Regarding AI, Baidu's founder, chairman, and CEO Robin Li's attitude is clear: "Baidu aims to be the first company to rebuild all its products using AI."

It's not about integration or integration, but about redoing and reconstructing. Subsequently, in 2023, Baidu released over 10 AI-native applications, including search, maps, cloud storage, and document libraries.

Since 2024, Baidu has accomplished several significant achievements in the AI field. First, it launched three lightweight models, ERNIE Speed, Lite, and Tiny, to reduce model prices. Then, it introduced three development tools, enabling everyone to use AI at a low cost. Thanks to the reduction in costs, as of mid-April, the number of developers in Baidu's PaddlePaddle community reached 13 million. What changes have these breakthroughs brought to Baidu? And how has the secondary market responded to this? Behind Baidu, how should China's internet giants explore the integration of AI and reality?

Collaborating with Apple, Samsung, Tesla, Baidu AI continues to see good news.

As early as 2017, Baidu's layout in the field of artificial intelligence already covered areas such as voice recognition, natural language processing, image recognition, and intelligent driving. At that time, Baidu's AI was more of a vision, assumption, and expectation.

After years of development, today's Baidu has closely integrated AI with users' daily lives. According to data released at Baidu's VAI Summit in May 2024, currently, 11% of Baidu's search services are generated by AI, with over 140 million library AI users and over 1.5 billion AI feature uses. In the words of Baidu Vice President He Junjie, "This is an era where everyone has the opportunity to get a ticket to the boat - everyone can participate and everyone can create."

However, in the current Chinese market, although countless users hold "tickets" and want to board the era's giant AI ship, it is not easy to own a "ship" that can navigate smoothly in the Chinese market. Due to the special linguistic context of Chinese, there are fewer corpora for spoken and dialectal languages, making it difficult to form a multimodal corpus. Moreover, mainstream foreign large models are trained based on English corpora. Take ChatGPT as an example; in the training data, the proportion of Chinese corpora is less than one-thousandth, while the proportion of English corpora exceeds 92.6%.

This leads to issues of "discomfort" in language habits and expressions when overseas AI enters the Chinese market, and Baidu happens to be able to solve this problem, thus becoming a "bridge" for overseas giants to enter the Chinese market.

In January 2024, Samsung released the Galaxy S24 series, introducing AI features such as real-time translation of calls and text messages, AI photo editing, and search triggered by circling images. In the United States, these services are provided by Samsung's own AI engine and Google. In China, Samsung chose to collaborate with Baidu and Meitu Xiuxiu. Among them, Baidu is responsible for handling "instant search," text summarization, and other AI functions.

In April 2024, Baidu Apollo released Baidu Maps V20 and simultaneously announced a collaboration with Tesla, with Baidu Maps' lane-level navigation debuting globally on Tesla. From actual test videos, lane-level navigation realizes functions such as 3D lane navigation, real-time traffic lights, camera reminders, lane crossing reminders, bus lane reminders, speed limit reminders, and intersection live navigation.

In June 2024, Apple proposed the concept of "Apple Intelligence" at the WWDC conference and reached a collaboration with OpenAI. Soon after, Apple was reported to be actively seeking cooperation opportunities with local Chinese enterprises, aiming to further promote the implementation and popularization of its "Apple Intelligence" service in the Chinese market. After several rounds of negotiations, it was finally reported that Baidu would provide AI functionality for Apple's iPhone 16, Mac system, and iOS 18 released this year.

Collaborating with international giants such as Apple, Samsung, and Tesla, on the one hand, represents the affirmation of Baidu's long-term data and technological accumulation in the Chinese context by international giants. Previously, a Samsung spokesperson stated that the company chose to collaborate with Baidu because Samsung believes Baidu provides the most competitive commercialized large-language AI model in the Chinese market.

On the other hand, Baidu's localized solutions can leverage Baidu's existing traffic advantages, truly increasing Baidu AI's channel entry points, providing possibilities for Baidu's revenue growth, and bringing broader imagination space to market capitalization.

The products have been reconstructed with AI, but the valuation is reluctant. What are the reasons?

However, compared to the continuous good news on the news front, Baidu's performance in the secondary market has been unsatisfactory. Since 2024, Baidu's share price has cumulatively declined by 25.61%; on June 24, Baidu's share price fell to $88.58, dipping to $87.66 during the overnight session, hitting a new low since November 2022.

Combined with the financial report, in Q1 2024, Baidu's total revenue was 31.5 billion yuan, with a net profit of 7.011 billion yuan, an increase of 22% year-on-year. Among them, Baidu's core business and online marketing business revenue were 23.8 billion yuan and 17 billion yuan, respectively, both exceeding market expectations. Despite exceeding expectations in performance and the continuous good news on the AI front, Baidu's valuation still hasn't been supported. Why is that?

From a business perspective, Baidu's basic disk is still mobile ecosystem business represented by search advertising. However, with the emergence of apps like Toutiao, Douyin, and Xiaohongshu, personalized recommendation content has gained more user time, and passive information reception has gradually replaced active information search. Users' ways of obtaining content have become more vertical rather than singular.

