Data as of September 11, 2026. Baidu is one of the strangest large-cap technology stocks in the market today. The company that once looked like China’s closest equivalent to Google is now trying to convince investors that search advertising is no longer the right lens through which to value it. In the second quarter of 2026, total revenue fell 4% year over year to RMB31.3 billion. That headline looks weak. Underneath it, however, Baidu’s AI Cloud Infrastructure revenue grew 50%, GPU Cloud revenue surged 283%, and the company’s broader AI-powered business reached RMB12.5 billion.
That contrast is the entire investment case. Baidu is not a clean growth story. It is a business in which an old engine is shrinking while a new one is scaling rapidly. The stock, around $91 on September 10 and with a market capitalization near $31 billion, is priced as if investors remain deeply skeptical that the transition will create enough durable profit to offset the erosion of traditional online marketing.
I think that skepticism is understandable. It is also exactly what makes the stock interesting. The market is not asking whether Baidu has AI technology. It clearly does. The question is whether that technology can become a business large enough, profitable enough and defensible enough to change how the entire company is valued.
The headline revenue number hides two different companies
Baidu reported second-quarter revenue of RMB31.3 billion, down 4% from a year earlier. Baidu General Business generated RMB25.2 billion, while iQIYI contributed RMB6.3 billion. If this were the only information available, the stock would look like a mature internet company slowly losing relevance.
But Baidu now separates what it calls its AI-powered business from legacy operations. That distinction matters. AI-powered business revenue reached RMB12.5 billion in Q2, up 25% year over year and equal to half of Baidu General Business revenue. Legacy business fell to RMB10.4 billion, down 23%.
In other words, the company is already near the point where the new business is as important as the old one. The transition is not hypothetical anymore. It is visible in the revenue mix.
GPU Cloud growth is the number that changes the story
AI Cloud Infrastructure revenue reached RMB7.3 billion in the second quarter, up 50% year over year. Inside that segment, GPU Cloud revenue increased 283%. That followed 184% growth in the previous quarter.
This is important because cloud infrastructure is one of the clearest ways Baidu can monetize the AI boom without relying on consumers to pay directly for a chatbot. Chinese companies need compute. They need access to accelerators, model training infrastructure, inference capacity and enterprise AI tools. Baidu has spent years building a full-stack ecosystem around chips, cloud infrastructure, models and applications.
The strategic advantage is integration. A company using Baidu’s cloud can access models, development tools and infrastructure from the same provider. That resembles the logic behind the largest US hyperscalers. The risk is also similar: capital intensity.
GPU Cloud growth sounds extraordinary, but serving that demand requires expensive hardware, data centers and power. Revenue growth alone does not tell us whether returns on capital will be attractive. Investors need to watch gross margins, depreciation, capital expenditures and utilization rates. A cloud business can grow very quickly while still destroying value if capacity is overbuilt or hardware becomes obsolete too fast.
Baidu is becoming an AI infrastructure company before the market believes it
At this stage, the most important mental shift is to stop treating Baidu only as a search engine. Search is still economically important, but the company is building an AI infrastructure layer that could be used across enterprise applications, autonomous driving, consumer services and advertising.
The business resembles a portfolio of AI options. Cloud is the most visible commercial engine. AI applications are a second layer. Autonomous driving is a third. AI-native marketing may become a fourth as search itself changes.
This creates complexity. It also creates optionality. A single product does not need to carry the valuation if several businesses become economically meaningful.
The problem: traditional advertising is still deteriorating
The bullish AI narrative should not obscure the weakness in Baidu’s old business. Legacy revenue fell 23% year over year in the second quarter. Online marketing has been under pressure as Chinese consumer demand remains uneven and user behavior shifts toward short-form video, social commerce and AI interfaces.
This matters because advertising historically produced attractive cash flow that funded Baidu’s research ambitions. If that profit pool erodes faster than AI businesses scale, the company could spend years in a transition with limited earnings growth.
There is also a strategic threat from generative AI itself. Search engines once monetized a list of links. AI assistants answer questions directly. That can reduce the number of commercial pages a user visits and change how advertising inventory is created.
Baidu therefore faces the classic innovator’s dilemma: it must disrupt the product that still finances much of the company.
AI-native marketing may be the bridge
Baidu reported RMB2.6 billion of AI-native marketing services revenue in Q2, approximately flat year over year. At first glance that looks unimpressive. I would not dismiss it.
The long-term question is whether Baidu can redesign advertising around AI-generated answers, agents and personalized commercial recommendations. If it succeeds, the company may preserve some of the economics of search even as the user experience changes dramatically.
