Baidu's "Desperate Measures", Can It Lead to a Rebirth?

$Baidu(BIDU.US) $BIDU-SW(09888.HK) The second quarter results were released after the Hong Kong stock market closed on August 20, 2025. Overall, they slightly exceeded expectations, but the current performance is largely priced in. The key lies in future guidance!

Specifically (focusing only on Baidu Core) :

1. Intelligent Cloud expected to grow at a high rate: Cloud business is currently a hot topic, and high growth is a given. However, without the surprise of the previous quarter, the 27% growth in the second quarter aligns with most market expectations, and the positive impact has long been factored into the valuation.

2. Advertising in a painful transition period: The pains of transition are becoming evident, with AI-generated content penetration in search results now at 64%, much more aggressive than Google's 26% penetration in the TOP20 search results (according to seoClarity). Meanwhile, monetization plans are temporarily on hold, prioritizing user experience, a "startup spirit" rarely seen on a mature commercial platform like Baidu.

Despite the sacrifices, there have been gains—Baidu App's MAU has continued to grow this year, reaching 735 million in the second quarter, with a net increase of over 10 million. Whether or not it justifies the current "sacrifice," the trend is at least improving. This is also evidenced in Google's case, where AI Overview has brought an overall increase in traffic in the short term.

Another secondary factor affecting Q2 advertising is the consolidation of YY, which requires internal offsetting of the original approximately 300-400 million quarterly YY advertising expenses, impacting growth by about 2 percentage points.

Since the company's proactive transformation has the greatest impact, future outlook depends on the company's pace. Currently, the market does not have high expectations for Q3, with the main focus on whether Q4 can rebound, benefiting from AI search, similar to how Google benefits from AI's enhancement of search volume, ad conversion efficiency, and the release of new inventory.

3. Apollo Go's overseas expansion: In the second quarter, Apollo Go made several moves, mainly targeting overseas markets, collaborating with Uber and Lyft in Asia, the Middle East, and Europe, with plans to deploy thousands of Apollo Go autonomous vehicles in the coming years. However, it will take at least until next year to see a significant impact on the group's performance.

4. Cost control under AI efficiency: The poor performance of advertising, which contributes significantly to profits, has a very noticeable drag on current profits (down 40% year-on-year), but it's not as bad as the market thought. The decline is mainly reflected in gross profit (due to both revenue decline and increased AI costs), while operating expenses in Q2 were actually kept under control.

Sales expenses grew by 2%, possibly related to promotional expenses for AI product launches in the second quarter, while R&D expenses continued to decrease by 14%, mainly due to optimization of traditional R&D staff under AI efficiency.

5. Capex expansion keeping pace: Compared to other giants in the same period, Baidu's capital expenditure is not only significantly smaller in scale but also shows a different trend. The company explains this by citing the accumulation of investments started 10 years ago and partial substitution by its own Kunlun chips.

However, starting this year, due to the group's full-stack AI transformation, Baidu's capital expenditure has resumed expansion, accelerating by 80% in the second quarter to 3.8 billion, although the absolute scale still lags behind peers. Regarding capital expenditure expansion, during this AI transformation window, Dolphin Research has always tended to interpret it positively.

6. Buybacks reduced again: The sudden increase in buybacks in Q1 gave Dolphin Research some hope, but the buyback pace slowed again in Q2, with the scale of buybacks this quarter only half of Q1. If buybacks continue at the Q2 pace in the second half, the annual buyback yield will only be 3.7% (=11.5/306).

Although the overall valuation of Chinese stocks has risen this year, reducing buyback motivation and lowering shareholder returns, for Baidu, which is currently under performance pressure, whether buybacks can support valuation confidence is particularly important.

7. Detailed financial report data overview

Dolphin Research's View

From the perspective of expectation deviation, the second quarter results are basically on target. The "shock" in advertising has already been priced in, and the "highlights" in profitability may bring some rewards, but clearly, the more critical aspect is whether a signal of an advertising inflection point can be provided in the short term.

However, qualitatively speaking, Baidu's aggressive transformation should not be interpreted solely negatively. On the contrary, if it can lead to a revival, regaining user reputation and mindshare, it is a solution of short-term pain for long-term gain. Baidu's new AI search framework is somewhat similar to Google's AI Mode—shifting from "keywords + links" to a "Prompt + task" multimodal content interaction approach.

Therefore, in conjunction with some of Google's actions, Dolphin Research is also reflecting on a question: Ultimately, does AI disrupt search or improve it?

The debate over whether Google's search volume increased in the second quarter has been suppressing Google's stock performance. But there is no doubt that the absolute scale has increased. One interesting driver is that to avoid errors caused by the high hallucination rate of native AI applications like GPT, users have added a new action of verifying results through Google search in actual use.

In this view, the search form still needs to be retained (displaying source links or related content retrieval), but the experience can be more vivid, whether it's text, images, videos, or other multimodal end-to-end input and output, or a clearer and more user-friendly result presentation. Currently, the new Baidu is exploring, and it remains to be seen whether this "burning bridges" move can bring miraculous effects.

But returning to rational investment, during a period when performance is under pressure but the inflection point is unclear, how to value is a big question.

1) If viewed as a whole group, with the pain in the advertising business, the value of Baidu's core business is no longer undervalued, with a market cap of 30 billion implying a P/E of over 15x this year. However, the innovative business itself is mainly loss-making, and for Baidu, which is already full-stack AI (basic large models, applications, cloud, chips), this valuation method is clearly disadvantageous and needs to be evaluated by segments.

