Baidu's search traffic has long been eroded by short videos and social platforms. To emerge from its current predicament, Baidu requires not just adjustments to its business model but also strategic restructuring.

@ New Technological Knowledge Original

Once soaring high, Baidu now seems to hover at low altitudes.

As a pioneer in the AI field, Baidu's share price surged to a high of $160 in early 2023, fueled by the AI concept. However, as market enthusiasm waned, its valuation gradually declined, and its share price has now plummeted, losing approximately 44%.

More concerningly, Baidu's advertising business, which sustains its operations, is crumbling. In Q1 of this year, its advertising revenue was surpassed by JD.com, reminiscent of being overtaken by Taobao and ByteDance a few years ago. Nowadays, Kuaishou's advertising revenue is closing in on Baidu's, with a growth rate several times higher, making it only a matter of time before it overtakes Baidu.

The latest financial report reveals that Baidu has seen revenue declines for two consecutive quarters, with advertising revenue experiencing negative growth for two straight quarters. Market patience with Baidu is waning. The P/E ratio has fallen from over 20 times a year ago to less than 11, even lower than that of many consumer stocks (e.g., Haidilao's P/E ratio is 20, and Pop Mart's is 74).

International capital giants have successively issued warnings, successively lowering their performance expectations for Baidu and their target stock prices.

However, amidst adversity, Baidu is attempting to find a breakthrough through a commercial 'self-rescue'. At the recently concluded 2024 Baidu AI Marketing Conference, Baidu announced its intensification of AI marketing and the upgrading of its business system to 'Baidu Banfei,' boldly claiming that it will unlock 10 times the commercial space in the next 2-3 years.

Caught between internal bottlenecks and external competitors, can Baidu's 'Banfei' plan help it soar, or is it just a futile struggle? The answer to this game between technology and commerce may be gradually unfolding.

01 AI Dividends Uncaptured

In 2023, the AI wave propelled Baidu to a share price high, but with the increasing emergence of commercialization dilemmas, Baidu is gradually showing signs of fatigue. The rapid development of AI technology, originally seen as an opportunity for Baidu to take off again, has failed to bring the expected dividends.

As a leader in the AI era, why can't Baidu please the capital market?

First, the advent of AI search indeed allows users to reach their destination faster, but this efficiency enhancement conflicts with Baidu's advertising model, on which it relies.

Traditional search engines profit through the 'keyword bidding advertising' model. When users enter keywords, advertisers pay for top ad placements, and these ads often need multiple exposures in search results to obtain clicks. The premise of this model is that users will spend time browsing multiple search results, thereby increasing the probability of ads being clicked.

However, the goal of AI search is to directly provide the most relevant answer, reducing the time users spend on the search results page. Currently, Baidu's AI search accounts for 20% of total searches, directly returning answers through intelligent Q&A. For example, when a user searches for 'which brand of dishwasher to buy,' traditional search might require clicking on multiple results for comparison, while AI search might directly provide a ranking recommendation or a summary of price ranges, eliminating the need for users to browse other information or click on ads.

The shortened user retention time directly weakens advertising effectiveness, which is evident in the data. In the third quarter of 2024, Baidu's online marketing service revenue declined by 6% year-on-year. In stark contrast to Baidu's declining online marketing revenue, platforms like Kuaishou and Bilibili have achieved robust double-digit growth in advertising revenue.

Currently, users are increasingly spending their time on short video and social media platforms, while comprehensive platforms like Xiaohongshu and WeChat Search, which retain users through rich content, are growing rapidly. Advertisers are also more inclined to advertise on these channels.

Secondly, although Baidu's AI technology has certain strengths, it lacks irreplaceability. Wenxin Yiyan has positioned Baidu among the first tier of domestic large models, but in an era where large models are ubiquitous, it does not exhibit a significant leading edge compared to Douyin's Doubao Assistant, Tencent's Hunyuan Large Model, or Alibaba's Tongyi Large Model.

In November 2024, Douyin Doubao's monthly active users (MAU) reached 59.98 million, with a year-on-year increase of 16.92%, while Baidu's Wenxiaoyan only had 12.99 million MAU, with a growth rate of only 3.33%. Additionally, KIMI from Moonlit Side is hot on its heels with 12.82 million MAU.

Industry expectations for performance improvements in large models are also cooling, and the diminishing marginal benefits of large-scale training are making AI commercialization even more challenging.

The transformation of the advertising model has failed to match the efficiency gains of AI search, and the technological advantage has not established a sufficiently wide moat. With traffic being divided and dividends difficult to realize, Baidu urgently needs to firmly grasp the construction of a deep ecosystem.

02 Unbreakable Profit Deadlock

On the path of AI commercialization, Baidu is not lying down but sprinting full steam ahead. However, the results show that despite launching a series of applications and a new business system, Baidu has yet to touch the crucial point of truly changing the market landscape.

In the past year, Baidu has concentrated its efforts on launching over 100 industrial applications, such as Free Canvas, AI Entity Top 100 List, Miaoda, etc., attempting to empower various industries through AI. Although these applications cover fields such as manufacturing, energy, transportation, and finance, with Baidu Intelligent Cloud helping enterprises fine-tune 33,000 large models and develop 770,000 enterprise-level applications, they are still some distance from becoming 'super apps' that change users' perceptions.

Robin Li has made it clear that Baidu's goal is not to launch a 'super app' but to help more enterprises create 'super useful' applications. However, although this B2B-focused strategy has achieved certain results in the industrial sector, such as Baidu Intelligent Cloud becoming the largest cloud service provider for the industrial implementation of large models in China, there are still no killer products in the C-end market that generate strong user stickiness.

