The triple transformation of the three major businesses has brought about an accelerated fission of value across Baidu's entire ecosystem.
As the wave of generative AI sweeps through the world, a year is enough time to completely transform a company into a “new look.”
As the first company in the world to launch a direct benchmark ChatGPT model, the development and commercialization exploration of Baidu-SW (09888) has received major attention from the global AI community and capital markets. One year after launching “Wenxin Yime,” Baidu handed over a brilliant questionnaire.
On May 16, Baidu announced its results for the first quarter of 2024. According to the data, Baidu's revenue during the reporting period was 31.5 billion yuan, and net profit (non-GAAP) attributable to Baidu was 7 billion yuan, an increase of 22% over the previous year. Both revenue and net profit exceeded market expectations.
The reason behind the impressive performance is that after Baidu switched to prioritizing artificial intelligence with the Internet as the center, AI began to release value after completely restructuring Baidu's entire business. Among them, AI search has improved B-side delivery efficiency while creating a better product experience for C-side users. Cloud services have achieved innovative breakthroughs driven by AI, and autonomous driving enabled by AI has successfully opened up the business model. The triple transformation of the three major businesses has brought about an accelerated fission of value across Baidu's entire ecosystem.
Goldman Sachs said that Baidu is accelerating towards a revenue/profit structure centered on general artificial intelligence. As enterprise users and advertisers adopt its general artificial intelligence solutions, this will bring Baidu incremental revenue of about 3 billion to 6 billion yuan in 2024; at the same time, revenue growth from cloud services should accelerate from 2023 levels, and efficient execution of operating expenses will enable Baidu to achieve a stable operating profit margin.
Through this financial report, we can get a glimpse of Baidu's real industrial value and commercial prospects as a leading AI model company in China, and how to re-give it a matching valuation will be a question that investors need to think deeply about.
AI-enabled search business is expected to reshape the competitive landscape of the market
As the ballast stone of Baidu's entire business system, the mobile ecosystem business is known as a “cash cow”. Its steady profits and cash flow have laid a solid foundation for Baidu's marathon-style long-term high-pressure investment in artificial intelligence. As the core application of the mobile ecosystem, search is the basic platform of Baidu's mobile ecosystem. With the rise of generative AI, the search industry has ushered in a new transformation.
Search is one of the main ways for the public to obtain information. It is an important channel for human-computer interaction, so search is also a huge traffic portal. In the AI era, AI-enabled searches can generate more accurate and direct results, allowing users to meet their needs in one step. Currently, the industry believes that with the immediate characteristics of search and the excellent user experience of search enabled by AI, search is the most likely application to become a Killer App in the AI era. As a result, whoever can reshape the search landscape with AI will master new traffic entrances. Google and Meta are also using AI to carry out related transformations to explore the latest AI+ search model.
As a leader in the domestic search field, Baidu has been deeply involved in the search industry for many years, and is also the first company in China to dabble in intelligent search. After the launch of Wenxin's words, Baidu quickly completed the restructuring of the search business with an AI model. The new search product has three major characteristics: extreme satisfaction, recommendation stimulation, and multiple rounds of interaction, which not only improved the quality of search, but also broke through the scope of application of search and expanded the boundaries of the product.
Currently, about 11% of Baidu's search results are generated by AI, which can provide users with more accurate, more organized, and more intuitive answers. At the same time, Baidu has recently added an “interactive” column to the search and completely changed it to a question-and-answer search. Various actions indicate that the speed at which AI is reconstructing searches is speeding up.
According to the conference call, Baidu's AI search products continue to increase in stickiness, and Baidu currently generates more and more search results from AI, and formats are becoming more and more diverse, such as text, images, third-party links, points of interest, etc., which can meet the comprehensive search needs of users.
At the same time, the launch of the smart device can be described as an important “gripper” for Baidu to improve its search business. It not only enriches Baidu's content ecosystem, but also enhances the search experience of C-side users, and effectively improves the delivery efficiency of B-side customers.
For example, on the C-side, you only need to search for Singapore Cultural Tourism on the Baidu App. The first result is the “Singapore Tourism Board” smart device. Users can directly inquire about travel guides such as popular attractions, food highlights, etc., as well as practical information on entry and departure, local transportation, etc., highlighting the accuracy and efficiency of information exchange.
