Source: Wall Street News
Li Yanhong pointed out that currently 11% of Baidu search results are generated by AI. In addition to mobile phones, Wenxin's big model has also begun to expand into the PC and electric vehicle fields. Baidu's “Radish Run” has accumulated more than 6 million service orders. In Wuhan, the proportion of orders for the “Radish Run” fully driverless service exceeded 55% in the first quarter, and continued to rise to 70% in April. Baidu expects this ratio to rise rapidly to 100% in the next few quarters.
Before the US stock market on May 16,$BIDU-SW (09888.HK)$/$Baidu (BIDU.US)$Financial results for the first quarter of 2024 were released. According to financial reports, Baidu's revenue for the first quarter was 31.5 billion yuan, a slight increase of 1% over the previous year, and is expected to be 31.43 billion yuan.
Net profit (non-GAAP) attributable to Baidu in the first quarter was 7.011 billion yuan, up 22% year on year, and net profit (non-GAAP) attributable to Baidu's core was 6.628 billion yuan, up 26% year over year, all exceeding market expectations. The core operating margin (non-GAAP) was 23.5%.
In a subsequent conference call, Li Yanhong, founder, chairman and CEO of Baidu, said that with the expansion of the scale of driverless driving operations and the continuous improvement of the cost structure, hardware depreciation costs will be significantly reduced this year; with the continuous improvement of operational efficiency and cost reduction, Radish Run will take the lead in achieving UE break-even within the Wuhan region in the near future.
Currently, 11% of search results on Baidu Search are generated by AI, so that searches can answer user questions more accurately, more organized, and more directly. Baidu Search's AI restructuring is still in its early stages. Overall, search is most likely to become a killer app in the AI era.
The following is a summary of the analyst Q&A session of the conference call:
Q1: Does AI technology help Baidu increase ad monetization rates? In what areas are advertisers seeing improvements? What areas can be further strengthened?
Answer: Baidu advertising monetization is the first business to benefit from AI. Starting in the second half of last year, it has generated hundreds of millions of yuan in incremental revenue every quarter. Baidu has been using Wenxin's big model to upgrade the monetization system, enhancing all aspects of advertising technology, including improving, adding features, improving the delivery system, automating creative generation, and adding strategic information to advertisers. This improvement prompted advertisers to increase their spending on Baidu, and AI-related incremental ad revenue increased sequentially in the first quarter. It is expected that this trend will continue.
Incremental revenue helped Baidu mitigate broader macroeconomic weakness. Advertisers can build their online presence and interact with potential customers using natural language through multiple rounds of conversations, and with the introduction of Agent Builder, advertisers can easily create customized AI assistants. When advertisers express their intentions to AI assistants, they can achieve their goals more effectively. Agents can also help enrich Baidu's content and improve Baidu's user experience.
Although it's still in its early stages, some agents have helped some advertisers achieve better ROI. For example, an online education customer used Agent Builder to create their AI agent, inject key insights such as product descriptions and subject matter expertise, and provide continuous feedback on Baidu's improvements. The agent significantly enhanced the company's online customer service by providing around-the-clock, high-quality consulting, and increased ad conversion rates by 20%.
AI assistants will be the main form of content and service in the new AI era. Baidu will continue to improve profitability, increase the conversion rate and ROI of advertisers, and also promote an increase in transactions directly generated on the Baidu platform. This transformation should help Baidu transform the traditional CPC model into a more effective CPS model.
Q2: How are some Baidu's price reduction initiatives affecting the AI cloud business? As competition intensifies, how profitable is Baidu Cloud? What should the customer's level of sustainable growth be?
Answer: GAI and the big model transformed the cloud computing industry from general computing to AI computing. Baidu's AI cloud revenue declined from a year-on-year decline in the third quarter of last year to an 11% increase in the fourth quarter of last year, and then accelerated to a 12% growth rate in the first quarter of this year. In the first quarter of this year, GAI and basic models accounted for 6.9% of AI cloud revenue. Traditional cloud businesses are also being supported using GAI and basic models.
