Advertising relies on macro, cloud is not heavy on quality, without hesitation to invest in AI (Baidu 4Q22 phone conference summary)

Below is $ Baidu.US 4Q22 earnings call summary, financial information can be reviewed in《 Surviving the pandemic and welcoming ChatGPT, is Baidu about to have a second spring? 》

I. CEO Overview

In the past quarter, Baidu's core revenue was affected by the rapid spread of the new crown epidemic and its impact on the wider macroeconomy. Despite this, we still achieved an increase in operating profit and profit margin.

Baidu has demonstrated its value as a trusted platform for seeking accurate and authoritative information at critical times (such as the peak of the new crown epidemic). As China's epidemic prevention and control measures relaxed and infection rates surged, people became enthusiastic about monitoring the progress of local epidemics. We quickly mobilized resources and launched the COVID-19 Index based on Baidu Search, providing real-time updates and epidemic forecasts. In addition, we also cooperated with well-known hospitals and medical experts to provide guidance and advice for public health. In the fourth quarter, health-related search queries increased by about 40% YoY, reaching a record high.

Although our fourth-quarter performance was affected by the new crown epidemic, we remain optimistic about the prospect of accelerating revenue growth in the next few quarters. So far, we have seen our online marketing revenue recover from the trough, especially in vertical fields such as healthcare, tourism, and lifestyle. Assuming no major macro-environmental disruptions, we expect this recovery trend to continue and increase our online marketing revenue by increasing advertising demand and stimulating sustained consumption in China. We also expect to resume cloud project implementation, sign new contracts, and accelerate cloud revenue growth.

In addition, we are at the forefront of the AI wave-the early investment has put us in a favorable position on the track, and now the industry is about to experience significant growth and monetization. In many years, we have realized the huge potential of AI in driving market expansion and increasing enterprise value. Today, AI has become the foundation of many of our products and services, and the recent popularity of ChatGPT will completely change many enterprises. Baidu has the ability to take advantage of the upcoming inflection point of AI, and we are also excited about the upcoming launch of Wenxin Yiyu (Ernie Bot), the latest conversation AI robot supported by our latest large language model.

We plan to fully integrate Wenxin Yiyu into all of our operating products and consumer-facing products. We first plan to embed Wenxin Yiyu into Baidu Search, which we believe will reshape the search, generation, and presentation of information, create new entry points for the next generation of the Internet, help us attract more users, and gain market share in the future in a profitable way; we will also integrate Wenxin Yiyu into Xiaodu to upgrade all smart devices and services. In terms of autonomous driving, we plan to integrate Wenxin Yiyu into our automotive solutions to further enhance the user experience. We also believe that pre-training large models play an important role in driving our business expansion. Obviously, we can benefit from the new AI inflection point, because we have made overall business preparations for this new trend in a timely manner. We will continue to focus on the company's long-term sustainable development. The past year of 2022 was extremely challenging, but we wisely utilized this time, laid a solid foundation for building a more efficient organization through cost optimization, organizational restructuring, and management reorganization. Therefore, in the second half of the year, non-GAAP operating profit margin and profit both improved year-on-year. We believe that more streamlined operations will help us achieve sustainable growth.

【4Q22 Operational Highlights】

AI Cloud: 4Q22 AI Cloud revenue increased by 4% year-on-year, mainly due to the impact of the COVID-19 pandemic at the end of the year. The profit margin of AI Cloud in the fourth quarter increased year-on-year, with two main factors at work: first, we continue to reduce low-profit businesses in IaaS and cloud solutions; second, we continue to standardize our AI solutions, striving to form specific use cases for traditional industries.

Intelligent driving: We continued to make significant progress in the fourth quarter, further strengthening our confidence in the strategy and business model. Apollo's goal is still to be the world's largest autonomous driving service provider (measured by completed ride numbers). In the fourth quarter, Apollo provided 561,000 rides on open roads to the public, a year-on-year increase of 162%. By the end of January, the cumulative number of rides exceeded 2 million, indicating that users are increasingly accepting Apollo's ride-hailing service. As we expand our operations, we are able to continuously improve our technology and improve our service. As mentioned in the past, labor costs are a major component of the cost of ride-hailing services, and removing safety operators from vehicles can significantly reduce this cost. In the fourth quarter, the number of fully unmanned ride-hailing services provided to the public grew rapidly, helping us expand our operations and reduce per-mile costs. In 2022, we also launched RT6, aiming to reduce the cost of Robotaxi for the first time to the price range of popular electric vehicles in the Chinese market, which will allow us to further reduce costs in the future. Baidu Apollo's automotive solution has made a breakthrough partnership with a multinational automaker aiming to provide AVP (Automated Valet Parking) function for one of their popular models in China.

