Baidu: Consumer confidence remains weak, waiting for policy transmission (3Q24 conference call summary)

The following is the summary of the Q3 2024 earnings conference call for $Baidu(BIDU.US)$BIDU-SW(09888.HK) . For the financial report interpretation, please refer to Baidu: No way to turn the tide, only a difficult path ahead .

1. Core Information Review of the Financial Report:

2. Detailed Content of the Financial Report Conference Call

2.1 Key Information from Executives:

2.1.1 Business Progress

1) Core AI Technology

a. The self-developed four-layer AI infrastructure optimization, ERNIE 4.0 Turbo's reasoning efficiency has improved by 48%, and reasoning costs are expected to continue to decline.

b. Launched lightweight models Speed Pro and Lite Pro, featuring low latency, high throughput, improved stability, and higher accuracy.

c. ERNIE's daily API call volume increased from 600 million in August to 1.5 billion in November, with daily token generation exceeding 1.7 trillion.

d. Launched the ERNIE iRAG model for text-image generation, addressing generation logic issues in e-commerce and marketing scenarios, significantly improving the accuracy of generated content.

e. ERNIE Agent

- Daily conversation volume reaches 15 million, with a Top 100 Agent list covering a wide range of scenarios.

- A typical case includes the intelligent agent developed for BYD, achieving a 70% second visit rate, significantly enhancing brand engagement and purchase intention.

- Developed intelligent agents for companies like Samsung and Yanghe, demonstrating the wide applicability and commercial value of large language model agents in brand interactions.

2) Search and Mobile Ecosystem

a. Over 20% of search result pages contain AI-generated content (up from 18% in August), with 70% of Baidu App's monthly active users engaging with AI-generated contentb.

b. AI-generated search results enhance user stickiness and complex query rates, with the frequency of multi-turn interactions continuing to grow, reaching tens of millions of daily interactions.

c. The number of monthly active users for the document library AI function surpassed 50 million in September, a year-on-year increase of over 300%. Subscription revenue increased by 23% year-on-year, mainly due to the launch of AI features and an increase in users' willingness to pay.

3) Customer Service Products

a. Customer Joy:

Through ERNIE-supported AI capabilities, it can independently resolve most user issues, reducing the need for manual intervention.

API call volume has increased significantly, with daily peaks reaching millions.

Collaborated with Yum China to validate product stability and intelligence.

b. Exam Treasure:

Generated missing answers and explanations through ERNIE 4.0 Turbo and ERNIE Speed Pro, with the effective number of questions in the question bank increasing by 150%.

Replaced manual writing of explanations, reducing related costs by 99.8%.

Achieved real-time AI tutoring services, allowing users to upload questions and receive instant answers.

The proportion of paying users increased by over 100%, with revenue growing by 246% year-on-year.

4) AI Cloud Business

a. Revenue Performance

AI cloud business revenue reached 4.9 billion RMB, a year-on-year increase of 11%, maintaining double-digit growth and achieving non-GAAP operating profitability.

Revenue related to General Artificial Intelligence (GenAI) continues to grow, with its proportion of AI cloud revenue rising from 9% in the previous quarter to 11%.

b. Technical Advantages and Infrastructure:

GPU cluster optimization: Managing tens of thousands of GPUs, achieving an efficiency rate of 99.5% for LLM training time.

Heterogeneous GPU integration: Further enhancing the ability to train and deploy models, controlling throughput loss from combining GPUs from different vendors to within 5%.

Strengthening market position: Consolidating leadership as a high-performance, low-cost AI infrastructure provider.

c. Product and Tool Development:

Model Builder tool: Supports efficient fine-tuning of lightweight models, with performance comparable to ERNIE 4.0, while improving inference speed and reducing costs.

CodeMate programming assistant: AI-generated code now accounts for one-third of the company's new code, improving internal R&D efficiency.

Instant Answer platform: Launched the first no-code tool, simplifying enterprise AI application development through natural language instructions.

d. Customer Demand:

• Revenue from mid-tier enterprise customers in the public cloud increased by 170% quarter-on-quarter, reflecting the growing demand for AI services among small and medium-sized enterprises.

