Baidu: How to unlock greater value in AI search? (2Q24 summary)

The following is the summary of the second quarter financial report conference call for $Baidu(BIDU.US) in 2024. For financial report analysis, please refer to " Facing Multiple Pressures, Baidu Seeks Transformation "

I. Review of Key Financial Information:

II. Detailed Content of the Financial Report Conference Call

2.1. Key Points from Executive Statements:

1) Business Progress

1. AI Cloud Business

a. The AI cloud business continues to drive overall revenue growth, with Baidu's core business non-GAAP operating profit margin increasing by nearly 2% year-on-year, and non-GAAP operating profit growing by 8% year-on-year.

b. Baidu AI cloud revenue reached 5.1 billion RMB, a 14% year-on-year increase.

c. The performance of generative AI is strong, with revenue share increasing by around 2% compared to the previous quarter.

d. By cross-selling CPU cloud services to GPU cloud customers, the AI cloud business is expanding brand awareness.

2. Intelligent Driving and Autonomous Driving

a. Apollo Go achieved 100% fully driverless operation in Wuhan, with service coverage reaching 9 million people.

b. The sixth-generation autonomous driving car RT6 has started large-scale road testing, expected to significantly reduce future costs.

c. In the second quarter, Apollo Go provided approximately 899,000 ride services, a 26% year-on-year increase.

3. Mobile Ecosystem

a. Due to macroeconomic and market competition factors, online marketing revenue in the second quarter decreased by 2% year-on-year.

b. Key offline vertical industries such as real estate, franchising, and automotive remain sluggish, with the company still committed to accelerating the transformation towards artificial intelligence.

c. Incremental advertising revenue from generative AI and large model enhanced advertising systems continued to grow sequentially in this quarter.

4. ERNIE Model

a. API calls for the ERNIE model in the second quarter exceeded 600 million times, with the introduction of 3 lightweight ERNIE models and ERNIE 4.0 Turbo.

b. By reducing API costs and optimizing the PaddlePaddle framework, Baidu further expanded the accessibility of AI applications.

5. Search Engine and Internal Products

a. With the support of the ERNIE model, generative content's share on Baidu's search results page increased from 11% in mid-May to the current 18%.

b. Baidu's subscription revenue from Baidu Wenku saw a significant year-on-year increase in the second quarter, benefiting from the application of generative AI.

6. Strategic Outlook

a. Baidu will continue to focus on technological innovation, especially in AI-driven smart driving, AI cloud services, and online marketing.

2) Financial Performance

1. Revenue and Profit Performance

a. Total revenue amounted to RMB 33.9 billion, basically flat compared to last year, with Baidu's core business revenue at RMB 26.7 billion, a year-on-year growth of 1%.

b. Online marketing revenue was RMB 19.2 billion, a 2% year-on-year decrease; non-online marketing revenue (mainly driven by AI cloud services) was RMB 7.5 billion, a 10% year-on-year increase.

c. Operating profit margin was 21%, non-GAAP operating profit margin was 26%.

d. Other net income was RMB 771 million, a 44% year-on-year decrease, mainly due to a decrease in net foreign exchange gains.

2. Costs and Expenses

a. Revenue costs were RMB 16.4 billion, a 1% year-on-year increase.

b. Operating expenses were RMB 11.6 billion, a 9% year-on-year decrease.

c. Baidu's core business operating expenses were RMB 10.2 billion, a 10% year-on-year decrease.

d. Sales, general and administrative expenses (SG&A) were RMB 4.8 billion, an 11% year-on-year decrease, accounting for 18% of Baidu's core business revenue, down from 20% in the same period last year.

e. Baidu's core business research and development expenses were RMB 5.4 billion, an 8% year-on-year decrease, accounting for 20% of core business revenue, lower than 23% in the same period last year.

2.2, Analyst Q&A

Q: In the next 2 to 3 years, how do you view the competitive landscape in the AI field? Can Baidu maintain its current first-mover advantage, or will the technological gap between key players in the industry gradually narrow? Besides technological leadership, what other key factors do you think are crucial in building a successful AI industry ecosystem?

A: In the next two to three years, competition in the AI field will be extremely fierce, and only profitable enterprises will survive. Last year, the Chinese AI industry was in the stage of basic model competition, and I predict that the industry will undergo consolidation, with only a few models surviving.

Currently, there is strong market demand for our generative AI products, and ERNIE continues to consolidate its market-leading position. I am confident in maintaining this leadership position. Our over a decade of investment in the AI field has given us a first-mover advantage and technological leadership. In March last year, we became the world's first listed company to launch a GPT-type model, and have since continuously upgraded models through our proprietary architecture.

In October last year, we launched China's first G4-type model ERNIE 4.0, and in June this year, we introduced the more powerful ERNIE 4.0 Turbo. We will continue to launch upgraded flagship models and further expand our competitive advantage through application-driven initiatives.