On the B-side, more and more small and medium-sized enterprises are also beginning to use Wenxin smart devices. Merchants can use natural language to create an exclusive smart device on AgentBuilder in a few words, making it a 7x24 hour online gold salesman, making it closer to users and better understanding user needs.
For example, Kai Tak Education created a study abroad consultant who knows business and students through AgentBuilder. In the first week of launch, it successfully distributed 1.55 million times and interacted with users 58,000 times. The number of leads converted increased exponentially, the conversion cost of effective leads was significantly reduced, and operational efficiency was greatly improved. Furthermore, the smart device created by the famous home furnishing brand Sophia can not only answer users' decoration requests, but also actively confirm specific needs with users and recommend offline stores, reducing the cost of effective leads by 30%.
Based on ease of use and efficient distribution efficiency, Baidu's smart devices are widely popular with customers. Up to now, more than 50,000 developers and tens of thousands of companies have joined the Wenxin Smart Platform. Currently, new smart devices are being created every day, distributed millions of times a day, to serve many vertical fields such as education, law, B2B, and tourism. In the future, as smart technology innovation continues to be iterated, user intentions will be matched with the most relevant product supply in a more natural way.
It is undeniable that AI-generated content will have a certain impact on the revenue of online marketing businesses in the short term, but investors should also recognize that AI-enabled search and the launch of smart devices have greatly improved the user experience, increased user stickiness, increased search usage and user time, and that smart devices may become “mini-programs” in the AI era. The formation of an intelligent ecosystem is expected to make Baidu a super traffic portal. Currently, Wenxin has surpassed 200 million users. Compared with the data released in December last year, its number of users has doubled.
With its first-mover advantage and innovative business model on smart devices, Baidu is expected to rapidly expand its market share in the search field, thereby reshaping the competitive landscape of the entire search industry and further consolidating the company's leading position in the search field. Therefore, from a long-term perspective, after the restructuring of AI, the fundamentals of Baidu's search business became more stable, and the intrinsic value continued to increase.
The big model drives the rapid development of cloud services, and the burgeoning demand for inference brings new long-term opportunities
As Baidu's mid-term growth point, cloud services achieved rapid growth during the reporting period. Revenue increased 12% year over year to 4.7 billion yuan. Of these, the proportion of revenue contributed by generative AI reached 6.9%, gradually expanding. Behind the impressive performance of this business, it has benefited from growing market demand and a rapid decline in the cost and threshold of Baidu Cloud service usage.
The rise of generative AI has significantly boosted the demand for the AIaaS market. This is because pre-training of large models requires large-scale computing power, and deployment to the user side for fine tuning or secondary training also requires certain computing power. At the same time, the application of large models and the migration to the end side led to strong growth in inference demand.
However, due to the concentrated release of market demand and chip-side restrictions, AI training costs and inference costs are at a relatively high level, so whoever can provide cost-effective stable computing power can quickly seize the AI cloud service market.
Faced with strong market demand and potential industry pain points, Baidu, which has been the number one in the AI public cloud service market for a long time, continuously optimizes the performance of large iterative models to reduce the cost of enterprises and developers calling the model.
After the Wenxin model was upgraded to version 4.0 in October 2023, Wenxin has reached the leading level in the industry in terms of understanding, generation, logic, and memory, and has achieved further significant improvements in general capabilities such as code generation, code interpretation, and code optimization, reaching the leading international level.
While the performance of the Wenxin large model soared, inference costs were drastically reduced. Based on the unique advantages of a four-layer technology stack, Baidu can form a closed loop of efficient feedback in chips, frames, large models, and terminal application scenarios to help large models continuously tune and iterate. Compared to a year ago, the algorithm training efficiency of Wenxin's big model was increased 5.1 times the original, the average weekly training efficiency reached 98.8%, the inference performance increased 105 times, and the cost of inference was reduced to 1%.
In order to avoid waste of resources and further reduce customer usage costs, Baidu launched the AI native application development tool AppBuilder and the model customization tool ModelBuilder in the cloud business during the reporting period, and launched 3 lightweight models and 2 specific scene models based on the 3.5 and 4.0 versions of the Wenxin Big Model flagship, forming a MoE (size model collaboration) usage method, and enterprises can call the model on the newly launched ModelBuilder, which can not only reduce costs but also improve response speed. For example, Xiaodu used this model. Ordinary questions and answers only used a small model and 4.0 for arithmetic questions. In the end, the response speed was increased by 2 times, and the cost was reduced by 99%.
