Baidu Earnings Preview: Can AI Cloud Keep Growing After a 79% Surge?

 

Baidu reports second quarter results before the US market opens on August 18, 2026, with management holding a call at 8:00 a.m. Eastern, 8:00 p.m. Beijing time the same day. Per Baidu's investor relations announcement, the release covers the quarter ended June 30.

 

The prior quarter set a benchmark that will not be easy to repeat. Per Baidu's first quarter results, AI Cloud Infra revenue was RMB 8.8 billion, up 79% year over year, while GPU Cloud revenue rose 184%. Per Futu's breakdown, AI Cloud Infra revenue grew from RMB 4.9 billion in the first quarter of 2025 to RMB 8.8 billion, with growth accelerating from 34% to 79%.

 

But what determines the quality of this report is not the AI growth rate itself. Run one subtraction on the disclosed figures: the core AI business added roughly RMB 4.5 billion year over year while the legacy business lost roughly RMB 4.2 billion. The two nearly cancel, which is why Baidu Core revenue grew only about 2%. The question for the second quarter is whether the AI increment begins to exceed the legacy decline by a meaningful margin rather than merely offsetting it.

 

 

 

Baidu reports second quarter results before the US market opens on August 18, 2026, with the call at 8:00 a.m. Eastern and a phone replay available through August 25.

 

First quarter AI Cloud Infra revenue was RMB 8.8 billion, up 79% year over year and accelerating from 34% a year earlier, while GPU Cloud revenue rose 184% after 143% in the prior quarter.

 

First quarter total revenue was RMB 32.08 billion, down modestly year over year, while Baidu Core revenue of RMB 26.0 billion rose roughly 2%, returning to growth after several quarters of decline.

 

Core AI-powered business revenue reached RMB 13.6 billion, up 49% and exceeding half of Baidu Core revenue for the first time at 52%, while the legacy business declined 29% to RMB 10.20 billion.

 

Total AI Cloud revenue combining AI Cloud Infra and AI applications reached RMB 11.3 billion, with AI-native marketing services revenue of RMB 2.30 billion, up 36%.

 

Margin direction is split: Baidu Core non-GAAP operating income was RMB 4.0 billion, up 39% sequentially at a 12% margin, while adjusted EBITDA fell 17% year over year to RMB 5.95 billion.

 

Operating cash flow was RMB 2.7 billion, the third consecutive positive quarter since turning positive in the third quarter of 2025.

 

Apollo Go delivered 3.2 million fully driverless rides in the first quarter with triple-digit ride growth, while Kunlunxin has achieved large-scale commercial deployment in a single AI computing cluster of over 30,000 accelerators.

 

 

 

The starting point is the three-layer structure of the prior quarter. Per Yahoo Finance, first quarter revenue dropped modestly year over year to RMB 32.08 billion, about $4.75 billion, with the decline largely due to the slowdown in its legacy business where the top line fell 29% to RMB 10.20 billion; the bright spot was the Baidu Core AI-powered business, which grew 49% to RMB 13.60 billion; AI-native marketing services revenue reached RMB 2.30 billion, up 36%; and Baidu app monthly active users reached 655 million in March 2026.

 

Per the first quarter earnings call transcript, management reported RMB 26.0 billion in total revenue for the quarter, up 2% year over year and marking a return to positive growth, with core AI-powered business revenue reaching RMB 13.6 billion, up 49%, and accounting for more than half of Baidu General Business revenue for the first time at 52%.

 

Note that sources differ slightly on Baidu Core revenue growth, placing it between 1% and 2% depending on the measure, and the company's official filing is authoritative.

 

 

This is the calculation readers should verify for themselves. Core AI-powered revenue of RMB 13.6 billion growing 49% implies a prior-year base near RMB 9.1 billion and an increment near RMB 4.5 billion. Legacy revenue of RMB 10.20 billion declining 29% implies a prior-year base near RMB 14.4 billion and a decline near RMB 4.2 billion.

 

Netting the two leaves roughly RMB 0.3 billion of increment against a RMB 26.0 billion base, corresponding to growth of roughly 1% to 2% and matching the disclosed figure.

 

What that arithmetic establishes is this: the structural transition is complete in mix terms, and 52% is a genuine milestone, but it is not complete in growth terms, because the AI increment currently only covers the legacy attrition. That is the core standard for judging the second quarter, namely whether the gap between the two curves begins to widen. The calculation above derives from disclosed figures and illustrates structure rather than providing precise reconciliation, which belongs to the official filing.

