Baidu (Minutes): Considering timely advancement of AI search commercialization

The following is $Baidu(BIDU.US)$BIDU-SW(09888.HK) the minutes of the Q4 2024 earnings call, for the earnings report interpretation please refer to《 “The Decline of Baidu, Even AI Can't Save It” 》

1. Core Information Review of the Earnings Report

1. Advertising Expectations Under Pressure, Focus on Outlook: Advertising continued to deteriorate in Q4, aligning with guidance and market expectations, also reflecting the backdrop of macro pressure and the decline in MAU of Baidu's mobile app. The decline in advertising is influenced not only by macroeconomic factors and competitive landscape but also by Baidu's control over monetization—expanding AI search penetration while still restraining the commercialization of AI search advertising.

Last quarter, management revealed that in November, the penetration rate of AI search in Q&A volume reached 20%, with a user penetration rate of 50%. However, since Q4, Doubao and Deepseek have brought significant changes to the AI market and the traditional search competitive landscape, so it is worth closely monitoring whether Baidu will make any changes to its subsequent development strategy for AI search in response.

2. AI Drives Accelerated Cloud Demand: In Q4, the growth rate of smart cloud in other businesses was 26%, which is impressive and reflects the surging downstream AI demand.

The company disclosed that in December 2024, the daily average API call volume processed by the Wenxin large model reached 1.65 billion (internal + external), with external API call volume increasing by 178% month-over-month, showing strong momentum. Under the transformation brought by Deepseek, downstream applications are expected to accelerate AI penetration this year, thus benefiting Baidu's smart cloud revenue and continuing to enhance growth.

3. Rapid Progress of Luobo Kuaipao: Excluding smart cloud revenue, the remaining businesses, such as Xiaodu, smart transportation, and autonomous driving, saw flat year-over-year growth and a quarter-over-quarter decline, partly due to the closure of Jiyue, which resulted in a loss of related ADAS revenue, confirmed as bad debt.

In Q4, Luobo Kuaipao received 1.1 million orders, a year-over-year increase of 36%, showing significant acceleration compared to Q3, mainly due to the accelerated progress under the optimized UE economic model, including deploying more vehicles and expanding into new areas. In November, Luobo Kuaipao entered the Hong Kong region and obtained autonomous driving testing permits. As of February this year, Luobo Kuaipao has achieved fully unmanned autonomous driving capabilities in all regions where it has been permitted to operate domestically 4. Profit under pressure, with potential impact from extreme losses provision: In the fourth quarter, in addition to the pressure from advertising, profits were also affected by the recognition of related costs from the closure of Extreme (one-time provision for severance compensation and other related losses of about 1 billion), resulting in a 22% year-on-year decline in operating profit. Excluding this impact, profit remained flat year-on-year, still under pressure. However, these impacts were largely in line with market expectations, and due to revenue meeting expectations and R&D expenses being lower than expected, overall operating profit slightly beat expectations.

5. Increased importance of shareholder returns: In the fourth quarter, Baidu's core free cash flow was negative, mainly due to the one-time impact from Extreme. Additionally, capital expenditures also saw a certain increase quarter-on-quarter. Dolphin Research speculates that due to the continuously increasing demand for the AI cloud, the company has made moderate increases in investment in related infrastructure.

In the fourth quarter, Baidu's repurchase increased quarter-on-quarter, but due to the small total scale, the impact was limited. Thus, for the entire year of 2024, Baidu's repurchase exceeded 1 billion USD, with zero dividends, resulting in an overall shareholder return yield of 2.3% based on a market value of 41.8 billion USD at the end of the year.

6. Detailed financial report data

II. Detailed content of the earnings call

2.1 Key information from executives

1. Strong momentum in AI cloud: Q4 revenue grew by 26% year-on-year, mainly benefiting from market recognition of Baidu's artificial intelligence capabilities, offsetting the weakness in the online marketing business.

2. Strategic partnerships: Collaborated with the State Grid of China to develop AI solutions (such as power dispatch and equipment diagnostics), thereby increasing spending on AI cloud services.