As of November 2024, the daily average invocation of Baidu's Wenxin Large Model has reached 1.5 billion times, but the primary demand is concentrated in the industrial sector. Applications at the consumer level, such as the Baidu App and Wenxin Yiyan APP, have not yet generated explosive traffic effects.

In comparison, super apps on other platforms appear to have more potential. For example, Douyin Doubao Assistant's monthly active users have exceeded 51.3 million, more than four times that of Baidu Wenxin Yiyan. While Baidu has the advantage of application richness, it still needs to make more breakthroughs to ensure that these applications directly address user pain points and truly integrate into daily life.

At the recently held 2024 Baidu AI Marketing Conference, Baidu launched 'Baidu Banfei' with great fanfare, announcing that it will unlock 10 times the commercial space with its new business system. This system includes a series of AI marketing tools such as Qingduo, Brand BOT, and Qingge, aiming to drive advertising creativity and placement through large models. However, whether advertisers will buy into this remains to be seen, given Baidu's inherent deficiencies in content ecosystem and traffic scale.

Baidu's search traffic has long been eroded by short videos and social platforms, which achieve more efficient advertising conversions with AI-generated content and full-site promotion. For example, Kuaishou's advertising revenue increased by 27.4% year-on-year in Q3 2024, reaching 16.7 billion yuan, and it is highly likely to surpass Baidu's advertising revenue in the next quarter.

Although Baidu Qingduo has the capability to generate advertising materials through AI, compared to short video platforms with larger content libraries and higher user retention times, Baidu's advantage is not significant.

Moreover, the price war in AI large model applications also poses profitability challenges to Baidu's business system. For example, Baidu Intelligent Cloud's main models have seen significant price reductions or are even offered for free, which, although increasing API invocations in the short term, has had a limited boost to overall revenue.

03 Difficulty in Adhering to Long-term Principles

The AI industry is facing a hard reality: high costs and inefficient monetization models are eroding industry patience.

According to Sequoia Capital, in 2023, the AI industry's cost on NVIDIA chips alone reached $50 billion, while the overall revenue was only $3 billion. This means that even with technology innovations driven by large models, AI enterprises still struggle to break out of the 'high investment, low return' cycle.

Baidu is also deeply entrenched in this quagmire. Currently, the business model of large models still primarily relies on traditional API invocation fees and subscription systems, but these models are obviously insufficient to support broader market demands. For example, although Baidu's Wenxin Large Model has achieved a daily average invocation of 1.5 billion times, most of this demand is concentrated in enterprise services rather than daily consumer use.

Meanwhile, there is a growing voice in the industry that the Scaling Law is coming to an end, and the stagnation of large model performance improvements further compresses future profit expectations.

Technology giants such as Tencent and Alibaba are also exploring the B2B market, attempting to integrate AI technology into government services, finance, and other industries. However, whether judged by user scale or revenue share, there is still a long way to go before reaching a mature business model.

On the other hand, the value of AI lies not in how many technological highlights it can generate but in whether it can be embedded into users' lives and change public habits. Baidu's problem is that it still hasn't shaken off its past shadow in terms of content and users.

Recently, Baidu's commercialization has also encountered troubles. Collaboration with rumored major client Apple seems less than ideal, encountering obstacles when adjusting Baidu's AI model for Siri.

Baidu's management is also relatively cautious and lacks boldness at this time, making it difficult to send a clear signal to the market.

For example, OpenAI officially opened its video generation large model Sora to users yesterday. Due to its popularity, the Sora website was temporarily overwhelmed. However, Baidu CEO Robin Li stated that 'Baidu will not do Sora' because the investment cycle is too long, and business benefits may not be realized in 10 or even 20 years. No matter how popular it is, Baidu will not do it.

In contrast, domestic internet giants have already been following up on the layout, especially ByteDance and Kuaishou, which are leading the video generation race. Once again, Baidu has lost its early advantage of 'getting up early'.

For a long time, Baidu has been controversial due to its pay-per-click ranking and content quality issues. From the pay-per-click ranking scandal in 2008 to the Weizexi incident in 2016 to the widespread dissemination of 'Search Engine Baidu is Dead' in 2019, Baidu has repeatedly fallen into public opinion vortices. These incidents have not only damaged Baidu's brand image but also greatly reduced users' trust in Baidu's content.

In the mobile internet era, Baidu attempted to establish a content ecosystem through Baijiahao, but its over-reliance on content supply from its own platform further decreased users' evaluation of search quality. A large amount of Baijiahao content is mixed in quality, repetitive, and inefficient, running counter to users' expectations of quickly obtaining high-quality information from search engines.

The popularization of AI applications ultimately relies on high-quality data and user trust. For Baidu to achieve a true commercial breakthrough, it first needs to rebuild user confidence in its content ecosystem. From questions about 'paying for placement' to criticism of 'declining search quality,' Baidu needs not only to repair its image but also to win back user trust through authentic and reliable content.

Perhaps what Baidu needs is a complete redefinition, transforming from a mere technology provider to a true connector in users' lives; from a content platform builder to an activator of content ecosystems; from an explorer of advertising monetization to a driver of business value through applications and services.

While Baidu was indeed the pioneer in leveraging AI technology, technological dividends are ephemeral, and the true determinant of success or failure lies in the user and market dynamics. To extricate itself from its current challenges, Baidu requires not merely an adjustment in its business model but also a strategic overhaul. Technology serves merely as a tool, with the genuine core resting in innovating content and crafting a value experience that is irresistible to users. Capturing the hearts and minds of users is paramount in Baidu's commercial transformation journey.

By focusing on user needs and redefining its content ecosystem through technological innovation, Baidu can uncover the key to re-engaging with the market. Concurrently, it must proactively shoulder greater social responsibility, setting an industry benchmark and striving towards genuine inclusive value.