baidu's AI business showed strong business vitality and growth potential. Over 30% of its AI business revenue in the third quarter meant that Baidu's revenue structure had qualitatively changed, and it accelerated towards a new stage of high growth driven by AI.

The full-stack self-developed AI layout creates unique competitiveness, and AI cloud services continue to increase

The reason behind Baidu's AI business-driven qualitative changes in revenue structure is not only due to its strategic direction of choosing to deeply integrate AI technology with the industry, but is also deeply rooted in the full-stack self-developed technology foundation it has built after being deeply involved in the AI field for more than ten years. The complete four-layer layout from chip, framework, model to application makes it the company with the most complete full-stack AI technology layout and the most significant synergy effects in the world. The deep accumulation of these four layers of linkage is the core source of Baidu's AI's strong explosive power.

The value of the four-layer full-stack layout of “chip-framework-model-application” is far more than simply stacking up technology. Instead, it has built a powerful closed-loop system that can achieve end-to-end optimization and self-strengthening flywheel effects from low-level computing power to top-level applications, and can achieve end-to-end optimization and self-strengthening flywheel effects. This makes Baidu a remarkable advantage that distinguishes Baidu AI in terms of technology iteration efficiency, depth of innovation, and breadth of commercial implementation. This is also Baidu AI's unique and difficult to replicate core competitiveness.

At the chip level, Baidu launched its self-developed third-generation Kunlun Core product P800 computing power chip in 2025, which has been widely used in reasoning and training tasks for various large models. The chip has already been deployed on a large scale in financial institutions such as China Merchants Bank, state-owned enterprises such as State Grid and China Steel Research, and universities such as Tongji University and Peking University. Furthermore, in August 2025, Kunlunchip successfully won the bid for China Mobile's billion-level mining project, which marks an important breakthrough in large-scale infrastructure projects.

While the P800 computing power chip is being commercialized on a large scale, Baidu officially released the next-generation Kunlun Core and Super Node product Tianchi in November. The new generation of Kunlun Core includes two products. Among them, the Kunlun Core M100 is optimized for large-scale inference scenarios, provides the ultimate cost performance ratio, and will be launched in 2026; the Kunlun Core M300 provides extreme performance for the training and inference needs of large-scale multi-modal models and will be launched in 2027. The Tianchi 256 supernode and Tianchi 512 supernode, which were released simultaneously, will be officially launched next year. A single Tianchi 512 supernode can complete trillion parameter model training.

As Baidu Kunlun Core continues to iterate, the AI computing power base it has built will become more solid and efficient. More importantly, Kunlun Core is moving from serving Baidu internally to large-scale external export, indicating that Baidu is transforming its accumulated computing power advantages into open industrial capabilities. This not only enables Baidu to accurately grasp the core opportunities of the AI era, but is also a key step in promoting the localization of computing power and consolidating the cornerstone of the development of China's AI industry.

At the framework level, Baidu's self-developed propeller deep learning framework is the largest deep learning platform in China. Currently, Propeller Wenxin Ecological Developers have reached 23.33 million, and service companies have reached 760,000. As an efficient dispatch of the underlying chip's computing power and providing a “scaffolding” and “toolbox” for upper model development, Flying Slurry is the key to achieving end-to-end optimization.

At the model level, Wenxin Big Model 5.0 is a native full-modal model released by Baidu at the 2025 World Conference. It has 2.4 trillion parameters and the ability to uniformly understand and generate text, images, audio, and video. The model uses end-to-end multiple rounds of reinforcement learning training, which significantly enhances the ability to call agents and tools.

According to evaluation data, Wenxin 5.0's text ability ranked second in the world and number one in the country in the LMarena Arena ranking. It is particularly outstanding in creative writing and understanding of complex issues. In more than 40 comprehensive evaluations, its multi-modal understanding and generation capabilities have reached the world's leading level, which is comparable to the top international models. Furthermore, the Wenxin X1.1 deep thinking model, which was released earlier, also ranked first in the domestic accurate command compliance list in the SuperClue evaluation.

At the application level, Baidu has a rich application ecosystem such as search, intelligent cloud, and autonomous driving, which can provide the technology with massive amounts of real data and refinement of diverse implementation scenarios. The generated user feedback and data can feed back the continuous optimization of lower-level models, frameworks, and chips, thus continuously strengthening the entire AI system.