 

 

 

Confirm first that the acceleration is real. Futu's breakdown shows AI Cloud Infra revenue growing from RMB 4.9 billion in the first quarter of 2025 to RMB 8.8 billion in the first quarter of 2026, with growth accelerating from 34% to 79%, and GPU Cloud growth accelerating from 128% to 184%.

 

The earnings call transcript frames it slightly differently, with management saying GPU Cloud revenue continued its strong trajectory from last quarter's 143% growth, accelerating further to 184% year over year. The two sources use different reference periods, one comparing with the prior year and the other with the prior quarter, so readers should confirm which base they are using.

 

On either measure the direction is upward rather than downward, which is uncommon for a business several quarters into its ramp.

 

 

There is an information gap here worth flagging. The company disclosed the absolute figure of RMB 8.8 billion for AI Cloud Infra but disclosed only a 184% growth rate for GPU Cloud, without an absolute figure.

 

That means the base underlying 184% cannot be determined from public information. Per Alpha Sense's summary, GPU Cloud became a meaningful contributor to total AI Cloud Infra revenue with 184% year-over-year growth, typically carrying stronger margins than traditional CPU cloud. The phrase meaningful contributor indicates it has reached scale, but it does not quantify it.

 

The practical approach is to treat 184% as a directional signal rather than an extrapolable growth rate. With the base unknown, the durability of triple-digit growth cannot be judged from that single number, and whether the second quarter discloses an absolute figure is itself a data point worth watching.

 

 

Absent a GPU Cloud dollar figure, token consumption offers another angle. Alpha Sense's summary shows the Qianfan MaaS platform expanding its model library to include popular models from Zhipu AI, MiniMax, Kimi and DeepSeek beyond ERNIE, with daily average token consumption from external customers growing to nearly seven times the level of a year ago in March.

 

Management's explanation points to a change in workload composition. Futu's breakdown notes management saying enterprise AI demand is shifting from training to large-scale inference deployment, and that the compute density required for inference far exceeds that of the training phase, driving exponential revenue growth.

 

That explanation is testable: if inference is genuinely replacing training as the dominant workload, the revenue curve should be smoother and more recurring in character than a training cycle, and that shows up in sequential data over subsequent quarters.

 

 

 

This is the dataset most needing cross-checking in the second quarter. Per TIKR's analysis, non-GAAP operating income for Baidu General Business was RMB 4.0 billion in the first quarter, up 39% sequentially, with a non-GAAP operating margin of 12%, while operating cash flow came in at RMB 2.7 billion, the third consecutive quarter of positive cash flow since turning positive in the third quarter of 2025.

 

But Yahoo Finance notes Baidu's margins remain pressured, with adjusted EBITDA declining 17% year over year to RMB 5.95 billion.

 

The two opposing figures are not contradictory: sequential improvement indicates scale effects emerging, while the annual decline indicates the cost structure relative to the prior year is still deteriorating. Judging which is more representative requires seeing whether the two lines converge in the second quarter.

 

 

The support comes from product mix. Alpha Sense notes GPU Cloud typically carries stronger margins than traditional CPU cloud, and TIKR's analysis notes GPU Cloud's 184% growth makes it a structurally higher-margin contributor to the revenue mix, with the AI-powered business crossing 52% of Baidu General Business revenue described by the CFO on the call as a mix shift that will support margin expansion as these businesses scale.

 

The drag comes from investment. Per Simply Wall St, analyst fair value estimates shifted from $105.77 to $101.60 as analysts factored in weaker online advertising trends along with higher AI investment needs, with a mix of both reduced and raised Street targets ranging from $124 to $205.

 

For investors tracking Chinese ADRs alongside other assets, cross-market relative strength often signals shifts in capital preference earlier than any single sector narrative, and on platforms covering multiple asset classes such as MEXC that comparison is more direct.

 

 

 

 

This thread is easily underweighted. Alpha Sense's summary shows Kunlunxin self-developed AI chips achieving strong market recognition for stability, efficiency, compatibility and versatility, with large-scale commercial deployment in a single AI computing cluster of over 30,000 accelerators, and having been optimised and validated for workloads across various models including ERNIE, DeepSeek V4, GLM-5.1 and MiniMax M2.7.

 

The composition of that list matters. ERNIE is Baidu's own model, while DeepSeek, GLM and MiniMax are all third parties. Optimising and validating for competitors' models indicates Kunlunxin is positioned not merely for internal use but as general-purpose compute hardware for external sale. That distinction materially affects the valuation framework: internally consumed silicon reduces cost, while externally sold silicon creates revenue.

