Report contents
Industry landscape
Where the company competes: Chinese search ads, now a six-way contest for attention
Baidu runs an auction for Chinese-language search results and the advertising attached to them, and that auction is no longer the whole company. General Business, the reporting unit that holds search, AI cloud and consumer apps, sat alongside iQIYI, which Baidu reports on its own line, in the third-quarter 2026 results the company published on November 16, 2026. Online marketing fell to 48% of General Business revenue from 62% a year earlier; the AI-powered line rose to 53% of the same base, on unaudited internal management accounts. The two shares overlap, because the AI line contains AI-native marketing, so the split describes a shifting mix rather than two clean halves.
Upstream, Baidu depends on content it can index, on GPU capacity, and increasingly on its own silicon through Kunlunxin. Downstream sit users who supply attention and advertisers who bid for it. The auction itself carries almost no marginal cost per query, which is where Baidu's historical margin came from; the AI buildout is capital-heavy and the cost sits with whoever owns the chips. When Baidu has to buy distribution rather than earn it, traffic acquisition costs take a cut of the same ad yuan.
The competitive boundary is national, not global. Search in China is licensed and content-regulated, and Google's 2010 exit removed the one rival with global scale, so Baidu's share never faced a foreign challenge. That license does not exclude domestic entrants, though, and it never was the real barrier. Index scale and a liquid advertiser auction were, and neither protects a company when users type their queries somewhere else.
Baidu now meets a different set of rivals in each arena. ByteDance takes search and feed advertising through Douyin and Toutiao; Tencent through WeChat; Alibaba through Quark; Xiaohongshu (Little Red Book) through community search. In AI assistants, Questmobile put peak daily active users during the 2026 Spring Festival at 145 million for ByteDance's Doubao, 74 million for Alibaba's Qianwen and 41 million for Tencent's Yuanbao, figures Morningstar analyst Chelsey Tam cited on March 10, 2026; Bloomberg reported that Baidu's Ernie trails DeepSeek and Moonshot. In AI cloud infrastructure, Baidu AI Cloud faces Alibaba Cloud, Huawei Cloud and Tencent Cloud. Dated public share estimates for Chinese search ad spending are thin, so the ranking rests on user and product evidence rather than a clean percentage.
Barriers diverge sharply by arena. Consumer attention looks easy to enter: a short-video app or a chatbot can capture search intent without building a crawler, an index or an ad sales force, and bundling does the distribution work. Six credible contenders for the same Chinese user is well past the point where entry barriers usually hold. AI cloud is the opposite: data residency rules, enterprise procurement cycles, switching costs and GPU capex gate entry, and four players carry most of the domestic market.
Where the moat question turns, then, is whether Baidu's strongest ground, enterprise cloud and compute, is big enough to matter against a consumer franchise whose walls have already been crossed.
Competitive advantage
Does the company have a moat? Weak, and only in cloud switching costs
Overall rating: Weak. Baidu still owns the default Chinese-language search box, but ownership no longer protects a return. The decisive evidence is profitability: estimated after-tax ROIC of about 2.5% in FY2025, 5.1% in FY2024 and 5.9% in FY2023, all well below the 9.7% WACC Morningstar used when it cut Baidu's economic moat rating to none from wide in March 2026. A company earning under its cost of capital is not being shielded from competition, however familiar its name. What survives is narrower and newer: switching costs inside enterprise AI cloud, and a domestic GPU business in Kunlunxin. Neither is yet big enough to lift the consolidated return, which is why the call here is Weak instead of None.
For most of the past decade Chinese search advertising behaved like a natural monopoly. Crawling, indexing and ad serving are fixed costs, a query costs almost nothing to serve, and advertisers gravitate to the deepest auction, so the leader's unit costs fell as it grew. The fixed cost has since moved. GPU capacity is now the industry's large capital commitment, and the companies pulling search intent away from Baidu (ByteDance, Alibaba, Tencent) already carried that cost for video, commerce and social. Rivals did not have to outspend Baidu to match its scale. They had the scale first and pointed it at search.