Mobile ecology: Mobile search queries and information flow distributed through the Baidu app continued to achieve double-digit year-on-year growth in this quarter. In December, Baidu's application MAU increased by 4.2% year-on-year. Despite the decline in revenue in the fourth quarter, our mobile ecosystem continues to generate high operating profit, operating profit margin, and significant cash flow due to increased efficiency.

1. Our online marketing revenue has recovered after the Chinese New Year holiday.

2. Retail and e-commerce continue to perform well, with year-on-year growth of more than 20%, accounting for more than 10% of total advertising revenue. Users not only come to Baidu to search for information and knowledge but also increasingly come to us to search for services and goods, driving the growth of Baidu's product search queries and transactions.

3. Baidu's GMV surged 84% year-on-year in Q4 2022 driven by search.

4. Despite macro challenges, short videos continued to drive YoY growth in feed revenue in the fourth quarter. Beyond feed, we believe that short videos are also advantageous for enhancing the user experience and monetization of search. The increasing prevalence of short videos in Baidu's mobile ecosystem should drive long-term online marketing revenue growth. We continue to help content creators enter our mobile ecosystem through AI-generated content, especially in the short video realm.

Our priorities are clear and focused:

Firstly, we are committed to revenue growth as we have seen a clear path to recovery.

Secondly, we are poised to leverage AI's enormous potential to drive growth. Our goal is to integrate AIGC into most of our products and services, and make our latest large language models available to consumers.

Thirdly, we will consolidate our foundation while maintaining healthy growth. Our mobile ecosystem will continue to generate robust profitability and cash flow. Our goal is to further expand the profitability of the AI cloud and surpass our internet peers. We will also continue to make strategic and prudent investments in autonomous driving to seize long-term market opportunities.

II. Key Business Data

Baidu achieved a B rating in the latest CDP climate change score, a significant improvement from D in 2021. (CDP is a non-profit organization that manages the global disclosure system for environmental impacts for investors, companies, cities, states, and regions)

AI Cloud:

As of the end of 2022, ACE Intelligent Transportation had been adopted by 69 cities (compared to 35 cities a year ago), with contract amounts exceeding RMB 10 million.

According to the China Public Cloud Market Report released by IDC in H1 2022, Baidu once again ranked first among AI cloud providers.

As of the end of 2022, the PaddlePaddle developer community has grown to 5.35 million people and served 200,000 enterprises, creating 670,000 models.

Autonomous Driving:

Baidu's autonomous ride-hailing service, Apollo Go, provided 561,000 rides in Q4 2022, up 162% YoY. As of the end of January 2023, the cumulative number of rides provided by Apollo Go to the public has exceeded 2 million. On December 30, 2022, Apollo Go obtained the first unmanned driving test license plate in Beijing, taking another step in providing fully unmanned ride-hailing services on public roads in the capital.

Mobile Ecosystem:

In December 2022, Baidu App MAU reached 648 million, a YoY growth of 4%.

Managed pages accounted for 45% of online marketing revenue on Baidu Core in the previous quarter. For the full year of 2022, this proportion was around 48%, higher than about 40% a year ago. IQIYI:

The average daily subscription of IQIYI this quarter was 111.6 million, compared with 97 million in the fourth quarter of 2021 and 101 million in the third quarter of 2022.

Other:

According to data from IDC, Strategy Analytics, and Canalys, as of the first nine months of 2022, Xiaodu is the number one smart display and smart speaker in terms of shipments in China.

Three, Q&A

Q1: From the perspective of industry, region, and macro trends, what is your view on the trend of advertising recovery since the Spring Festival holiday?

A1: As we enter the first quarter, the number of inflections of COVID-19 has surpassed its peak and is now on a downward trend. People's daily lives are gradually returning to normal, and the operations of many companies are also recovering. As the economy continues to recover, our advertising revenue has gradually improved. In studying our advertising verticals, we have seen substantial improvements to our largest verticals, including healthcare, travel, lifestyle, and franchise, which are almost all offline businesses. Online businesses, especially online gaming businesses, have seen comparatively weak recoveries.