 

The same summary notes Baidu's full-stack AI capabilities providing differentiated advantages spanning underlying infrastructure to applications, enabling stable and reliable compute supply while optimising end-to-end performance and cost effectiveness. Futu's breakdown describes that four-layer architecture as cloud infrastructure including Kunlunxin, the PaddlePaddle deep learning framework, foundation models ERNIE and Qianfan, and AI applications, noting the core value lies in inter-layer synergy creating a cost moat.

 

 

Apollo Go's data also needs layering. The earnings call transcript shows 3.2 million fully driverless rides in the first quarter, sustaining triple-digit year-over-year growth in total rides.

 

But the profitability path points abroad. Per BigGo Finance, Apollo Go partnered with CAR Inc. at Haikou Airport to offer a rental model where driverless vehicles wait for passengers right at the arrival terminal, while a new autonomous ride-hailing app launched in Dubai, and Robin Li noted overseas markets hold higher pricing potential and could deliver much stronger profitability than China, where ride fares are low.

 

The implication is direct: domestically, ride growth and profit growth are not synchronised, and genuine unit economics improvement depends on overseas expansion. What matters in the second quarter is therefore not total rides but progress in overseas cities and any commentary on per-ride pricing.

 

 

First, the gap between the core AI increment and the legacy decline. Those were roughly RMB 4.5 billion and negative RMB 4.2 billion in the first quarter, nearly cancelling. A materially wider gap in the second quarter would give the improvement in headline growth a real foundation.

 

Second, sequential rather than year-over-year AI Cloud Infra figures. The RMB 8.8 billion is a clear absolute value, and sequential change strips out base effects, making it the cleanest read on growth durability.

 

Third, whether GPU Cloud discloses an absolute figure. Only a growth rate exists today, and a dollar figure in the second quarter would make that line independently valuable for the first time.

 

Fourth, whether the year-over-year direction of adjusted EBITDA converges with the sequential direction of operating income. The first quarter showed the former down 17% and the latter up 39%, and a quarter of data is needed to establish which represents the trend.

 

Fifth, commentary on capital expenditure. Alongside rapid AI Cloud growth, the pace of compute investment determines free cash flow, and three consecutive quarters of positive operating cash flow is a record that needs maintaining.

 

Sixth, specific progress and pricing on Apollo Go's overseas deployment, the only route to improving its unit economics.

 

 

On volatility expectations, a historical calibration is warranted. Per StockTitan, Baidu's prior earnings-date record showed an average move of negative 0.64%, while recent insider data showed net selling as a risk factor to monitor alongside the upcoming report. The same summary notes that average derives from a two-event sample, limiting its representativeness.

 

Valuation is also not in extreme territory. Yahoo Finance notes that on a forward-adjusted basis, Baidu's stock trades at a non-GAAP price-to-earnings ratio of 14.82 times, a bit higher than the industry average of 12.98 times.

 

Together those two data points describe a structure different from other AI-linked names this week. On a stock trading near 300 times sales or up 183% over a year, both beats and misses get amplified significantly. On a stock at under 15 times forward earnings with an average earnings-day move below 1%, the tolerance band is comparatively wider.

 

To be explicit, the historical average move is not a forecast of this reaction, the sample is small, market conditions vary, and the actual response may deviate substantially from the historical pattern.

 

 

The first risk is the pace of legacy decline. Legacy revenue fell 29% year over year in the first quarter, and a wider decline would keep offsetting the AI increment, making the return to positive headline growth difficult to sustain. This is what analysts referenced as weaker online advertising trends when trimming fair value estimates.

 

The second is the balance between investment and margin. Adjusted EBITDA fell 17% year over year while analysts cited higher AI investment needs, meaning the trade-off between growth and margin persists near term.

 

The third is fading base effects. High growth rates in AI Cloud Infra and GPU Cloud rest partly on smaller prior-year bases, and as absolute scale grows a natural deceleration in year-over-year rates is mathematically inevitable. The question is the pace of that deceleration.

 

The fourth is competition and regulation. Major participants in China's cloud market are all advancing their own AI infrastructure, while autonomous driving commercialisation depends on the cadence of local regulation.

 

On scenarios, the base case is AI Cloud Infra maintaining strong growth while the year-over-year rate eases from 79%, Baidu Core revenue growth ticking modestly higher, and attention shifting toward margins and capital expenditure. A second is the AI increment clearly exceeding legacy attrition, pushing headline growth into mid-to-high single digits and allowing the valuation framework to migrate from an AI company with legacy businesses toward a growth name. A third is a widening legacy decline or further margin pressure, where a forward multiple under 15 times limits the downside but extends the time required for any re-rating.