Baidu continued to focus on achieving high-quality revenue growth by reducing low-profit businesses this quarter. Regarding the GAI and basic model business, the market is still in a very early stage. Baidu's focus is still on the education market and expanding its influence in more enterprises. It is expected that business-related AI profit margins will increase further and be higher than traditional cloud businesses.
Competitors' pricing policies are changing, and it's actually quite common for cloud vendors to adjust pricing for certain products. This is a trend we have observed many times in the past. Baidu's cloud products have expanded from traditional CPU clouds to high-value AI products and services. Therefore, changes in the industry have minimal impact on the development of the AI cloud business on CPU cloud pricing.
Q3: Given the shortage of chips, how can Baidu maintain its delivery promise, differentiate value, and strengthen its lead in large models?
Answer: Baidu is working on a unique 4-tier AI architecture stack to optimize the cost and performance of models to ensure that customers and developers can easily build apps using tools such as Agent Builder, App Builder, and Model Builder.
Baidu recently made a breakthrough by integrating GPUs from different vendors into a large computing cluster, which can use less advanced chips for efficient model training and influx. Baidu's deep learning framework Pedal helps reduce model training and costs through continuous innovation and enhancements. Baidu's deep learning framework Pedal is compatible with more than 50 different chips, and its developer community has grown to 13 million.
Baidu is making AI more accessible and affordable through lightweight LLM and model development toolkits, and is confident to become a leader in China's AI ecosystem.
Q4: What is the main drag on ad growth compared to peers? What trends have you seen in ad budgets and advertiser sentiment, particularly in the second quarter, April, and May. How should normalized ad growth be considered in 2024?
Answer: Baidu's online marketing revenue increased 3% year over year in the first quarter, and weak macroeconomics led to a weak advertising business. Advertisers mainly come from small and medium-sized enterprises from all walks of life and are highly sensitive to the macro environment. Advertiser sentiment remained weak in the first quarter in verticals such as real estate and franchising. Entering the second quarter, there was no improvement in advertisers' sentiment, and online marketing revenue growth is expected to weaken. Despite the challenges, online marketing is expected to remain Baidu's main business.
Q5: Regarding autonomous taxis, can management provide this year's developments, strategic initiatives, and geographical coverage of services? Furthermore, with regard to the goal of achieving break-even operation of UE in Wuhan mentioned earlier, can you explain the strategy for continuously optimizing UE? What is the expected fleet size for this year? How will these measures impact costs?
In 2023, Apollo made significant progress in improving the economic efficiency of regional units in key cities. Let me use Apologos in Wuhan, our largest operating base to explain how we did this.
Since the commencement of commercial operations in 2022, the economic efficiency of our operating units has continued to improve, mainly due to the expansion of the scale of driverless services and the reduction in bicycle costs.
Currently, the fleet of fully autonomous vehicles in Wuhan has grown to 300 vehicles, a threefold increase from a year ago. The operating area and service hours have also been significantly expanded due to the government's recognition of our technology, and the service has covered more than 7 million people. Our operating hours have been extended to 24 hours a day, 7 days a week. These expansion measures have increased the number of passengers and distance traveled by bicycles per day, thereby increasing revenue.
In terms of cost control, we have increased the deployment of fully automated driving services by showing consistent safety operation records. The proportion of fully driverless orders has risen to 70%, and is expected to reach 100% soon, which will help us reduce costs associated with safety personnel.
Meanwhile, large-scale production of our sixth-generation autonomous taxi RT6 is underway, and each vehicle (not including batteries) is expected to cost less than $30,000. The RT6 will be the main model for our future fleet expansion, helping to reduce bicycle depreciation costs, further enhance unit economic efficiency, and drive us to profit.
Looking ahead to this quarter, we plan to expand the fleet of fully autonomous vehicles operated in Wuhan to 1,000 vehicles by the end of the year, an increase of more than three times over the end of last year. Our focus remains on improving the economic efficiency of regional units, reducing Apologo bike losses, while continuing to improve operational efficiency and reduce costs. We believe Apollo will first achieve a break-even balance in the economic benefits of operating units in Wuhan. Once this is achieved, we will be able to rapidly scale up our operations.
editor/tolk