Market position: first in web search, one of six where queries start
Market position: Moderate. Baidu remains the largest Chinese general search engine, and the evidence does not show another company displacing it from that specific position. It is no longer the gateway, though. Chinese users now search inside Little Red Book, social apps, short video and AI chatbots, Morningstar's Chelsey Tam wrote in March 2026. Ad budgets follow that attention, and Baidu's own results show the cost: non-AI general business revenue fell 20% year over year in the fourth quarter of 2025.
Stability: the revenue trend runs one way
Stability of market position: Weak. Baidu no longer reports online marketing revenue separately, so the ad franchise has to be tracked through the general business line, and a clean five-year share series cannot be built from the filings. The usual rule of thumb is that a five-point drift in average share over five years means entry is not barred; the disclosed revenue trend points past that threshold, in the direction of loss. Morningstar's downgrade is an independent second read on the same evidence, and it moved two notches at once.
Profitability: returns below the cost of capital
Profitability: Weak. The returns Baidu reported over FY2021–FY2025 never reach the 15%–25% band that signals a durable advantage, and they fall sharply in the final year.
| Fiscal year | ROE (%) | ROA (%) | Gross margin (%) | Net margin (%) |
|---|---|---|---|---|
| FY2021 | 4.9 | 2.9 | 48.3 | 8.2 |
| FY2022 | 3.3 | 2.0 | 48.3 | 6.1 |
| FY2023 | 8.1 | 5.1 | 51.7 | 15.1 |
| FY2024 | 8.7 | 5.7 | 50.3 | 17.8 |
| FY2025 | 2.0 | 1.3 | 43.9 | 4.3 |
The ROIC estimate shows why the equity returns are thin. For FY2025, operating income of RMB 10.37 billion taxed at the 18% effective rate implied by the year's provision gives NOPAT of roughly RMB 8.5 billion, estimated. Invested capital, meaning equity of RMB 266.3 billion plus RMB 93.0 billion of debt less RMB 24.6 billion of cash, was about RMB 334.7 billion. That is where the 2.5% comes from; the same method gives about 5.1% for FY2024 and 5.9% for FY2023.
Two features of the balance sheet make that measure harsh on the legacy ad business. The denominator carries RMB 36.8 billion of goodwill, and it consolidates the entire AI buildout, which is capital-hungry by design and still early. Even allowing for both, the direction is the finding: returns fell as the ad mix thinned, and the capital going into AI has not yet earned anything close to the group's cost of capital.
Margins: scale that is not spreading costs
Margins: Moderate. Revenue was RMB 124.49 billion in FY2021 and RMB 129.08 billion in FY2025, roughly 1% a year, so Baidu is not buying growth with price. The mix inside that flat line changed. Gross margin fell about eight points between FY2023 and FY2025, and net margin dropped from 17.8% in FY2024 to 4.3% in FY2025, a year that included a quarterly net loss. On that revenue base, one point of gross margin is worth about RMB 1.3 billion, so the compression cost roughly RMB 10 billion of gross profit, close to twice the year's net income of RMB 5.59 billion. The comparison that can be made is with Baidu's own history, and AI infrastructure is adding fixed costs faster than the ad auction sheds them.
Profitability decides the rating
Market position is Moderate and margins are Moderate, but profitability carries the extra weight and it is Weak: below the cost of capital in each of the three years estimated, with the return falling rather than rising. That combination caps the overall rating. The strongest argument for None is that the legacy auction probably still earns high returns on its own; the group figures cannot separate it from the AI spend, and the revenue base that generates those returns is the part that is shrinking.
Where the advantage comes from and how long it can last: cloud contracts outlast the ad auction
Three things are worth testing in this industry: whether an enterprise will pay to avoid tearing out an AI stack it has already built, whether a larger query base still lowers cost per query, and whether Baidu's own silicon gives it a cost edge rivals cannot simply buy. Only the first shows protection today.
The number that tests the whole position is AI-powered revenue growth measured against the decline in non-AI general business revenue, together with what that growth costs in capex. Morningstar's March 2026 model assumes the non-AI general business compounds down at 11% a year over 2026–2030 while AI-powered businesses grow 23%. On the cash side, FY2025 free cash flow was negative RMB 16.43 billion against positive RMB 13.1 billion in FY2024; in the June 2026 quarter the gap between operating cash flow of RMB 3.44 billion and free cash flow of negative RMB 7.95 billion implies capex of roughly RMB 11 billion in three months, estimated. Growth is being purchased before it is being earned.