 

Status distinctions: all first quarter financial data, operating metrics and management commentary come from the company's official releases and public earnings call records. Second quarter data is entirely unknown before August 18. Analyst targets, fair value estimates, valuation multiples and historical earnings-day statistics come from third parties, change constantly and differ between sources. Every subtraction and ratio calculation from public data here illustrates structure rather than providing precise reconciliation.

 

 

What deserves the most attention in this report is not 79% or 184% but a difference you have to calculate yourself. Core AI-powered revenue of RMB 13.6 billion growing 49% implies an increment near RMB 4.5 billion. Legacy revenue of RMB 10.20 billion declining 29% implies a reduction near RMB 4.2 billion. Netting them leaves roughly RMB 0.3 billion against a RMB 26.0 billion base, matching the disclosed growth of about 2%. In other words, the AI business is currently filling a hole rather than creating growth. The 52% share is a genuine milestone, but it measures structure rather than momentum. What the second quarter actually has to answer is whether that gap has started to widen.

 

Three misreadings look likely. The first is treating 184% as an extrapolable rate. The company disclosed RMB 8.8 billion for AI Cloud Infra but gave only a growth rate for GPU Cloud with no dollar figure, leaving the base unverifiable. With an unknown base, the durability of triple-digit growth cannot be inferred from one number, and whether the second quarter discloses an absolute figure for the first time is itself worth watching. The second is overlooking the internal contradiction in margin direction. Baidu Core non-GAAP operating income rose 39% sequentially while adjusted EBITDA fell 17% year over year, one up and one down. Sequential improvement reflects scale effects and the annual decline reflects the cost structure against the prior year, and until they converge no single-direction conclusion is secure. The third is underweighting what the Kunlunxin validation list implies. The models optimised and validated include DeepSeek V4, GLM-5.1 and MiniMax M2.7, all third parties. Supporting competitors' models indicates positioning as externally sold general-purpose compute hardware rather than internal supply, and the former creates revenue while the latter only reduces cost, with entirely different valuation consequences.

 

Three verifiable metrics deserve tracking. First, the sequential dollar figure for AI Cloud Infra rather than the year-over-year rate, since RMB 8.8 billion supplies a clean base and sequential change strips out base effects, making it the most direct read on genuine demand intensity. Second, whether operating cash flow sustains its positive streak, at RMB 2.7 billion in the first quarter and the third positive quarter since the third quarter of 2025, since that line determines whether the AI investment cycle is self-funding. Third, specific commentary on Apollo Go's overseas pricing, since Robin Li has stated overseas markets hold higher pricing potential and could deliver much stronger profitability than China, making progress on overseas revenue per ride more informative than total ride counts.

 

For cross-asset investors, the lesson is that identical GPU cloud businesses carry entirely different risk structures depending on how they are funded. This week's comparison is clean: one US neocloud runs full-year capital expenditure at roughly 2.9 times revenue with interest expense near four times adjusted operating income, another runs capital expenditure at roughly seven times revenue, while Baidu's GPU Cloud growth occurs against three consecutive quarters of positive operating cash flow, with investment supported by an existing business rather than debt. That difference is almost invisible while demand rises, but it determines who has time to wait when demand slows. Crypto knows the pattern: identical mining businesses expanded from own cash flow versus expanded through equipment-collateralised borrowing end very differently in a downcycle. When evaluating compute assets, establish the source of capital and the rigidity of repayment obligations before looking at the growth rate. That sequence sits closer to the substance of the risk than the reverse. From a risk management standpoint, on a name at under 15 times forward earnings with an average earnings-day move below 1%, position sizing is less urgent than for this week's very high-multiple names, though that equally means the scale of any single re-rating is limited.

 

This analysis rests on the company's official releases, public earnings call records and credible reporting available now. Actual second quarter data, legacy business trends and the pace of investment could each change the conclusion, and no single scenario should be treated as a fixed expectation.

 

 

 

Before the US market opens on August 18, 2026, covering the quarter ended June 30. Management holds an earnings call at 8:00 a.m. Eastern, 8:00 p.m. Beijing time the same day, with pre-registration required to receive dial-in details, a passcode and a unique access PIN. A phone replay is available through August 25, and a live and archived webcast will be available on Baidu's investor relations site.

 

 

First quarter AI Cloud Infra revenue was RMB 8.8 billion, up 79% year over year and accelerating markedly from 34% a year earlier, with revenue rising from RMB 4.9 billion in the first quarter of 2025. Within that, GPU Cloud revenue grew 184% year over year, continuing from 143% in the prior quarter. Total AI Cloud revenue combining AI Cloud Infra and AI applications reached RMB 11.3 billion.