In terms of regional recovery, second-tier and third-tier cities are recovering faster than first-tier cities because many people spend more time resting with their families before returning to work in first-tier cities.

Q2: What is your initial view on the trend of advertising recovery in the coming months?

A2: First, we believe that we will benefit from China's reopening and macro recovery. Therefore, we expect the demand of advertisers to recover, which will greatly increase our advertising revenue. I would also like to emphasize that in the process of economic and commercial recovery, advertisers are more willing to spend on performance-based advertising rather than pure brand promotion. Search has been proven to be the most effective form of performance-based advertising because users have a clear intent when they use search. Search ads link user intent with the most relevant products and service offerings, which is why we hope to receive more marketing funds from advertisers in the long term. We will continue to benefit from the continuous improvement of China's macro economy.

1. E-commerce and short videos will continue to drive our advertising business's long-term growth. Overall, we believe that in the long run, our advertising revenue growth rate should exceed that of China's GDP growth rate.
2. We are very excited about AIGC. We believe that by integrating Wenxin Yiyuan into Baidu applications, especially Baidu search, we can enhance the user experience, help us attract new users and promote user participation, and also stimulate advertisers' interest in Baidu. Therefore, this will drive our long-term revenue growth.

Q3: Many people are looking forward to Baidu's progress in ChatGPT and AIGC. Can the company talk about the future layout planning in this field?

A3: We are excited about the progress of ChatGPT and AIGC, which will change many things. We are developing Wenxin Yiyuan, which is a new conversational AI tool trained using large language models. We will combine Wenxin Yiyuan with Baidu search and open it to the public in March. We released the Wenxin big model in March 2019. Behind the Wenxin Yanyan is the Wenxin big model with over 100 billion parameters, which is trained by processing billions of user search requests and for other APPs, making it the most advanced Chinese language model. It not only includes language but also an understanding of Chinese culture. Wenxin 3.0 big model is now a very localized AI basic model in the Chinese market. Compared with models developed by overseas companies, Wenxin big model is more suitable for the Chinese market.

In addition, our revenue-based deep learning framework, PaddlePaddle, has also developed well, with millions of developers using the platform. There is strong collaboration between the framework layer and the model layer. Pre-training AI costs a fortune, and we believe our full-stack AI capabilities will enable us to build the most efficient large language model and support various applications, including search, content generation, and any vertical area that can significantly improve productivity.

We are actively integrating Wenxin Yanyan and Baidu search services. Users will soon be able to interact directly with the new generative language model, which will enhance and upgrade the traditional search experience and attract more users. Such features are likely to become a new traffic entry point for Internet users, thereby expanding the market size of search. At the same time, it will also help our advertisers, content creators, and merchants.

Regarding AI Cloud, our AI Cloud has full-stack capabilities, including cloud infrastructure layer, open-source deep learning framework layer, big model layer, and application layer. We are providing MaaS (Model-as-a-Service), which can help many business owners build their own models and applications on Baidu Cloud and bring significant positive changes and improvements in many areas, including efficiency, better decision-making, and improved customer experience.

To sum up, Baidu has built our moats in large language models and AI basic technology with powerful AI capabilities. In addition to expanding the market size of search, we are also able to help many industries establish their models, develop their applications and improve productivity. We believe that AI will change the rules of the game in the cloud computing market. AI is greatly changing many industries, and we are very excited about what is about to happen. Many partners have announced the integration of our Wenxin Yanyan service.

Q4: As for the growth of the AI Cloud business, will the interruption of project revenue recognition caused by the pandemic turbulence in 4Q22 drive the revenue growth rate in 1Q23? What is the outlook for the company's AI Cloud business in 2023? Which products are the main growth drivers in the cloud service market, and how do you view the development prospects in the next few years?

A4: After reopening, as our cloud service team began to focus on ongoing projects, we did see cloud revenue gradually recover. The cycle of cloud project revenue recognition varies from a few months to several quarters. Given this, our AI cloud revenue growth is expected to regress this year. At the same time, we will continue to focus on reducing some low-profit businesses, as we are dedicated to improving the profitability of our cloud business. This move will have an impact on our ad revenue growth and the future. Our goal is still to keep the growth rate of cloud services ahead of our internet peers. In the long run, moving traditional industries to the cloud and using AI to improve efficiency are two unchanging trends. I believe that digital and intelligent transformation will be the main driving force for growth. It is certain that we will benefit the most from this transformation because we believe that Baidu has a unique position.