5) Mobile Ecosystem

a. Online Marketing Revenue:

• Online marketing revenue decreased by 4% year-on-year, mainly due to macroeconomic pressures, upgrades to AI-driven search systems, and the evolution of the internet content ecosystem
• The improvements brought by Generative AI (GenAI) and Large Language Models (LLM) to advertising systems have driven a continuous increase in incremental advertising revenue, reflecting the company's ongoing efforts to optimize monetization systems and advertising platforms.

b. Large Language Model Agent:

Functionality has been further optimized, with significant improvements in response accuracy and intent recognition capabilities.

As of September, approximately 20,000 advertisers have generated advertising spending through the Agent.

Case study: A healthcare client, after creating its own Agent, achieved more intelligent conversations and high-quality customer service through continuous optimization, resulting in a 13.6% increase in users' willingness to initiate conversations and a 50% increase in lead conversion rates (data from July to September).

c. Future Outlook:

Despite the online marketing business being in a transitional period, Baidu will continue to drive AI transformation, committed to creating long-term value for users, customers, and shareholders.

6) Intelligent Driving

a. Technological Progress:

The sixth-generation autonomous vehicle RT6 has been operating on public roads in multiple cities in China.

The proportion of fully unmanned driving operations nationwide reached 70% in the third quarter and further increased to 80% in October.

Achieved 100% fully unmanned driving operations in Chongqing.

b. Operational Scale:

In the third quarter, Apollo Go provided approximately 988,000 public travel services nationwide, a year-on-year increase of 20%.

As of October, the cumulative number of public travel services exceeded 8 million, further consolidating Baidu's leadership position in the intelligent travel field.

c. Future Expansion:

Regulation and Deployment: Although the technology is ready for large-scale deployment, the company will patiently cooperate with the improvement of regulatory frameworks.

New Cities and Operational Models: Exploring entry into more cities and adopting a light-asset model to enhance operational flexibility and uncover new growth potential.

2.1.2 Financial Performance

1) Overall Performance

a. Total Revenue: RMB 33.6 billion, a year-on-year decrease of 3%.

b. Baidu's core business revenue: RMB 26.5 billion, basically flat compared to the same period last year.

Online marketing revenue: RMB 18.8 billion, a year-on-year decrease of 4%.

Non-online marketing revenue: RMB 7.7 billion, a year-on-year increase of 12%, mainly driven by AI cloud business.

c. iQIYI revenue: RMB 7.2 billion, a year-on-year decrease of 10%.

2) Cost and Expense Side

a. Operating Costs: RMB 16.4 billion, a year-on-year increase of 1%.

b. Operating Expenses: RMB 11.2 billion, a year-on-year decrease of 5%.

c. Baidu's core business operating expenses: RMB 9.9 billion, a year-on-year decrease of 5%.

• Selling, General and Administrative Expenses (SG&A): RMB 5 billion, a year-on-year increase of 4%, with the proportion of core business revenue increasing from 18% to 19%
• Research and Development Expenses (R&D): RMB 4.9 billion, a year-on-year decrease of 13%, accounting for 19% of core business revenue, down from 21%.

3) Profitability

a. GAAP Operating Income: RMB 5.9 billion.

• Baidu's core business operating income: RMB 5.7 billion, operating profit margin of 21%.

b. Non-GAAP Operating Income: RMB 7 billion.

• Non-GAAP Baidu's core business operating income: RMB 6.7 billion, operating profit margin of 25%.

c. Net Profit:

GAAP net profit attributable to Baidu: RMB 7.6 billion, diluted earnings per ADS of RMB 21.6.

Non-GAAP net profit attributable to Baidu: RMB 5.9 billion, diluted earnings per ADS of RMB 16.6.

GAAP net profit of Baidu's core business: RMB 7.5 billion, net profit margin of 28%.

Non-GAAP net profit of Baidu's core business: RMB 5.7 billion, net profit margin of 21%.

d. Other Income: Total RMB 2.7 billion, a year-on-year increase of 14%.

e. Income Tax Expense: RMB 814 million, a year-on-year decrease of 37% (compared to RMB 1.3 billion in the same period last year).

4) Employee Count

As of September 30, 2024, the total number of employees in Baidu's core business is approximately 31,000.

2.2 Q&A Analyst Questions

Q: How should we assess the penetration speed of generative artificial intelligence in search results? What does management consider the optimal penetration rate? Additionally, can you share the expected duration of the transformation of search products? Finally, please update us on the latest progress in monetizing AI search and when we can expect the commercialization of AI-generated search results?

