Baidu: Deeseek becomes 'Soul Ferry'? Life and death still depend on yourself

$Baidu(BIDU.US) released its Q1 2025 earnings after the Hong Kong market closed on May 21, slightly exceeding expectations overall. Simply put, DeepSeek was a big help!

Details (focusing on Baidu Core):

1. DeepSeek offered at the lowest price online, driving explosive growth in Smart Cloud: Baidu was the first platform to fully capitalize on the DeepSeek boom. After DeepSeek went viral, Baidu quickly integrated it and launched a major promotion, offering the lowest-priced computing power online. This directly boosted Smart Cloud growth from 28% last quarter to 42% this quarter.

2. User growth "a long drought ends with sweet rain": Beyond cloud, mobile Baidu traffic also benefited from DeepSeek. In mid-February, the full version of DeepSeek was integrated into Baidu Search, revitalizing user engagement. Mobile Baidu's monthly active users surged by 45 million sequentially, reaching a record high of 724 million.

3. Advertising underperformed, relying on AI search monetization to fill the gap: Advertising revenue declined 6% YoY, as expected. While this aligned with company guidance, given that the macro environment in Q1 wasn’t as bad as anticipated, some institutions had slightly raised their expectations for Baidu’s ad performance last month. This made the actual results disappointing from an expectations perspective.

This further indicates that the new traffic was primarily engaged with AI search, with traditional ads unable to benefit directly. Baidu plans to gradually commercialize AI search in Q2, so management’s rollout plan and milestones will be key to watch.

4. Improving economic model in innovation businesses: Gross margin fell 7 ppts YoY, mainly due to shifts in business mix. However, a rough breakdown shows that other businesses continued to improve margins, likely due to factors like Apollo Go’s switch to RT6 and accelerated cloud growth.

5. AI-driven internal efficiency gains continue: While Baidu hasn’t yet reaped much external benefit from AI, internal efficiency gains have been ongoing for nearly a year. In Q1, aside from increased sales promotions, R&D expenses continued to decline, reflecting AI’s impact on operational efficiency—a trend also seen in other R&D-heavy internet platforms.

6. Capital expenditure resumes expansion: The current expansion should be viewed positively. Baidu’s capex has lagged behind other internet giants, and last year, to protect profits and cash flow, it even deliberately scaled back. This was also tied to Baidu’s stockpiling of AI chips and in-house chip development. However, in Q1, possibly due to preemptive buying amid chip restrictions, capex rebounded with 40% growth.

7. Share buybacks increase—is Baidu finally loosening up?: Q1 free cash flow was negative due to higher AI-related operating expenses and short-term debt repayment. Net cash (excluding debt) stood at RMB 133.7 billion ($18.6 billion). Baidu Core repurchased $445 million in net cash in Q1, continuing the sequential increase. This brings cumulative buybacks under the 2023 plan (3 years, $5 billion) to $2.1 billion. The pace remains slow unless the company never intended to fully execute the plan.

However, with a longer quiet period in Q1, if buybacks normalize to over $500 million per quarter for the rest of the year, annual repurchases could nearly double to $2 billion. At the current market cap of $30.7 billion, this would represent a 6.5% yield. For a low-growth core business, this return is decent—assuming execution aligns with expectations.

8. Detailed earnings data

Dolphin Research’s View

With low expectations for ads, Q1 hinged on whether cloud could deliver. Thanks to DeepSeek’s 助攻 and promotions, AI Cloud came through.

Valuation-wise, Baidu’s $29 billion market cap, net of $19 billion in net cash, implies just $10 billion for the core business. Against 2025’s RMB 18 billion net profit, the PE is just 4x.

While Chinese internet stocks are cheap, Baidu’s valuation is jaw-dropping. But is it a value trap or opportunity? For now, Dolphin Research believes that until Apollo Go and Smart Cloud show clearer progress, it’s still a value trap.

First, ads—Baidu’s reluctant cash cow—remain critical. Management emphasized accelerating AI search monetization, but per the call and follow-ups, this is still in very early stages, requiring step-by-step progress.

This implies that while Q1 ads were weak, Q2 won’t see a rebound—in fact, Q2-Q3 could worsen.

As for Apollo Go, Dolphin Research has long noted that L4 models are capital-intensive. Running in one street versus city-wide are two different things, and scaling nationally is another challenge. Even if successful, it’s a long-term, heavy-investment business—short-term valuation support remains more narrative-driven.

In this context, cloud’s "significant" beat may attract some capital inflows. But Baidu’s underlying competitiveness and long-term growth potential remain unconvincing based on Q1 alone.