Our AI cloud provides 4 layers of full-stack capabilities, including cloud infrastructure layer, open source deep learning framework layer, large model layer, and ultimately the application layer. All 4 layers can work seamlessly together and provide end-to-end solutions. Compared with other solutions built by different vendors, our cost is lower. In addition, end-to-end solutions can actually be upgraded in a shorter amount of time. Once we make any improvements to any of these layers, the overall effect will improve. So if users are in industries such as transportation, manufacturing, energy, and utilities, I think Baidu's cloud services should be the most efficient. Because in addition to the technology and advantages I just mentioned, we have a deep understanding of the industry, and we have accumulated solid experience by successfully delivering high-quality industry solutions. Most importantly, as Robin mentioned earlier, we plan to integrate Wenxin Yiyu into our cloud services and AI products. Therefore, this should help many companies build their innovative applications, explore new businesses, or improve their operational efficiency. So we believe this will be an important step for our AI cloud and may reshape the entire cloud industry in China.

Q5: How is the IT spending for cloud services and when will it break even?

A5: We continue to improve the profitability of cloud services by cutting low-profit businesses and continuing to build scalable AI applications. This has been our consistent strategy for many years. Among them, personal cloud services will continue to generate considerable operating profit margins. While the overall enterprise cloud business is still losing money, both gross profit margin and operating profit margin are improving, and we expect the profit margin to continue to improve in the past few quarters. We are expected to make AI cloud a profitable business in the next few years.

Q6: What is the investment plan for the company in 2023, and what is the profit margin of the core business? What are the long-term targets for R&D and SG&A expenses?

A6: We started to cut costs from the end of 2021, including cutting some businesses where we don't have advantages. We carefully review and restructure our business situation. Now we have a more streamlined operation system than before. In the past year, the total number of employees in Baidu's core business has actually decreased by 8% year-on-year.

For AI cloud, we are gradually phasing out low-profit businesses. We also strengthened control over variable costs. In 2022, Baidu's core business operating expenses decreased by 9% year-on-year, SG&A decreased by 15% year-on-year, and R&D decreased by 3% year-on-year. In the second half of 2022, Baidu's core business Non-GAAP OPM achieved a year-on-year growth.

In 2023, we will maintain strict discipline in terms of expenses, balance revenue growth and profit margins. We will continue to focus on our marketing expenses, especially customer acquisition costs. We will carefully manage all other variable costs. In terms of R&D, we are working hard to improve the efficiency of our team, allowing them to focus on building technology, products, or services that create more value for users and customers.

At the same time, we will continue to invest in growth and new opportunities while closely monitoring the returns on capex. We will continue to invest in cutting-edge technologies such as large language models and generative AI because we believe they will transform many businesses and provide huge opportunities. We can further improve the efficiency of reallocating resources to support new initiatives, such as Wenyi Yiyuan. In summary, we will not hesitate to invest in our promising new businesses.

So in conclusion, our advertising business should continue to generate high profits and cash flows, which are the foundation of our business.

For AI Cloud, our goal is to continue to expand the profitability of personal clouds. Enterprise clouds will continue to cut low-profit businesses and focus on core industries.

We will continue to standardize our AI solutions, which will provide considerable profits in the future. We will continue to invest in the future, but at the same time, we will carefully monitor operational and financial indicators to evaluate our long-term performance and investment returns. We believe we still have room to optimize costs and expenses, but in the process, we will dynamically allocate resources and balance growth and profit margins.

Q7: How is the UE model of the Robotaxi business in the company? When will it achieve breakeven, and what is the technology development roadmap?

A7: The main costs of Robotaxi are labor costs and vehicle costs-

In terms of labor costs, completely unmanned driving means no safety officer. This helps reduce labor costs. In 2022, we made breakthrough progress in fully unmanned Robotaxi services. Now we provide free unmanned Robotaxi services in Wuhan and Chongqing. For example, in Wuhan, we continuously expand our business by increasing the coverage, daily operating hours, and fleet size. In December 2022, we were approved to test fully unmanned Robotaxi services in Beijing Economic and Technological Development Zone. This is an important step.