A: We are continuously driving the transformation of our search business through AI technology. Currently, over 20% of search pages and more than 70% of monthly active users have interacted with AI-generated content. However, these figures only represent our initial achievements in the AI-driven search transformation process and do not fully reflect the overall application of AI-generated content, nor should they be the sole metric for measuring progress. What truly matters is our long-term goal of comprehensively transforming the search experience through the ERNIE model.

Our aim is to achieve a revolutionary transformation in search through AI-generated content, providing users with seamless on-demand and personalized information services, anytime and anywhere, in various forms to meet their needs. Through the ERNIE model, we can gain deep insights into the intentions behind users' complex queries, thereby generating content that better meets their needs. At the same time, we are committed to enriching content formats, including AI summaries, images, videos, agents, posts, and even digital humans. These different formats can be dynamically combined to create a personalized search experience for users and optimize content generation and presentation to better meet user preferences.

This series of initiatives has significantly enhanced user experience and driven higher user engagement. The AI-driven search transformation has already shown initial results in certain user metrics, which has strengthened our confidence in continuing this workThrough continuous exploration, we are breaking the boundaries of AI applications and gradually unlocking its unprecedented potential. In every quarter in the future, we will further showcase the results of AI-driven search transformation.

In terms of commercialization, we have made initial progress in certain verticals such as law, education, and B2B services. For example, the advertising agency function based on the ERNIE model has enhanced effective sales leads for advertisers, creating value for clients while optimizing user experience. As the ERNIE foundational model continues to evolve, we expect the agency function to unlock greater revenue opportunities in the future. However, we are still in the early stages of exploring AI search monetization and are gradually advancing in a mature and robust manner. In the short term, we will prioritize enhancing user experience rather than rushing to achieve commercialization.

This is highly consistent with our strategic direction, which prioritizes long-term value creation over short-term gains. Although we may face certain pressures in the short term, these trade-offs are worthwhile as they pave the way for achieving our long-term vision. In addition to the search business, we are also leveraging the ERNIE model to comprehensively upgrade multiple consumer products in the mobile ecosystem, such as information streams, Wenku, and Haokan Video. By optimizing content production, distribution mechanisms, and commercialization capabilities, we are enhancing user experience and engagement.

Early positive results further reinforce our strategic confidence. We expect that after this adjustment period, our advertising business will see improvements next year.

Q: The recent rapid growth in API call volume, what are the main driving factors? Additionally, have there been any potential "killer" applications in the B2B or B2C markets? Finally, what is the management's outlook on the medium to long-term application prospects of ERNIE?

A: The growth in ERNIE's API call volume is significant, driven mainly by the following factors:

(1) Model capability enhancement: ERNIE continues to optimize, significantly improving its intelligence level and the efficiency of solutions, enabling more applications to provide value to users;

(2) Decreased inference costs: The continuous reduction in inference costs allows more clients to afford and adopt this technology;

(3) Toolchain support: We provide clients with convenient toolchains to help them easily customize models based on specific application scenarios to meet their needs.

The rapid growth in API call volume is driven by both internal and external demand. We began promoting product transformation in the second quarter of last year, first upgrading the monetization system to improve advertisers' return on investment, bringing in incremental revenue of hundreds of millions each quarter. At the same time, we have expanded AI transformation to consumer-facing products, including Baidu Search, Wenku, Baidu Cloud, Input Method, and Maps, which have over 100 million monthly active users. These products with a large long-term user base are undergoing comprehensive innovation through AI.

This year, we accelerated the AI transformation of the search business, displaying more generated content on search results pages. The search business, due to its natural fit with language and text understanding, is highly aligned with the capabilities of large language models (LLMs) and has the potential to become a "killer" application in the era of generative AIIn addition, millions of intelligent agents have been deployed to answer user and customer questions, further driving the growth of API call volume.

Although most API calls currently come from upgrades of consumer products, the acceptance of ERNIE in the B2B market is also rapidly increasing. External API calls grew by approximately 240% year-on-year in the last week of the third quarter, widely applied in fields such as online education, social media, catering services, healthcare, legal consulting, and recruitment, indicating that enterprises recognize the value of powerful models and are willing to invest.