Basic models have limited value if not practically applied. Currently, ERNIE handles approximately 600 million API calls per day, generating around 1 trillion tokens, setting a record in China. We extensively use ERNIE to enhance Baidu Search and Baidu Wenku, turning them into AI-native applications, which has provided us with valuable user insights And support the continuous upgrading of ERNIE. At the same time, we are committed to making ERNIE more accessible and cost-effective for widespread application in solving real-world problems. We continuously reduce the cost of model reasoning, expand model combinations, and develop toolkits to enable customers, partners, and developers to more efficiently leverage ERNIE's capabilities.

Looking ahead, we will continue to iterate on ERNIE's capabilities, continuously improving the model through practical applications and feedback from ecosystem partners. The synergistic effects of the ecosystem, advanced model capabilities, and application-driven approach will further strengthen our leadership position in the field of AI.

Q: What progress has the company made in using generative AI and large language models to transform traditional search? In AI search, what key performance indicators (KPIs) does the company focus on? Has there been a change in user behavior since the introduction of AI in the search business? Additionally, what are the company's future plans for promoting AI search? Are there specific KPI targets, such as the proportion of search results provided by generative AI by the end of this year or in the coming years?

A: We are deeply committed to leveraging cutting-edge AI technology to enhance our search capabilities and integrating search with speed into one application. Since the second quarter, we have accelerated the pace of using ERNIE to transform search, with the main goals of improving user experience, providing more features, and increasing user engagement.

Currently, generative AI provides 18% of search results, which are more comprehensive and relevant than traditional search results, effectively compensating for the decline of open web systems. Additionally, when generative AI results meet the user's current query needs, we recommend other content or services to enhance user engagement and dwell time. Baidu Search's new interactive features also allow users to refine questions through multi-turn dialogues, making the search experience more natural and efficient.

We are also rapidly expanding ERNIE agents in search results to provide intelligent assistance to users. Through an open platform, developers can create various agents to meet different users' personalized needs. Our vision is to build a powerful agent network collectively constructed by numerous developers to address a wide range of challenges and opportunities.

With these improvements, Baidu Search becomes more diverse and powerful, and we will focus on long-term sustainable development, continuously optimizing user experience. Generative AI is transforming search into a multifunctional creative platform, offering a wider range of services. In the future, we will focus on long-term sustainable development rather than short-term profitability and continuously optimize our products based on user feedback.

Q: What are the driving factors behind the accelerated growth of cloud revenue this year? How much of it is incremental revenue related to AI? What contribution do these revenues make to the long-term growth of AI cloud business? Additionally, how does the profit margin of AI cloud compare to traditional cloud business? In the intense market price competition, how does the company view the long-term development goals of the cloud business?

A: Our AI cloud revenue, after a year-on-year decline in the third quarter of last year, resumed positive growth in the fourth quarter and continued to accelerate this quarter, with a 14% year-on-year growth. This growth is mainly driven by strong demand for generative AI and Large Language Models (LLMs) across various industries, as well as our strong and efficient AI infrastructure reputation on the MaaS platform The core revenue driver of AI Cloud is related to Generative AI and basic models, which has increased from approximately 5% in the fourth quarter of last year to nearly 9% in this quarter. In order to meet the diverse needs of customers, we offer a wide range of products and solutions, including model training and hosting services. We have also observed a significant increase in GPU customers' spending on public cloud, reflecting their high satisfaction with our services and enhancing our confidence in expanding market share.

Furthermore, to encourage usage, we provide free API calls for three types of ERNIE models and have reduced the price of flagship models. This has led to a surge in ERNIE API calls from over 200 million times per day in mid-May to over 600 million times, equivalent to generating over 1 trillion basic tokens per day. We believe that as ERNIE's efficiency and performance continue to improve, the demand for API calls will continue to grow rapidly.

In terms of traditional cloud services, by cross-selling CPU cloud services, we have observed a significant increase in revenue from these customers using CPU cloud services. Overall, we remain confident in the strong growth of AI Cloud revenue and plan to continue driving long-term growth. As for profitability, we are committed to achieving sustainable healthy growth. Our AI Cloud business is in good health and continues to generate non-GAAP operating profits, with profit margins improving.

Regarding market competition, the generative AI market is still in its early stages, and providing cost-effective services is key to market growth. We are not only striving to enhance the capabilities of ERNIE models but also continuously reducing inference costs. Considering the economies of scale as a market leader, we expect to continue expanding market share and believe that the normalized profit margin of generative AI-related businesses will exceed that of traditional cloud businesses.

Q: Can management quantify the impact of soft advertising revenue, analyze the self-erosion effects brought about by macroeconomic weakness and the implementation of generative AI, and competition in the search field separately? What are the core growth points for advertising revenue in the second half of this year? If we expect to see a trend change by the end of the year, is it mainly driven by macroeconomic improvement or AI monetization? Looking ahead to 2025 and beyond, what is the sustainable growth rate of advertising revenue?