Demand-side: switching costs hold renewals, not new workloads
Demand-side advantage: Moderate. An enterprise that builds models and data pipelines on Baidu AI Cloud faces months of re-certification, staff retraining and procurement work to move, and Chinese data-residency rules keep most of those workloads domestic. That is genuine captivity, and Morningstar credits the AI enterprise cloud with switching costs. The limit is size and scope: the cloud is a minority of revenue, which is why Morningstar judged the captivity insufficient to support a moat.
The protection is also asymmetric in a way that matters for a challenger. Switching costs defend accounts already installed; they do nothing for the next workload, because nothing is locked in before the contract is signed. Baidu AI Cloud is the smaller of four domestic platforms alongside Alibaba Cloud, Huawei Cloud and Tencent Cloud, so most of its future revenue must come from placements that are still contestable. If inference pricing turns into a sustained price war, Baidu's installed base survives and its growth does not. That would show up as rising capex against flat or falling AI cloud revenue, and it is the outcome to follow.
Economies of scale: the auction keeps its economics, the queries leave
Scale advantage in search: Weak. The search barrier was never the technology. Fixed spending on crawl, index and ad systems paid off only at volume, the deepest auction won the advertisers, and that produced the roughly 50% gross margin of FY2023 and FY2024. The mechanism required users to start at Baidu. When a short-video app or a chatbot answers the same need inside an app the user already has open, the fixed cost stays with Baidu and the query does not, so fewer queries spread over the same systems and sales force raise unit cost. The 20% fall in non-AI general business revenue in the fourth quarter of 2025 is that arithmetic showing up in the accounts.
Generative answers add a second squeeze. An AI summary occupies the space where a page of links and ads used to sit, so even queries that stay at Baidu may monetize worse than they did. The advantage here is now a liability with a slow decay rate, and the test is whether the non-AI decline narrows toward single digits as AI-native marketing is folded into the line, or keeps compounding.
Supply-side: Kunlunxin is real, its moat is not established
Supply-side advantage: Moderate at most. Designing its own accelerators lets Baidu cut cost in its own cloud and insulate supply from export controls, and Morningstar describes Kunlunxin as a first-tier GPU company in China by revenue scale and customer quality, while noting that disclosure is limited ahead of its IPO. The difficulty is that chip design is a cost and performance edge, not customer captivity. A faster part wins a benchmark and can lose it back the next cycle, Chinese buyers of accelerators have other domestic options, and Baidu's own software stack does more to retain developers than any single chip generation. If Kunlunxin is listed separately, Baidu's economic interest becomes a stake rather than an integrated cost advantage, which weakens the link to the parent's returns.
Moat strength and the two questions that decide it
Ranked against the usual hierarchy, Baidu's strongest remaining asset is partial customer captivity in enterprise cloud, the auction's scale has flipped from protection to burden, and the chip business is a cost edge of the kind that is easiest to copy. Brand is the false moat in this story: Baidu is one of the best-known names in China, and recognition did not hold the revenue line. A substitute here is not a better search engine but a different place to type the question, and it costs the user nothing, so price cannot defend the position.
What would raise the rating is AI cloud revenue growing faster than the non-AI general business shrinks, with group ROIC back above the cost of capital for more than a single year. What would lower it to None is another year of double-digit decline in the general business while capex keeps climbing. Two questions decide the outcome: does non-AI general business revenue stop falling before AI capex stops rising, and can Baidu AI Cloud win new workloads, not just renew existing ones, once inference price competition starts?
Generated by DeepVal AI. Data and models may contain errors or be outdated. For research only; not personalized investment advice.
Continue reading the full report
- Comparison and summaryHow Baidu Inc compares with its closest rivals, and our moat rating
Free to sign up · 20 research points · unlocking this report uses 10
20 points cover 2 full reports. No card needed.
Research reflects its generation date. Preview the first two modules.
More on Baidu Inc
Baidu Inc
Business, moat and key questions
Open company page