As generative AI becomes the core of our product line, ERNIE is driving Baidu's transformation from an internet-centric business to an AI-centric business. This transformation will unlock new revenue sources and further consolidate Baidu's market leadership.

Q: Can you share the current trends in advertising demand? In particular, have you observed a significant improvement in advertiser sentiment or advertising spending following the implementation of the latest stimulus policies? Additionally, what is management's view on the macroeconomic outlook as we approach 2025?

A: Our advertising business is closely related to the macroeconomic environment, especially since a large proportion of our advertising business comes from offline small and medium-sized enterprises (SMEs). These SMEs are deeply connected to the domestic consumer market and are highly sensitive to macroeconomic expectations; their vitality and recovery speed are important indicators of our business performance.

In the third quarter, we observed continued sluggishness in industries such as real estate, franchising, and healthcare. As of the fourth quarter, there has not yet been a significant improvement in advertising spending patterns, and consumer confidence remains weak. However, we are encouraged by the strength and timeliness of the recently launched stimulus policies, as well as the positive measures emerging in the market. Although these policies will take some time to transmit to SMEs and boost their confidence in advertising spending, we maintain a cautiously optimistic outlook on the recovery trend in the future.

With the macro environment stabilizing and domestic consumption warming up, we expect the confidence of SMEs to recover rapidly. As Baidu is the preferred platform for helping clients acquire customers, it has mature advertising effectiveness and can seize opportunities when SMEs resume advertising spending. After market confidence rebounds, we anticipate a significant recovery in advertising business, and with Baidu's extensive reach and unique value proposition tailored for SMEs, we are confident in achieving strong growth momentum during the market recovery.

Q: Given the current pressures facing the advertising business and the company's plans to accelerate and expand its AI transformation, can management elaborate on the current business focus, investment and resource allocation priorities, and expectations for profit margin trends in the fourth quarter?

A: We will continue to focus on our AI strategy, viewing it as a near-term priority and a long-term strategic focus. Given that AI-driven search reform is still in its early stages and AI-generated search results have not yet been monetized on a large scale, we expect our online marketing business to continue facing certain pressures in the short term. However, we firmly believe that this strategic direction has enormous long-term value.

According to this strategy, we will further strengthen product innovation, focusing on advancing AI search, while continuously increasing investment in enhancing ERNIE model capabilities and optimizing our layout in AI cloud services and autonomous driving, in order to maintain healthy profit levels and clarify our path to profitabilityThese measures are crucial for consolidating our leading position in China's technological innovation sector.

Therefore, profit margins may be in a period of adjustment in the short term. Looking ahead to 2025, we will focus more on the efficient allocation of resources, prioritizing support for high-growth areas while ensuring alignment with the company's long-term strategy.

Q: Can you analyze the performance of personal cloud and enterprise cloud? What are the latest dynamics in the current market competition landscape? What is the outlook for cloud business, especially the expected contribution of generative AI, and how do you view the assumptions regarding long-term profit margins?

A: Growth of cloud business driven by generative AI: Since last year, China's cloud industry has accelerated its transformation towards AI computing, benefiting from the development of generative AI and foundational models. In this quarter, revenue related to generative AI accounted for 11% of total AI cloud revenue, doubling from the 5% reported in last year's Q4. Management is confident in the sustainability of this growth trend.

Personal cloud: This quarter, personal cloud revenue was affected by short-term business adjustments, leading to a slowdown in growth. However, the upgrade of personal cloud driven by ERNIE is expected to mitigate short-term impacts and create more opportunities for long-term growth.

Enterprise and public sector cloud services: A major part of AI cloud revenue, maintaining strong growth momentum, with growth rates exceeding the overall growth of AI cloud business. Particularly strong demand for model training and inference in industries such as internet, education, and finance.

Major clients in industries such as internet, technology, and automotive have increased their usage and spending on GPU public cloud. The number of clients and spending from small and medium-sized enterprises has also significantly increased, especially in growth sectors like marketing software. The current cloud market has increasingly high demands for the capabilities of foundational models, and considering the needs for AI infrastructure, specialized technology, and capital investment, it is expected that only a few foundational models will survive in the market. Baidu is confident that ERNIE will become a leader.