 

 

By mix, yes. First quarter core AI-powered revenue of RMB 13.6 billion grew 49% and exceeded half of Baidu Core revenue for the first time at 52%. But the transition is incomplete in growth terms: calculated from public data, the AI increment was roughly RMB 4.5 billion while the legacy business declined 29% to RMB 10.20 billion, a reduction near RMB 4.2 billion. The two nearly cancel, which is why Baidu Core revenue grew only about 2%.

 

 

Because both directions coexist. Baidu Core non-GAAP operating income was RMB 4.0 billion, up 39% sequentially at a 12% margin, indicating scale effects emerging. But adjusted EBITDA fell 17% year over year to RMB 5.95 billion, indicating the cost structure relative to the prior year is still deteriorating. Analysts trimming fair value estimates cited both weaker online advertising trends and higher AI investment needs.

 

 

Per summaries, Kunlunxin self-developed AI chips have achieved strong market recognition for stability, efficiency, compatibility and versatility, with large-scale commercial deployment in a single AI computing cluster of over 30,000 accelerators, and have been optimised and validated for models including ERNIE, DeepSeek V4, GLM-5.1 and MiniMax M2.7. The presence of multiple third-party models on that list indicates positioning beyond internal use.

 

 

Primarily overseas. First quarter deliveries reached 3.2 million fully driverless rides with triple-digit year-over-year growth in total rides, while a new autonomous ride-hailing app launched in Dubai and Apollo Go partnered with CAR Inc. at Haikou Airport on a rental model with vehicles waiting at the arrival terminal. Robin Li noted overseas markets hold higher pricing potential and could deliver much stronger profitability than China, where ride fares are low.

 

 

Historical data does not support expecting an extreme move. Summaries show Baidu's prior earnings-date record with an average move of negative 0.64%, though that average derives from only a two-event sample and is limited in representativeness. On valuation, the stock trades at a forward-adjusted non-GAAP price-to-earnings ratio of 14.82 times against an industry average of 12.98 times. Compared with names trading at hundreds of times sales, the tolerance band is wider, though that also caps the scale of any single re-rating.

 

 

Four. First, the gap between the core AI increment and the legacy decline, roughly RMB 4.5 billion against negative RMB 4.2 billion in the first quarter. Second, the sequential dollar figure for AI Cloud Infra rather than the year-over-year rate, with RMB 8.8 billion as the base. Third, whether GPU Cloud discloses an absolute figure for the first time. Fourth, whether the year-over-year direction of adjusted EBITDA converges with the sequential direction of operating income.

 

 

This article is provided for informational and research purposes only and does not constitute investment advice, financial advice, legal advice, tax advice or any recommendation to transact. The financial data, operating metrics and management commentary referenced here come from the company's official releases and public earnings call records; second quarter data is entirely unknown before publication and no forecast of it is made here; sources differ slightly on Baidu Core revenue growth and the GPU Cloud reference base, which the text notes where applicable, and the company's official filings are authoritative; analyst targets, fair value estimates, valuation multiples and historical earnings-day statistics come from third parties that change constantly and differ between sources; and every subtraction and ratio calculation from public data in this article illustrates structure rather than providing precise reconciliation or a determination about the company's operations. Prices of crypto assets, equities and other related financial instruments can move sharply over short periods, particularly around earnings windows, investors may lose their entire principal, and American depositary shares of companies reporting in renminbi carry additional currency risk. Historical performance, technical indicators and on-chain data cannot guarantee future outcomes, and the historical average earnings-day move is not a forecast of this reaction. Readers should conduct their own independent research, verify official information directly, and evaluate any decision against their own financial circumstances, investment objectives, experience and risk tolerance, consulting a qualified professional adviser where appropriate. The MEXC Crypto Pulse team accepts no liability for any direct or indirect loss arising from use of or reliance on the information in this article.

 

 

James Mitchell specializes in technical analysis, market trends, and trading strategies for both Bitcoin and altcoins. Based in London, he has over 10 years of experience in financial markets. Before joining MEXC Learn, James worked as a senior analyst at a leading European investment firm, where he developed expertise in risk management and quantitative trading. His transition to cryptocurrency markets began in 2017, and he has since become recognized for his data-driven approach. He holds a Master's degree in Financial Economics from the London School of Economics. His analytical approach combines traditional technical analysis with on-chain metrics to provide readers with actionable insights.

 

Areas of Expertise:

 

 

Want the fastest access to MEXC's latest updates? Join our official Telegram group now!

MEXC Platform: Download App | Zero Fee Trading | Buy USDT Now

Join MEXC Community: X (Twitter) | Telegram | Discord

Get Insights: MEXC Learn | Terms of Service | MEXC Blog | MEXC News