In 2023, we plan to further expand our fleet and scale of unmanned operations. We will continue to reduce labor costs, allowing more vehicles to remove safety officers, and more and more cities in China support Robotaxi services. In addition, as we mentioned earlier, expanding our operational scale will help us improve all technologies and further improve safety performance. As we expand our scale, we will further improve our overall technology and gain the trust of regulatory agencies. Apollo remains the world's largest Robotaxi service provider. Last year, we provided more than 1.5 million orders to the public. Now, top cities like Beijing have an average of 15 orders per vehicle. This number is approaching the average daily volume of taxis. We believe that by continuously improving our technology, there will be more and more cities offering autonomous taxi services in the future.

In terms of vehicle costs, our Apollo RT6 has brought the cost of Robotaxi down to the common price range of electric vehicles on the market for the first time. We believe that vehicle costs will continue to decrease in the future. As the electric vehicle market in China is developing rapidly, the entire supply chain is highly localized. There is still a lot of room for cost reduction. Combined with these two factors, we believe that the UE model will continue to improve. In addition, once we scale up, we can use large language models to improve our technology. This will help us further improve safety performance.

Q8: After the reopening in China, how did the company's user traffic and average usage time change, especially in large cities?

A8: After the epidemic policy was reopened, people started to spend more time outdoors. In January, the overall average usage time of our main APP increased by 6.6% year-on-year, which is better than the growth rate of the entire APP market - this is because we have diversified APPs to meet the various needs of users.

Baidu APP is one of the few super apps in China. Baidu search has unique value. We use technology to help people quickly find the most relevant and authoritative information online. So every time users want to search for information and knowledge, they come to Baidu. This becomes more obvious when users return to work. Even if they want to learn about the development of their local education or examine high school scores, these needs can be satisfied through Baidu search and other applications.

In the past year, we have used technology to enrich content and services, providing users with better search experiences, information flow experiences, and many other product experiences. People use the Baidu APP not only to search for information, but also to find products and services. These efforts have consolidated Baidu's leading position in China's mobile internet industry.

Another highlight, of course, is Wenyuanyiyan. As I mentioned before, it is likely to become a new traffic entry for people on the Internet. In addition, using AI to generate short video content is a typical application of generative AI. We believe that over time, AIGC will help us accumulate more short videos on our platform, increasing the number of video views and advertising revenue.

Q9: Many domestic companies have cooperated with Baidu's Wenyuanyiyan. Can the company provide detailed information on cooperation models, monetization opportunities, and how they compete with other internet companies offering similar services?

A9: Since we announced the launch of Wenyuanyiyan, many companies have contacted us and expressed strong interest in cooperation. Some companies have announced cooperation with us to integrate Wenyuanyiyan into their products or services. More and more companies are realizing that generative language models will change their industry, and they want to seize this opportunity. The attention that Wenxin Yiyuan has received also demonstrates Baidu's leading AI capabilities, and the AI era has finally arrived.

In terms of monetization opportunities, Wenxin Yiyuan will be integrated into Baidu Search to improve the user experience, and users will rely more on us to complete various tasks, which can expand the search market. We are using AIGC to expand our content, such as text, images, and videos, which will create an excellent opportunity for us to attract new users, increase user usage time, and user stickiness.

For cloud customers, they will be able to leverage our full-stack AI capabilities, and they can develop their applications based on our AI framework and basic models, which will drive the revenue growth of our cloud services. Our full-stack AI capabilities are very unique. It includes cloud service layer, framework layer, model layer and application layer, and we have strong influence in all four layers. Our Wenxin 3.0 model has been trained on billions of daily user search requests and other corresponding training from other Baidu apps. The training of Wenxin model uses well-tagged massive data, which helps to update quickly and iterate. The entry barrier for AI is very high, requiring years of continuous investment, and we have a first-mover advantage. The technology behind Wenxin Yiyuan is called reinforcement learning with human feedback, and the human feedback part is crucial. As the first promoter, we are getting more and more human feedback. And further improve our capabilities, leading anyone else. And now we are focusing on preparing for the release of early products and continuously improving them after launch. We also focus on Wenxin Yiyuan to create areas of direct value.

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