Take the continued rapid growth of the AI cloud business as an example. This is mainly due to the significant increase in market demand for large model training and inference, as well as the continuous optimization of Baidu Smart Cloud's own performance and service quality. During the reporting period, Baidu upgraded its brand strategy for the Qianfan platform. The new “Baidu Qianfan” focused more on enterprise production scenarios and provided one-stop services such as models, agent development, and data intelligence. Currently, the Qianfan platform has more than 460,000 enterprise users, and the number of Agents developed has exceeded 1.3 million. It is the core platform for enterprise-level agent development. At the same time, the Baidu Baige platform launched a training template that supports rapid fine-tuning of the pi0.5 model, making it the first AI computing platform in China to provide this capability, helping to efficiently transform cutting-edge technologies such as physical intelligence into implementable products.

Based on recognition of Baidu Smart Cloud's excellent product performance, it has become the common choice of 65% state-owned enterprises, 100% systemically important banks, 800+ financial institutions, TOP 10 mobile phone manufacturers, TOP15 automobile brands with sales volume in the Chinese market, TOP10 NEV companies, etc., and has also won the title of No. 1 penetration rate in the intelligent industry.

Corresponding to this is the “China AI Public Cloud Service Market Share, 2024” report released by IDC in August, which shows that Baidu Smart Cloud has steadily ranked first with a market share of 24.6%, and has been the champion in the AI public cloud market in China for six consecutive years for a total of ten times. It can be seen from this that Baidu's advantage in the smart cloud field continues to be consolidated. With the explosion of reasoning demand in the future market, the smart cloud business, as Baidu's medium-term growth curve, will continue to drive Baidu's revenue growth in a steady manner.

Summarize:

Baidu's separate disclosure of AI business revenue in this financial report is not just a simple revenue division, but a key strategic development. It shows that Baidu's AI strategy has gradually progressed from the investment period to the harvest period. This is not only a milestone in its own development, but also provides the industry with a viable path for large-scale commercialization of AI, and is also its strategic declaration to fully sprint into AI transformation. Undoubtedly, as Baidu continues to invest in the AI field to consolidate its technical barriers and competitive advantage, and through comprehensive AI restructuring of its core business, it is expected to maintain steady growth.

There are not a few brokerage institutions and investment banks that are optimistic about Baidu's future development prospects. Recently, nearly 20 institutions, including Goldman Sachs, Morgan, and CICC, have raised Baidu's ratings and target prices. For example, Goldman Sachs said that Baidu's AI industry is shifting to an “inverted pyramid” structure, and AI models and AI applications should obtain greater value. It gave Baidu a target price of HK$150 for Hong Kong stocks and $154 for US stocks, all rated as “buy.”

Cathay Pacific Haitong also recently raised Baidu's revenue forecast for 2025 to 2027 to RMB 13.7 billion, RMB 138.5 billion and RMB 149.2 billion, and revised the adjusted net profit forecast to RMB 17.6 billion, 20.2 billion and RMB 29.8 billion. Based on the SOTP valuation method, it adjusted Baidu's Hong Kong stock target price to HK$176.

As the dominant player in the domestic AI field, Baidu is already an unavoidable target for investment in the AI industry. The dual “flywheel effect” that has been formed within its business system and can drive each other will make Baidu's future accelerated growth more certain. Baidu's first-tier flywheel is a “technical collaboration flywheel”: its core competitiveness is rooted in the “chip-framework-model-application” full-stack self-developed AI layout. This end-to-end architecture enables each technical level to achieve deep collaboration and joint optimization, thereby continuously improving the basic performance and efficiency of the entire AI system. This is the underlying driving force for flywheel rotation.

The second-tier flywheel, on the other hand, is a “commercial backfeed flywheel”: Baidu has deeply embedded cutting-edge AI capabilities and restructured the entire product line, which has significantly enhanced market competitiveness, which directly accelerates commercial monetization. The growth benefits brought about by commercialization have also provided more abundant financial support for AI technology research and development and ecological construction, forming a virtuous cycle of investment and return.

These two major flywheels do not operate in isolation. The technical flywheel provides a competitive core for commercial flywheels, while commercial flywheels deliver fuel for continuous operation to the technical flywheel. The two intertwined and accelerated each other, and jointly built a solid moat and growth momentum for Baidu to become a giant in the AI era. Under the continuous impetus of these dual flywheels, Baidu now has the potential to become a giant in the AI era, which is worthy of investors' attention.