The advantage is distribution. Baidu App had 644 million monthly active users in June. The company does not need to acquire every user from scratch. It can introduce AI features into products that already have enormous reach.
The challenge is monetization without damaging trust. AI-generated answers that feel like advertisements will be less useful. The best model may be one where commercial intent is detected naturally and advertisers pay for measurable actions rather than simple clicks.
AI Applications are growing more slowly, but they matter strategically
Revenue from AI Applications was RMB2.5 billion in Q2, up only 3% year over year. This bucket includes products such as Baidu Wenku, Baidu Drive and enterprise AI tools. The growth rate is far below Cloud, but the economics may eventually be better because software layers require less incremental capital than infrastructure.
Baidu said AI daily-active-user penetration across Wenku and Baidu Drive increased 27.4% year over year in June. That is a useful adoption signal. The next question is whether higher engagement translates into higher subscription revenue, enterprise contracts or stronger retention.
One of the most important changes in AI is that models themselves may commoditize faster than applications. If multiple models become good enough, value shifts toward distribution, proprietary data, workflow integration and customer relationships. Baidu already owns those assets in China.
Apollo Go is becoming a real operating network
Apollo Go is the part of Baidu that attracts the most futuristic headlines. The robotaxi platform has expanded far beyond small pilot projects. By Q2 2026, Baidu said Apollo Go’s global footprint had reached 28 cities and its fleets had accumulated more than 350 million autonomous kilometers, including more than 240 million fully driverless kilometers.
The international expansion is notable. Apollo Go began open-road testing in London with Uber and Lyft, launched fully driverless commercial operations in Dubai, received permits for fully driverless testing in Hong Kong and tested in Switzerland with PostBus.
This is no longer just a Chinese technology demonstration. Baidu is trying to become a global autonomous-mobility platform.
The obvious comparison is the robotaxi debate around Tesla. In our Tesla Cybercab analysis, the key issue is whether autonomy can move from technology milestone to scalable economics. The same test applies to Baidu. Miles driven are impressive, but investors ultimately need to know revenue per vehicle, fleet utilization, cost per ride, maintenance expense and capital intensity.
Robotaxis could be worth a lot — or almost nothing
Autonomous driving creates one of the widest valuation ranges in modern technology. If robotaxis operate at scale with low intervention rates, high utilization and attractive unit economics, a network can become enormously valuable. If regulatory constraints, hardware costs or low utilization prevent scale, billions of research spending may never produce adequate returns.
Baidu’s advantage is operational experience. Hundreds of millions of autonomous kilometers generate data that newer entrants do not have. The company also has partnerships that can reduce the need to build every local customer channel itself.
The risk is that robotaxi economics are intensely local. A system that works in Wuhan may not immediately work in London or Dubai. Regulation, road design, insurance, vehicle standards and labor costs all differ.
I therefore treat Apollo Go as an option inside the Baidu valuation rather than the core source of value today. That option becomes more valuable as commercial deployments prove repeatable economics.
The balance sheet gives Baidu time
One reason the transition remains investable is financial flexibility. Baidu is not a speculative startup that needs external funding every year. It has a long-established internet business, significant cash resources and the ability to fund AI investment internally.
The company also continues returning capital to shareholders. Baidu reported $259 million of repurchases since the beginning of Q1 2026 under its current buyback program.
Buybacks matter when a company trades at a depressed valuation, but only if management does not need that capital for higher-return investments. In Baidu’s case, the trade-off is particularly important because AI infrastructure and autonomous driving are capital hungry. Investors should want both discipline and ambition.
At roughly $31 billion, what is the market actually pricing?
At around $91 per ADR, Baidu’s equity value was roughly $31 billion in early September. That is a surprisingly modest valuation for a company with hundreds of millions of users, a rapidly growing AI cloud business and one of the world’s largest autonomous-driving programs.
The low valuation tells us the market does not trust the headline AI story. Investors are discounting at least four risks: the erosion of advertising, heavy AI capital spending, China-specific governance and geopolitical risk, and uncertainty over whether new businesses will ever earn margins comparable to the old search franchise.
I think that skepticism is rational. But valuation changes the question. Baidu does not need to become the Chinese equivalent of every US AI leader simultaneously. It needs AI Cloud and applications to grow enough that the shrinking legacy business stops determining consolidated earnings.
A simple scenario framework
| Scenario | AI business | Legacy business | What it means |
|---|---|---|---|
| Bear | Growth slows sharply | Continues double-digit decline | Transition absorbs cash without restoring growth |
| Base | 20–30% growth | Declines but becomes smaller | Revenue mix shifts enough to stabilize the company |
| Bull | Cloud and applications compound above 30% | Stabilizes through AI monetization | Baidu is rerated as an AI platform rather than a search relic |
The base case does not require perfection. If AI-powered business compounds at a healthy rate while legacy revenue shrinks into a smaller portion of the company, consolidated growth can eventually turn positive again. At that point, the valuation narrative changes.