3. Adoption of ERNIE: ERNIE has demonstrated industry-leading capabilities in instruction following and state-of-the-art RAG technology (which minimizes hallucinations). In December, ERNIE processed about 1.65 billion API costs daily, with external API costs increasing by 178% quarter-on-quarter. Rapid growth was observed across various business sectors, such as education, e-commerce, entertainment, and recruitment.

4. Open-source decision: Open-sourced the ERNIE 4.5 series and launched a free ERNIE bot for end-user customers.

5. Product innovation:

(1) Baidu Wenku: Transformed into an AI-driven creative platform, integrating Baidu Cloud to achieve a seamless workflow. Launched free Canvas in January to enhance content creation capabilities.

(2) Search transformation: 22% of search results now contain AI-generated content (such as short videos and digital humans). In January, 83% of monthly active users interacted with generative AI content during searches, with a year-on-year increase of 2% in search queries per user on the Baidu App in December. Enhanced query result matching improved user metrics.

6. Autonomous driving:

(1) Apollo Go: In Q4, Apollo Go provided approximately 1.1 million rides to the public nationwide, a year-on-year increase of 36%. In January, the cumulative number of rides provided to the public exceeded 9 million. Recently, it has successfully transitioned to 100% fully autonomous driving operations nationwide. It is expected that the number of rides will grow even faster by 2025.

(2) Global Expansion: Focused on a light asset model and safety, reshaping mobility.

7. Future Outlook: Continue investing in artificial intelligence to maintain technological leadership; prioritize improving Gen-AI-supported search and explore monetization through agents; in December, over 27,000 advertisers used the ERNIE agent daily, with rapid growth in healthcare, legal, education, and B2B sectors.

8. Vision: Baidu aims to consolidate its leading position in the field of artificial intelligence innovation, creating long-term value through AI cloud, ERNIE, and autonomous driving.

2.2 Q&A

Q: Strategic thinking behind the open-source ERNIE 4.5 series and free ERNIE bot, views on the current competitive landscape of foundational models, and Baidu's plans to maintain its lead.

A: The generative artificial intelligence and foundational model market is still in its early stages but is developing rapidly. The success of DeepSeek will accelerate the adoption rate of foundational models, making them more accessible and affordable. We are entering a transformative phase where a large number of new AI applications and use cases will emerge.

From DeepSeek, we learned that open-sourcing the best models can greatly promote application; people will try open-source models out of curiosity, thus driving broader application. ERNIE 4.5 will be the best model Baidu has ever created, and open-sourcing the ERNIE 4.5 series is based on deep confidence in our technological leadership, stemming from Baidu's ten years of R&D investment, continuous technological innovation, and its unique position as one of the few companies globally with four layers of AI capabilities. ERNIE has demonstrated strong market appeal, with daily API costs skyrocketing from 50 million to 1.65 billion within a year. Through open-sourcing, we believe more developers and users will recognize the true value of ERNIE, promoting broader application and expanding its influence in more scenarios.

Similarly, making the ERNIE bot freely available allows more users to compare it with other models on a fair basis. Whether open-source or closed-source, foundational models are only truly valuable when they can effectively solve fundamental problems on a large scale. Baidu has consistently adopted an application-driven approach to improve ERNIE iteratively, transforming products and serving enterprise clients. Internally, it has successfully transformed consumer-facing products like Baidu Search and Baidu Wenku while externally enhancing the model and application development experience for enterprise clients through the MaaS platform.

ERNIE possesses industry-leading capabilities in instruction following and state-of-the-art enhanced generation technology, minimizing hallucinations and achieving widespread application across various scenarios. Its comprehensive toolchain allows clients to customize any model for their application scenarios. In the future, Baidu plans to accelerate ERNIE's iteration based on performance improvements and cost reductions, continuing to develop it in areas with the greatest impact on the real world.

Q: Adjustments/changes in search business in the future, current/target mix of generative AI results, user metrics as AI integration deepens, and competition between Baidu Search and AI chatbots like DeepSeek and Doba.