Thus, the only near-term support is buybacks: $450 million in Q1. If annual buybacks hit $2 billion (with $2.9 billion remaining under the plan—one of the few Chinese companies likely to leave unused quota), the current buyback yield would be 7%.

As a no-growth value stock, this is middling versus peers like Vipshop and JD. Overall, a clearer business inflection point is needed, and Dolphin Research remains cautious.

Detailed Earnings Analysis

Baidu is rare among internet firms in splitting earnings into:

1. Baidu Core: Traditional ads (search/feed) and innovation businesses (Smart Cloud/DuerOS/Apollo);

2. iQiyi: Membership, ads, and content licensing.

The two segments are clearly delineated, and with iQiyi as a separate listed entity, Dolphin Research breaks both down. Due to ~1% inter-segment eliminations (RMB 200-400 million), Baidu Core sub-metrics may slightly differ from reported numbers but don’t affect trends.

I. Is AI Search Monetization a "Breakthrough" or "Drinking Poison to Quench Thirst"?

Baidu Core ads fell 6.1% YoY in Q1, in line with guidance. But given the macro environment wasn’t as bad as feared, this still disappointed slightly raised institutional expectations.

On ecosystem erosion, mobile Baidu traffic rebounded sharply with DeepSeek’s help, adding 45 million users. But the new traffic is sticky to AI search, leaving traditional search ads unable to benefit.

In standalone AI chatbot traffic, Ernie lags further behind. Among peers, DeepSeek and Yuanbao surged in Feb-Mar, but by April, only Doubao remained dominant. Beyond Doubao—which benefits from massive traffic and entertainment use cases—standalone platforms struggle to sustain growth without existing ecosystems like Baidu Search or WeChat.

Baidu will commercialize AI search in Q2. The rollout plan and milestones are critical for growth expectations.

II. Smart Cloud’s Surge: Powered by DeepSeek’s Lowest-Price Strategy?

Within Baidu Core’s non-ad businesses, ~70% of revenue comes from Smart Cloud; the rest is mainly autonomous driving and smart speakers. Q1 other revenue hit RMB 9.4 billion, up 40% YoY, accelerating sharply. Smart Cloud grew 42% with DeepSeek, far above the 25% consensus.

When DeepSeek went viral in February, Baidu not only integrated it swiftly but also launched aggressive pricing—free for half a month, then the lowest rates online. In mid-March, Baidu released Ernie 4.5 (its first multimodal model) and inference model X1, with leading benchmarks and even "prettier" pricing than DeepSeek.

Baidu AI Cloud’s acceleration shows strong enterprise demand for AI, with Baidu enjoying a growth 红利 as a "shovel seller."

Beyond Smart Cloud, other businesses (Xiaodu, smart transportation, autonomous driving) grew 35% YoY, partly from YY’s consolidation (~RMB 500 million revenue).

Among standalone innovations, autonomous driving stood out:

Apollo Go completed 1.4 million rides in Q1, up 75% YoY (vs. 36% in Q4), accelerating due to improved unit economics, more vehicles, and new regions. In April, it was approved for open-road testing in Hong Kong.

It has also expanded overseas to Dubai and Abu Dhabi, with road tests in Dubai starting in May.

III. AI-Driven Internal Efficiency; Innovation Business Models Improve

Group gross margin fell to 52% YoY due to mix shifts. Despite lower-margin other revenue rising 2 ppts QoQ, gross margin held flat, indicating improving unit economics in innovations, especially autonomous driving.

Capex reflects Baidu’s AI server/chip investments. Q1 marked a clear expansion, following Q4’s sequential uptick. While chip stockpiling may have contributed, DeepSeek’s revolutionary model development—slashing trial costs—clearly spurred demand, prompting upstream "shovel sellers" to ramp up supply.

On opex, AI’s internal efficiency gains persist, especially in R&D headcount. Q1 R&D expenses fell 16% YoY, with SBC down 36%—partly from a 15-20% 市值缩水， but also likely headcount optimization.

Baidu Core’s Q1 operating profit was RMB 4.2 billion (16% margin), down 3 ppts YoY, roughly in line. Business mix shifts will long pressure group profitability; until ads stabilize, profits will lag sales. The 拐点 hinges on AI search monetization—though user backlash is a risk.

For value-focused investors (ignoring sum-of-parts，看重现金流/ROE), this profit erosion is painful. Without significantly larger buybacks, Baidu’s appeal diminishes.