It is worth noting that AppBuilder has drastically lowered the development threshold. Developers can use natural language, develop an AI native application in just three steps, and easily publish and integrate into various business environments. Take the 24-hour “Resident Assistant” launched in Zhumadian, Henan as an example. The product is based on the AppBuilder Knowledge Q&A application (RAG framework), which can accurately answer questions with regional characteristics, improving the efficiency of grass-roots grid workers serving residents by 15%, increasing residents' service response speed by 30%, and increasing residents' satisfaction by 50%.
Based on excellent performance, rapidly decreasing usage costs and usage thresholds, more and more enterprises and developers are calling or developing AI native applications based on the Wenxin Big Model, which has led to a rapid rise in the number of calls to the Wenxin Big Model. Currently, the average number of daily calls to the Wenxin Big Model API has exceeded 200 million. Compared with the more than 50 million calculations disclosed in the Q4 financial report in '23, the number of daily calls has increased 300% month-on-month. The Wenxin Big Model has become the most advanced and widely used AI basic model in China.
At the same time, well-known companies in various industries have accelerated cooperation with Baidu Intelligent Cloud based on Wenxin's big model. In this quarter's conference call, Li Yanhong revealed that after Baidu Smart Cloud cooperated with China's Samsung and Honor in the previous quarter, mobile phone manufacturers such as Xiaomi, OPPO, and Vivo all connected to the Wenxin Big Model API this quarter. The field of cooperation for large models expanded from mobile phones to PCs and electric vehicles, and Lenovo and NIO also reached a partnership with Baidu during the quarter. Li Yanhong said that the expansion to smart devices has provided ample opportunities for the Wenxin model to be adopted by larger users.
IDC believes that in the next five years, next-generation artificial intelligence driven by big models and generative AI is expected to drive the entire industry through a downward phase and return to the era of high growth. Moreover, starting this year, with the accelerated spread of smart terminals, demand for inference will explode this year and next. Dongwu Securities said it is optimistic that Baidu's AI model capacity improvement and scenario expansion will drive cloud computing revenue growth. In the future, as the model matrix is improved to explore more customer needs, the cloud business is expected to continue to contribute incrementally.
It is worth noting that Baidu GPU Cloud has been widely recognized by existing and new customers due to its high performance, low cost, and low threshold. As the number of GPU cloud customers continues to increase, Baidu can cross-sell CPU cloud services to GPU cloud customers. This is expected to allow customers to switch to Baidu public cloud, which will further open up the growth space for Baidu's smart cloud business. Predictably, the smart cloud business, as one of Baidu's new growth curves, will continue to drive Baidu's revenue growth in a steady manner.
The inflection point of the industry is approaching, and large-scale applications of Radish Run are expected to accelerate
As the strongest expression of Baidu's artificial intelligence technology, intelligent driving may become Baidu's biggest hot spot this year. Because the autonomous driving industry has entered an important inflection point, large-scale applications are expected to be launched at an accelerated pace from this year.
On April 6, Elon Musk (Elon Musk) posted on social media X announcing that “Tesla will release the autonomous taxi Robotaxi on August 8.” Tesla's entry proved the viability of autonomous driving technology. Furthermore, Musk's visit to China on April 28 shows that Tesla FSD's entry into China is almost unresolved; it is just a matter of waiting for a specific schedule. Tesla FSD's plan to enter China will surely accelerate the commercialization and application of autonomous driving technology in China.
Baidu is probably the only company in China that can compete with Tesla's driverless cars. As the first and only company in China to provide fully driverless autonomous travel services to the public on open roads, Baidu's accumulation in autonomous driving is deep enough. Since 2013, it has been deploying autonomous driving. It has continued to invest in high-pressure, marathon R&D for ten years, and is driving long-term development with technological innovation. Currently, Baidu Apollo L4 autonomous driving safety operation test mileage has exceeded 100 million kilometers, and has more than 5,000 autonomous driving patent families. Among them, the number of high-level autonomous driving patents is the highest in the world.
A fully unmanned autonomous vehicle fleet under Baidu's autonomous driving commercialization platform Radish Express has already entered the five cities of Beijing, Shenzhen, Wuhan, Chongqing, and Shanghai. As of April 19, Radish Express had completed more than 6 million trips nationwide. During the reporting period, the number of Radish Run service orders reached 826,000 times.