A: Our advertising business is currently under dual pressure from external factors and AI-driven search transformation. External challenges include macroeconomic weakness and intensified competition, especially in real estate, franchising, and automotive sectors. Additionally, changes in Chinese user behavior have also impacted our online advertising business, with 18% of search results currently containing generative AI content but not yet monetized. Despite these challenges, we believe that technological innovation is key to future success. We are leveraging generative AI and ERNIE to comprehensively transform the mobile ecosystem and have already seen the effects of improved monetization of the advertising system.

However, upgrading existing systems with AI is only a short-term strategy. Our long-term goal is to create lasting user value through AI-driven search transformation. While AI-enhanced search products may take time to show revenue contributions, we are confident in ultimately building a more diverse, efficient, and user-friendly AI ecosystem that will drive sustainable revenue growth Q: Can the company explain the role of AI agents in actual usage and monetization? How to promote the adoption of AI agents by small and medium-sized enterprise advertisers, especially those with weak data and technical capabilities? What is the expected promotion speed of AI agents by the end of this year? In addition, can the management share the latest progress of AI monetization, especially the transition from CPC mode to CPS mode? How do advertisers feedback on the effectiveness and future budget allocation on the Baidu platform?

A: Currently, there are over 10,000 ERNIE agents on our platform. The advantage of ERNIE agents is that anyone can create them, whether they are users, content creators, service providers, advertisers, or companies, with very low barriers to entry. The value of ERNIE agents goes far beyond being a monetization tool, significantly enhancing the user experience in our mobile ecosystem. For example, a user engaged in a 288-round, 5-hour conversation with a lawyer agent on our platform when in conflict with law enforcement, benefiting users, advertising law firms, and our platform significantly.

Currently, ERNIE agents are still in the early stages but are rapidly developing. We are pleased to see advertisers using ERNIE agents achieving better sales conversion rates than traditional methods, especially in fields such as law, education, and B2B that require long decision-making processes and personalized pre-sales consultations. As ERNIE agents become more intelligent, they have the potential to drive direct sales and create more value for advertisers, transitioning from CPC mode to a more effective CPS mode.

We believe that ERNIE agents have the potential to become a new type of website system in the AI era. While early ERNIE agents may not always significantly improve the user experience, as they provide richer interactions, increase engagement, and offer personalized services, monetization opportunities will naturally arise.

Q: Baidu's core business saw an expansion in operating profit margin in the second quarter. Is this expansion sustainable? What is the company's outlook for the full-year 2024 trend of Baidu's core business profit margin?

A: While we will not provide specific guidance for the coming quarters, I can share some insights into our business focus and its impact on costs and expenses. For the AI cloud business, we are gradually phasing out low-profit businesses to ensure continued operating profit on a non-GAAP basis. In the long term, the normalized profit margin of generative AI-related cloud businesses should be higher than traditional cloud businesses. Currently, generative AI-related revenue accounts for 9% of our total cloud revenue.

In the field of intelligent driving, we are improving operational efficiency, reducing losses, especially making progress in the autonomous taxi business. Overall, we believe that high-quality growth and investment should be balanced. Over the years, we have maintained a solid track record of revenue growth and will continue to build the future on this foundation through strict cost control.

Despite the current challenging search business environment and pressure on advertising revenue, we remain committed to providing an improved experience for users, especially accelerating the AI-driven transformation in Baidu search. In the short term, the proportion of AI-generated content in search results has significantly increased but has not yet been monetized This kind of transformation may put pressure on profit margins in the short term, but we will gradually achieve long-term returns by improving AI infrastructure efficiency and technical expertise.

Q: With the government accelerating the approval of more cities for unmanned taxi licenses, how will this affect Baidu's unmanned taxi business opportunities and related costs? What are the goals of Apollo Go for this year and the coming years? How do we plan to replicate the success in Wuhan nationwide? For example, are we considering partnerships or other forms of collaboration to drive future development? Could management share their views on the changing competitive landscape in the unmanned taxi market?

A: We have been investing in autonomous driving technology for over a decade, always maintaining focus and patience. With regional profitability on the horizon, our vision is gradually becoming a scalable reality.

Apollo Go has now become the world's largest autonomous taxi service provider, earning public and regulatory trust with a track record of safe operations, allowing us to expand service coverage. Since launching in Wuhan in the third quarter of 2022, we have steadily increased the number of fully autonomous vehicles and service trips, currently achieving 100% fully autonomous operation in Wuhan, significantly reducing costs. We expect costs to further decrease with the official operation of RT6 vehicles.

Although our market share in Wuhan is currently less than 1%, with the gradual expansion of scale, China's travel market has enormous potential, and robot taxis will naturally drive demand with more affordable services, opening up new opportunities. Apollo Go aims to provide increasingly affordable, convenient, and safe travel options. Moreover, entering this market requires a strong belief in technology, long-term investment, and robust capital reserves, areas in which Apollo Go excels, setting the industry benchmark.

As for future business models, currently all our vehicles are company-owned, bearing hardware capital expenditure and depreciation costs, but we are open to various business models and partnerships. Our short-term focus remains on achieving profitability in major cities, while the long-term goal is to seize the huge market opportunity through technology licensing and fleet operation in a light asset model, ultimately providing better transportation solutions for society.

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