A: Currently, about 22% of search results pages contain AI-generated content, with more high-quality listings. Baidu has fundamentally changed search, making it more powerful and efficient than ever. With ERNIE, Baidu has expanded search beyond text and links, providing various content formats, including short videos, live broadcasts, intelligent agents, nodes, and shopping features, which can be dynamically combined to create a personalized search experience. While continuously transforming search, Baidu is also developing capabilities for deep personalization to adapt to each user and their preferences.

These efforts have led to better user outcomes, with 83% of monthly active users interacting with generative AI content in January, and the search query volume per user on the Baidu App increased by 2% year-on-year in December. Taking the Spring Festival holiday as an example, the ERNIE intelligent agent created value for advertisers, effectively bridging the gap when many small and medium-sized enterprises were on holiday but specific industry customer demand peaked. The ERNIE intelligent agent helped companies identify and filter highly relevant sales leads, seamlessly serving customers even with reduced staff during the holiday.

Regarding AI chatbots and search, we are still in the early stages of the AI revolution driven by foundational models. Whether it is native AI chatbots like DeepSeek or Baidu's AI-driven products, they are different ways to explore AI's potential.

As a leader in the search market with hundreds of millions of users, Baidu maintains flexibility and a comprehensive market perspective by adopting the best innovations and integrating genuinely innovative AI features, continuously penetrating and enhancing the search experience for its vast user base. Search is inherently closely related to language and text understanding, which perfectly aligns with the capabilities of large language models, placing Baidu at the forefront of the AI-enabled search transformation.

Search is evolving into an integrated platform that goes beyond AI chatbots, providing not only intelligent answers but also guiding users from information retrieval to in-depth analysis, task completion, and ultimately delivering comprehensive services and solutions. While AI chatbots represent an important early stage of AI application development, they are not yet fully mature. For Baidu, the key is to maintain a rapid and steady pace of AI transformation, focusing on discovering users' true needs for the next generation of search.

Q: The core advertising growth outlook for Q1 and the full year of 2025, macro assumptions, and opportunities and timing for Gen-AI search monetization.

A: In recent months, various policies have been introduced, such as monetary easing, entity-related policies, and trade policies, which will ultimately benefit the economy but require time to implement. Baidu's advertising business is highly correlated with small and medium-sized enterprises that are particularly sensitive to the macro environment, and the competitive landscape remains challenging, with no significant improvement seen in Q4.

However, Baidu's efforts to transform search with financial models are steadily progressing, and the AI transformation will continue to enhance user experience and create new opportunities. This quarter has further deepened the AI transformation of search with positive improvements in user metrics, which will help drive revenue growth and open new monetization opportunities.

Currently, about 22% of the total query volume corresponding to AI search results has not yet been monetized on a large scale, and monetization will advance once the AI search features are sufficiently refined. Based on these factors, Baidu believes that the advertising business has reached a turning point, and it is expected that the application business will gradually improve in the first half of 2025, with performance in the first half better than the Q4 quarter and further improvement in the second half compared to the first half.

Q: Outlook for AI cloud business growth in 2025, driving factors behind growth momentum, management's outlook on the AI cloud market in 2025, methods for enterprise cloud demand, and incremental opportunities brought by artificial intelligence to this field.

A: In Q4, Baidu's AI cloud revenue accelerated year-on-year growth to 26%, driving a 17% increase in total revenue for 2024, with revenue related to generative AI nearly tripling year-on-year in 2024. The growth momentum stems from the rising demand for ERNIE and AI infrastructure, as well as the market's high recognition of Baidu's technological leadership, successfully attracting a diverse customer base and establishing a strong opportunity pipeline.

In terms of AI infrastructure, partnerships have been established across multiple sectors, including the internet, automotive, and smart devices, leading to healthy growth in the customer base and continuous market share expansion. The MaaS Qianfan platform offers comprehensive foundational models, providing various third-party foundational models in addition to ERNIE with full-stack capabilities and end-to-end optimization to ensure performance and stability, as well as tools for model fine-tuning.

For the AI cloud market, rapid growth is expected from 2024 to 2025. On one hand, the widespread discussion of large language models during the Spring Festival has raised public awareness of foundational models and prompted more people to consider how to leverage them to improve business; on the other hand, the performance of foundational models will improve, and costs will decrease, leading more enterprises to integrate foundational models into various business processes, driving explosive growth in API costs and subsequently increasing spending on Baidu AI cloud. In terms of profits, Q4 non-GAAP operating profit margin continued to expand year-on-year, and Baidu is confident that AI cloud revenue will maintain strong growth and continue to generate positive operating profit in 2025.