It is worth noting that in the Wuhan region, the number of unmanned orders for Radish Express is more than four times that of the same period in 2023. In China, only Baidu is currently capable of completely unmanned autonomous driving. Especially in complicated road conditions like Wuhan, it will take quite a while for Tesla to achieve Baidu's current level of completely unmanned operation. It will take as little as 3 years and as much as 5 years.
Also, at the Apollo Day 2024 open day on May 15, Baidu announced the latest development results of its intelligent driving business. It unveiled Apollo-ADFM, the world's first large-scale autonomous driving model that supports L4 level driverless driving applications. This is Baidu's core technical foundation for achieving city-level driverless driving. With Apollo ADFM, it only takes half a year for L4 fully unmanned travel to deploy a city from 0 to 1.
At the same time, Baidu Radish Express officially released an unmanned vehicle equipped with Baidu Apollo's sixth-generation autonomous driving system solution. Compared with the 5th generation car, the vehicle cost was directly reduced by 60%, only 200,000. The ride experience surpassed that of the Mercedes-Benz S, comparable to Maybach, and once again set a new industry record. From now on, the first batch of sixth-generation unmanned vehicles delivered will be launched in batches in Wuhan. The deployment of 1,000 unmanned vehicles will be completed in Wuhan during the year, so that more users can enjoy green, safe and beautiful travel.
With the continuous iteration of the system improving efficiency and drastically reducing the cost of sixth-generation autonomous driverless vehicles, the unit economic efficiency of Radish Run has been continuously optimized under the stimulus of the scale effect, and the potential for profit has been shown. Baidu said that Radish Express will achieve balance of payments in Wuhan by the end of 2024 and fully enter the profit period in 25 years. This will make Radish Express the first platform in the world to commercialize profits for autonomous driving.
More importantly, once Radish Fast becomes profitable in Wuhan, it means that Baidu's autonomous driving operation model has been opened up, and the viability of its commercialization has been verified to the market, then Baidu autonomous driving can accelerate its expansion to more cities, and with the support of Apollo-ADFM, it can complete the deployment of a city in just 6 months, and the large-scale application of Baidu autonomous driving will be fully launched.
In addition to autonomous driving, another commercial model in Baidu's intelligent driving segment, automobile intelligence will also usher in the accelerated rollout of ASD. Currently, ANP3, the only visual urban pilot assisted driving product in China, has been upgraded to Baidu ASD (Apollo Self-Dirving), which is the largest autonomous driving model, and will soon be mass-produced and launched on all Jiyue cars, so that smart driving can be driven all over the country, and in places with Baidu maps. In the future, mass-produced passenger cars equipped with ASD can become “safe and reliable drivers” just like Radish Run, so that more users can enjoy the safe, comfortable, and efficient travel experience brought by leading autonomous driving technology.
It can be seen from this that Baidu's diversified commercialization strategy for intelligent driving has gradually entered a harvest period. Autonomous driving is at an inflection point, and ASD for intelligent vehicles is expected to be rolled out at an accelerated pace. These are all concrete signs of Baidu's accelerated commercialization of intelligent driving. With the full empowerment of the leading domestic Wenxin model and Baidu's world-leading autonomous driving technology, the foundation of intelligent driving as an emerging growth engine for Baidu in the long term will become more and more solid.
summed
As a leader in the AI industry, Baidu Wenxin's leading model and the potential value market unleashed by the overall restructuring of its own business can be seen by all. The reshaping of the search pattern, the acceleration of cloud service growth, and autonomous driving are about to enter a profit period. Changes in the three major business lines will surely drive a revaluation of Baidu's intrinsic value, but judging from the reflection in the capital market, no timely and effective pricing has been made. This means that Baidu's true value is not fully reflected.
If you look at Baidu's value from a different perspective, it will be more persuasive. Generative AI will bring about the fourth round of industrial revolution, and it has become a market consensus. In the new era, giants of the era will definitely be born, and Baidu has become a core target that cannot be circumvented in the artificial intelligence era. The strength shown in AI already has the potential to develop into a global giant. Currently, it is only the beginning of a new era for Baidu to embrace AI, and it is also a good time to lay out Baidu in the long term.