China risk is part of the price
No Baidu analysis is complete without discussing China. US investors own ADRs in a company operating under Chinese regulation, and the technology sector remains strategically sensitive. Data, AI models, autonomous driving and advanced chips all intersect with national-security policy.
That creates a permanent discount relative to comparable US technology companies. The discount may narrow or widen, but it should not be assumed away.
At the same time, China is one of the world’s largest digital economies. A company with Baidu’s user base, enterprise relationships and local infrastructure has opportunities that foreign competitors cannot easily replicate.
The right approach is not to ignore geopolitical risk or exaggerate it. It is to demand a valuation that compensates for it.
Competition may be the bigger operational risk
Baidu competes with Alibaba, Tencent, ByteDance and a growing ecosystem of Chinese model developers. Cloud customers can switch providers. Consumer AI products can gain and lose attention quickly. Autonomous-driving competitors include technology companies, automakers and specialized startups.
The company’s defense is its full stack: infrastructure, models, distribution, applications and autonomous-driving data. But full-stack strategies can also become expensive if every layer requires world-class investment.
I would watch whether Baidu increasingly focuses capital on businesses with clear commercial traction rather than trying to win every AI category.
What I would monitor over the next four quarters
Five metrics matter most to me. First, AI Cloud Infrastructure growth: does it remain above 30% as the base becomes larger? Second, GPU Cloud utilization and margin: does explosive demand translate into economic returns? Third, legacy marketing decline: does the contraction slow? Fourth, AI application monetization: does user adoption become revenue growth? Fifth, Apollo Go commercialization: do international deployments produce repeatable economics rather than only impressive mileage statistics?
If three of those five move in the right direction, the probability of a successful transition rises materially.
Why the stock can rerate without spectacular total revenue growth
Turnarounds are often misunderstood because investors focus only on consolidated revenue. The mix can matter more. A company replacing low-confidence, declining revenue with higher-growth strategic revenue can deserve a higher multiple even before total sales accelerate.
Baidu’s AI business now represents roughly half of General Business revenue. If that share moves toward two-thirds while Cloud remains a major growth engine, investors may begin valuing the company on forward AI earnings instead of backward-looking search declines.
This is similar to what happens when an industrial company shifts from hardware toward recurring services. The same dollar of revenue can be worth more if its growth, retention and margin profile improve.
The strongest bull argument
The bullish case is not that Baidu will suddenly become China’s Nvidia, Google and Waymo combined. That is too simplistic. The strongest argument is that the market currently values the company as a deteriorating internet franchise while AI-powered businesses have already become half of General Business revenue.
If Cloud growth remains strong, if AI applications monetize gradually and if Apollo Go proves commercially viable in even a subset of its markets, the sum of those pieces could justify a materially higher enterprise value.
The strongest bear argument
The bearish case is equally straightforward. AI infrastructure may grow quickly but at lower margins and much higher capital intensity than search advertising. Legacy marketing may continue shrinking. Robotaxis may require years of investment before generating meaningful profit. In that world, Baidu replaces a high-margin business with several lower-return businesses and the apparent transformation creates little shareholder value.
That is the risk investors should focus on. Revenue transformation alone is not enough. Return on invested capital must eventually improve.
My view on Baidu stock
Baidu is one of the more asymmetric large Chinese technology stocks because expectations are already low. Around $31 billion of market value, investors are not paying a premium for the AI narrative. They are demanding proof.
I would not buy the stock simply because GPU Cloud grew 283%. Extraordinary growth rates from a small base can fade quickly, and AI infrastructure is expensive. I would buy it only with the thesis that Baidu’s AI-powered business can become the dominant economic engine before the legacy advertising franchise deteriorates too far.
That thesis is becoming more credible. AI-powered business reached RMB12.5 billion in Q2. AI Cloud Infrastructure grew 50%. Apollo Go has moved from local experiment to international deployment. The pieces are real.
The missing piece is profitability at scale.
If Baidu demonstrates that the new AI businesses can produce durable cash returns, the stock may eventually stop being valued as a declining search company. At that point, the biggest surprise would not be faster revenue growth. It would be a higher multiple.
Sources
- Baidu Investor Relations — Q2 2026 Results
- Baidu Investor Relations — Q1 2026 Results
- StockAnalysis — BIDU market price history
This article is independent analysis, not investment advice. Valuation scenarios are illustrative and not price targets.