Baidu's basic disk, which relies on survival, has encountered significant challenges in the mobile internet era. AI has always been seen by Baidu as a trump card to turn the tide. In its latest financial report, Baidu emphasized that "the steady progress of generative AI technology has become a key driver of the company's business." However, judging from Baidu's first-quarter net profit, the 22% year-on-year increase (7.011 billion yuan) relies more on cost control and non-operating item contributions rather than substantial growth in the main business.

From the perspective of revenue composition, Baidu's core business revenue was 23.8 billion yuan, representing a year-on-year increase of 4%, maintaining steady growth. Among them, online marketing revenue was 17 billion yuan, representing a year-on-year increase of 2.7%. On the one hand, this growth rate has slowed compared to previous years, reflecting the intensification of market competition and the adjustment of advertiser placement strategies. On the other hand, this growth rate is also slower than that of AI-driven Google and Meta. According to the latest quarterly financial reports, Google's search advertising revenue was 62.659 billion US dollars, representing a year-on-year increase of 14%; Meta's advertising revenue reached 35.635 billion US dollars, representing a year-on-year increase of 26%.

In terms of non-online marketing revenue, Baidu achieved 6.8 billion yuan, representing a year-on-year increase of 6%. This growth is mainly due to the rapid development of the intelligent cloud business. As an emerging business area for Baidu, intelligent cloud is gradually winning market share with its technical strength and innovation capabilities. However, the intelligent cloud business also faces fierce market competition and pressure from technological iterations. Domestic and foreign cloud vendors are enhancing their cloud business competitiveness through AI. To this end, Baidu invested 5.4 billion yuan in R&D in the first quarter and needs to continue increasing investment in the future to maintain its competitive advantage.

In addition, as another important scenario for Baidu AI's landing, Baidu's autonomous driving service, Apollo Go, provided approximately 826,000 rides in Q1, representing a year-on-year increase of 25%. The cumulative number of rides provided has exceeded 6 million. However, from a revenue perspective, Apollo Go is still in a loss state and is expected to reach the break-even point next year. Tianfeng Securities previously released a research report stating that with the further enrichment of road test data, the areas and population covered by Apollo Go's business are increasing, and Baidu's intelligent driving technology is expected to usher in a paradigm shift moment.

From this perspective, Baidu's AI has not yet demonstrated intuitive driving force in terms of performance in the short term. It will take time for AI to save Baidu.

How to better integrate with AI, Chinese technology giants still need to work hard.

In fact, the problems faced by Baidu also plague all AI enterprises. Almost all Chinese and American technology giants are actively exploring and promoting the commercialization of AI. For them, AI is a standard underlying capability of technology giants and a "trend" they want to seize as much as possible.

According to ITJuzi data, in terms of the number of investments, Tencent and Baidu have invested in 87 and 84 AI industry companies, ranking first and second globally. Alibaba, ranking fourth globally, has also invested in 65 companies, second only to Google. From this perspective, Chinese companies are hoping to catch up with deficiencies in underlying computing power and chips by increasing investment in AI companies.

However, from the perspective of overseas technology giants, we can see another outcome.

Taking several US stock market giants as examples, before the AI wave arrived, companies like NVIDIA, Microsoft, Apple, Google, and Tesla already had their respective software businesses and hardware carriers, such as NVIDIA's CUDA and chips, Microsoft's Windows and PCs, Apple and Google's systems and mobile phones, Tesla's FSD system and new energy vehicles, robots, and so on.

Therefore, when the AI wave came, they had more to do with "reinforcement," selectively improving model algorithms and technologies based on their existing resources. This can also be verified from the number of mergers and acquisitions:

Taking Apple as an example, from 2024 to the present, Apple has made a total of 15 mergers and acquisitions, primarily of small technology companies, and the companies acquired are often highly related to their main business, either improving old businesses or opening up new ones, overall proceeding from business and strategy.

From this perspective, Chinese and American technology giants have given two different answers to the problem of AI commercialization: Chinese companies prefer equity investment and ecological cooperation to control AI companies, thereby maximizing the alignment of AI with their own businesses; while American companies have a stronger willingness to merge and acquire, both supplementing their original businesses and expanding their AI commercialization footprint through mergers and acquisitions.

Which of these two approaches is superior? Currently, we cannot give an accurate answer. But it is certain that even without advanced process chips, Chinese technology giants still have opportunities to catch up. After all, China now has many startup companies and AI developers exploring the landing of large models on the C-end and B-end. In vertical fields where large models are combined with industries, an industrial ecosystem for promoting and replicating industry-wide large models is being formed;

In the main C-end market, such as large models + office, large models + entertainment, large models + information acquisition, thousands of companies have flooded in, forming an AI economic miracle after the internet economy. In other words, we are likely to gain the upper hand in the industry application landing of large models.

It's not hard to imagine how the reconstruction of existing AI technology into apps like TikTok, WeChat, and Office will change our lives and work scenarios.

AI will eventually change everything, but first, it will change ourselves.

Source: Hong Kong Stock Research Society