The non-GAAP operating profit margin of AI cloud has improved year-on-year, mainly due to the continuous growth of generative AI-related revenue and the pursuit of high-quality revenue. Looking ahead, management is confident in the long-term strong revenue growth and healthy profitability of the AI cloud business.

Q: What is your view on the future competitive landscape of autonomous driving, especially after several autonomous driving-related companies recently went public in collaboration with large automotive companies and platforms? What is the strategic development direction of Apollo Go mobility service, including how to plan future business expansion and growth plans?

A: The autonomous driving industry has a very high threshold, requiring top-notch technical capabilities, excellent talent reserves, as well as long-term investment and strong capital support. Baidu has established an unparalleled advantage in these areas, laying the foundation for industry leadership through over a decade of continuous investment and establishing a global technological advantage.

Apollo Go has provided over 8 million mobility services, becoming the world's largest autonomous driving mobility platform. In terms of hardware, the cost of Apollo RT6 in the mass production phase is less than $30,000, possessing the strongest cost competitiveness in the market.

The current autonomous driving market is still in its early stages, and competition helps to drive market growth and foster a more friendly and innovation-supportive regulatory environment. Baidu believes that a healthy regulatory environment is key to the industry's development and is prepared to expand its business

Baidu is actively seeking international expansion opportunities, focusing on cities with conditions for large-scale fully autonomous driving operations (such as Wuhan and Chongqing). At the same time, Baidu is exploring a lightweight asset business model to maintain operational flexibility and efficiency, laying the foundation for future growth.

Q: What are the latest developments regarding the company's capital allocation strategy, particularly concerning shareholder return plans, including stock buybacks or potential dividend distributions?

A: Since going public in 2005, the company has been committed to creating long-term value for shareholders through growth in China's leading technology sectors and a continuous stock buyback program. Over the past few years, we have repurchased approximately $1 billion worth of stock on average each year. This year, through ongoing buyback efforts, our total number of outstanding shares has decreased. Although the pace of buybacks may vary by quarter, we consistently view the stock buyback program as one of the core ways to deliver value to shareholders.

Currently, we are focused on executing a sustainable and regular stock buyback program while also being open to evaluating other potential shareholder return methods. We firmly believe in Baidu's long-term sustainable growth potential and ensure that shareholder trust is rewarded. The fundamental way to serve shareholder interests is to build a solid business foundation and seize growth opportunities.

In the current wave of technological transformation centered around generative artificial intelligence and foundational models, Baidu possesses strong AI infrastructure, technical expertise, and ample financial and human resources to implement our strategic plans. We aim to maintain investment flexibility when necessary to drive the development of new businesses, thereby delivering long-term value to shareholders.

Q: Can you outline the future technology roadmap for Wenxin? What are the key milestones in the company's AI model development, and when are they expected to be achieved?

A: Since the launch of ERNIE last March, we have continuously enhanced the capabilities of foundational models, particularly our flagship model.

In October last year, we released China's first QPD4-level model with globally leading capabilities—ERNIE 4.0. The ERNIE 4.0 Turbo launched in June this year further improved performance. Based on these milestones, we will continue to optimize the flagship model, providing higher performance, more accurate outputs, and broader support for user needs. A new version of ERNIE is expected to be launched early next year to further consolidate our leading position in the foundational model field.

Compared to overseas companies, we have adopted an application-driven strategy. We believe the true value of foundational models lies in supporting practical applications that are widely used and meet user needs. Over the past 18 months, our model development has consistently focused on solving real-world problems based on market demand, achieving significant progress, including reducing model hallucinations, improving accuracy, and launching customized models that meet diverse needs.

To make the ERNIE model more user-friendly and cost-effective, we have enhanced performance, reduced inference costs, and optimized response speed. Our unique advantage lies in a clear strategic direction and defined core capabilities, prioritizing the development of technologies that are highly aligned with our business and can create maximum value. Earlier this year, we expanded our visual foundational model capabilities to the autonomous driving business, further strengthening our leading position in this fieldIn addition, we are actively exploring multimodal capabilities and applications based on the advantages of language models, aiming to create collaborative value and develop more potential possibilities. In various fields, we always focus on the efficiency of resource allocation, optimizing the foundational model to maximize its impact on the business while maintaining market leadership.

Risk Disclosure and Statement of this Article: Dolphin Investment Research Disclaimer and General Disclosure