2) Under SOTP, the most conservative approach is to assume the company holds onto search and waits for decline, eventually distributing dividends in a breakup style (the actual amount is questionable), which is low PE advertising (5x P/E, estimated 15 billion USD * 90% group discount = 13.5 billion USD) + net cash (15.4 billion USD * 90% low dividend discount = 14 billion USD) = approximately 27.5 billion USD market cap.

3) But making this assumption now is also inappropriate, not to mention whether self-revolution can ultimately bring "rebirth," Baidu Cloud is still in the spotlight, and Apollo Go's overseas expansion is not without highlights in the future. These new businesses may attract growth-focused funds, and if mainstream fund styles can shift, Baidu will at least not return to the stage of having no stories to tell in previous years.

Therefore, in this case, we base it on advertising (13.5 billion USD) + cloud (assuming 28 billion revenue this year, 3x EV/Sales * 90% group discount = 10.5 billion USD) totaling 24 billion USD for the two core businesses, with net cash at a higher discount of 80% (under normal development, Baidu will continue to maintain a low dividend state), corresponding to a total of 35 billion USD. It is worth noting that if Baidu continues to choose the current low shareholder buyback execution, the net cash discount needs to be further increased.

And the additional 3 billion USD for Apollo Go (assuming 10,000 vehicles in 2028, 3x EV/Sales, discounted back to 2025 as 3 billion USD * 90% group discount) is seen as an upward option in a high sentiment environment. From the perspective of certainty and cost-effectiveness, before the advertising inflection point is clearly reached, the best way to play Baidu is to leave a safety margin based on one's own risk preference while maintaining range trading.

Below is a detailed interpretation of the financial report

Baidu is relatively rare among internet companies in breaking down its performance into:

1. Baidu Core: Covers traditional advertising business (search/information flow advertising) and innovative businesses (intelligent cloud/DuerOS Xiaodu speakers/Apollo, etc.);

2. iQIYI Business: Membership, advertising, and copyright sublicensing, among others.

The separation of the two businesses is clear, and with iQIYI as a separately listed company providing detailed data, Dolphin Research also breaks down the two businesses in detail. Due to approximately 1% (2-4 billion) offset items between the two major businesses, the detailed data of Baidu Core split by Dolphin Research may slightly differ from the actual reported figures, but it does not affect trend judgment.

I. Advertising in a painful transition

In the second quarter, Baidu Core advertising declined by 15.4% year-on-year, basically in line with guidance and market expectations. The decline is due to multiple factors, including external reasons such as competition and environmental pressure, with the aggressive AI transformation of its search business being the main influencing factor.

Therefore, the pace of recovery is in the company's hands, and it is recommended to pay attention to how the outlook is presented in the conference call. Currently, market expectations are low, but there is still focus on "whether Q4 will rebound" and "how much it will rebound."

Although the comprehensive AI transformation is under short-term commercialization pressure, some positive signs are also visible, with Baidu App's MAU having continuously net increased this year.

Looking purely at AI chatbot terminal traffic, apart from Doubao having a clear relative advantage, other platforms either experience short-term fluctuations under heavy promotion or continue to slow down. In terms of embedded AI applications, each platform mainly relies on the traffic advantage of their original products.

II. Intelligent Cloud expected to grow at a high rate

In Baidu Core's other businesses (non-advertising), nearly 65% of revenue comes from Intelligent Cloud, with the remaining 35% mainly from autonomous driving technology solutions, smart speakers, etc., and the second quarter also includes YY live streaming revenue (estimated 800 million to 1 billion based on company communication).

In the second quarter, other business revenue was 10 billion, growing by 34% year-on-year. Among them, Intelligent Cloud grew by 27%, basically meeting market expectations. Other businesses outside of Intelligent Cloud grew by 15% year-on-year.

Although AI Cloud met expectations, Dolphin Research originally had higher expectations for Baidu AI Cloud based on the accelerated growth trend of Tencent Cloud. When advertising is under pressure, it is very much needed for AI Cloud to exceed expectations and stand out.

The most outstanding part of its innovative business is still the intelligent driving business:

In the second quarter, Apollo Go received 2.2 million orders, growing by 148% year-on-year, with a significant acceleration compared to the previous quarter. This should include new volumes from Hong Kong, the Middle East, and other regions that just started testing in the second quarter.

In terms of overseas business advancement, Apollo Go made several moves in the second quarter, collaborating with Uber and Lyft in Asia, the Middle East, and Europe, with plans to deploy thousands of Apollo Go autonomous vehicles in the coming years. However, it will take at least until next year to see a significant impact on performance.

III. AI efficiency brings continuous personnel optimization

Under the dual impact of revenue decline and increased AI costs, the group's gross margin continued to decline to 49.7%.

In terms of capital expenditure, Baidu's investment in AI servers and chips accelerated in the second quarter, and although the scale still lags behind, Baidu is catching up with the giants in terms of trend.

In terms of operating expenses, the impact of AI efficiency on internal operations continues, especially in traditional R&D manpower. R&D expenses continued to decrease by 13% in the second quarter, with total SBC expenses down 31% year-on-year, mainly reflecting the impact of personnel optimization.

Ultimately, Baidu Core's operating profit in the second quarter was 3.3 billion, with a profit margin of 13%, down 8 percentage points year-on-year, slightly exceeding expectations. Changes in business structure will have a long-term impact on the overall group's monetization efficiency, and as long as the advertising pressure period does not end, profit performance will continue to lag behind revenue.