Q: Trends in profit margins, core profit margin levels for Q1 and the full year of 2025, room for optimizing operational efficiency, and annual investment plans.

A: Despite short-term pressures, Baidu remains optimistic about the sustainable impact of strategic AI investments, focusing on enhancing resilience across all aspects of business operations and achieving positive progress. In the online marketing business, advertising revenue is expected to gradually improve, primarily due to plans for large-scale monetization of AI-generated search results and the ability to seize growth opportunities when the macro environment improves.

The AI cloud business has demonstrated strong growth with continuously improving profit margins and with an expanding market share, unique self-developed four-layer AI architecture, and leading AI capabilities. Baidu is confident in maintaining strong momentum. For the Robotaxi (Apollo Go) business, the focus is on narrowing losses by improving operational efficiency and enhancing user experience while also exploring innovative operational models.

In 2025, Baidu will optimize resource allocation to high-growth opportunities aligned with long-term strategies, with investments continuing to focus on projects with significant future potential and high ROI, such as enhancing ERNIE capabilities, deepening AI transformation in search, strengthening AI cloud services, and expanding autonomous driving projects, while also striving to improve operational efficiency and promote collaboration among business units to maximize the impact of strategic investments.

Q: Latest business progress of Robotaxi by 2025, fleet size targets, unit economics, value proposition, industry inflection points, competitive landscape, and Baidu's competitive and scaling approach.

A: Baidu has invested in autonomous driving technology for over a decade and has become a global leader in this field through Apollo Go. With complex traffic conditions in China and ride-hailing prices approximately 1/7 of those in the United States, Baidu has successfully operated in this environment, showcasing its exceptional technology and operational capabilities. The newly designed sixth-generation autonomous vehicle RT6 is the most cost-effective taxi in the world.

In the fourth quarter, Apollo Go provided approximately 1.1 million ride services nationwide, a year-on-year increase of 36%, with over 9 million cumulative ride services offered to the public in January. Baidu has achieved 100% fully autonomous driving operations in China, setting a new industry benchmark.

In November last year, Apollo Go received the green light for open testing in Hong Kong, demonstrating its ability to adapt autonomous driving technology to different traffic systems. This year is a key year for expansion, with expectations that fleet size and ride volume will grow faster than ever, while actively seeking partnerships with ride service providers, local taxi companies, third-party fleet operators, and other potential partners to efficiently expand using a light-asset model and maintain flexibility. The market is still in its infancy, and competition helps accelerate market development and create a more favorable innovative regulatory environment. Baidu's mission is to provide a safer, more convenient, and comfortable travel experience for more users.

Q: Baidu's strategic focus for business investment in 2025, resource allocation for search, autonomous driving, cloud services, and foundational models, management's outlook on capital expenditures for the entire year of 2025, and shareholder returns.

A: Baidu's capital allocation for 2025 will follow these strategic considerations: continuing to prioritize enhancing artificial intelligence capabilities as a long-term strategic priority, deepening the AI transformation of its product portfolio, especially in search; for AI cloud, promoting the application of AI infrastructure, the MaaS platform Tiangong, and ERNIE among enterprise clients to meet their growing demand for AI products, solutions, and cloud services; in autonomous driving, focusing on expanding domestic operational scale, exploring innovative operational models, and expanding into international markets, while ensuring strict management of investment returns and effective loss control. The two key priorities for capital expenditure decisions are strengthening technological leadership and accelerating the application of AI products and services across various fields.

Regarding shareholder returns, over $1 billion in stock has been repurchased since the beginning of 2024, significantly higher than the total repurchase amount in 2023, with a total of $1.7 billion repurchased under a $5 billion repurchase plan (as of December 2025). In the future, there are plans to accelerate the repurchase program to reward shareholders for their long-term trust and support.

Risk Disclosure Statement of this article: Dolphin Research Disclaimer and General Disclosure