Baidu GPU cloud surges 283%, but investors punish unfinished AI transition

Baidu, Inc. (NASDAQ: BIDU; HKEX: 9888) reported second-quarter 2026 revenue of RMB31.33 billion, down 4% year over year, even as its artificial intelligence businesses continued to expand rapidly enough to represent about half of the company’s General Business revenue. Baidu Core AI-powered Business generated RMB12.5 billion, increasing 25% from a year earlier, while AI Cloud Infrastructure revenue climbed 50% to RMB7.3 billion and GPU Cloud revenue surged 283%. The counterweight was a 23% decline in Legacy Business revenue to RMB10.4 billion and a 19% contraction in online marketing services to RMB13.1 billion, leaving the company with a genuine AI transition but not yet a complete economic replacement for the businesses being disrupted. Net income attributable to Baidu fell to RMB2.32 billion from RMB7.32 billion a year earlier, adding another reason investors reacted negatively despite the headline AI growth.

The quarter creates a more complicated investment case than simply describing Baidu as a beneficiary of China’s AI expansion. Artificial intelligence has already moved from an experimental adjacency to one of the largest economic engines inside the company, yet AI-powered revenue declined 8% sequentially from the first quarter and AI Cloud Infrastructure revenue fell 17% quarter over quarter. Baidu therefore has to prove not only that its AI operations can grow rapidly from year-ago levels, but that they can expand consistently enough to offset continuing structural erosion in traditional advertising and internet services.

How much of Baidu’s business has actually shifted from internet advertising toward artificial intelligence?

Baidu’s internal revenue breakdown shows just how far the transformation has progressed. Core AI-powered Business generated RMB12.5 billion in the second quarter, compared with RMB10.0 billion a year earlier, increasing its share of Baidu General Business revenue from 38% to approximately 50%. Legacy Business moved in the opposite direction, falling from RMB13.6 billion to RMB10.4 billion, while other General Business revenue declined from RMB2.7 billion to RMB2.3 billion. Baidu is consequently reaching the point where artificial intelligence is no longer being financed primarily by a dominant legacy operation, because the AI business itself has become comparable in size to the traditional business it is intended to replace.

The transition is especially visible inside the company’s conventional revenue classification. Online Marketing Services generated RMB13.1 billion, down 19%, while the broader “Others” category produced RMB12.1 billion, up 21%, with AI Cloud among the principal growth drivers. One year earlier, online advertising represented 62% of Baidu General Business revenue and other activities represented 38%; by the second quarter of 2026, that mix had narrowed to 52% and 48%, respectively. This is a substantial shift for a company historically dependent on search advertising, although it also demonstrates why weakness in marketing remains large enough to suppress consolidated growth.

The difficult part is that the transition is occurring while total General Business revenue is shrinking. Baidu General Business revenue fell 4% year over year to RMB25.2 billion, which means the expanding AI businesses are still replacing lost revenue rather than adding enough incremental scale to produce overall growth. A more decisive inflection would require AI Cloud, applications and AI-native marketing to expand faster than the decline in conventional advertising over several consecutive quarters.

Why does Baidu’s 283% GPU Cloud growth matter more than the broader 50% AI Cloud increase?

AI Cloud Infrastructure remained the largest individual component of Baidu’s AI-powered business, generating RMB7.3 billion and increasing 50% year over year. Within that operation, GPU Cloud revenue rose 283%, accelerating from 184% growth in the first quarter, according to Baidu. The distinction is important because GPU Cloud represents public cloud-based artificial intelligence computing demand, providing a more direct measure of whether enterprises are actually paying Baidu for scarce compute infrastructure rather than merely experimenting with AI applications.

The acceleration also supports Baidu’s attempt to build a vertically integrated artificial intelligence business spanning models, cloud infrastructure, applications and its internally developed Kunlunxin accelerator technology. That architecture potentially lets the company participate in several layers of enterprise AI spending, from compute infrastructure through software deployment, instead of relying only on model subscriptions or consumer-facing products. Baidu has separately been pursuing a proposed spin-off and listing of Kunlunxin, which could make the economics of its semiconductor exposure more visible if that process advances.

However, sequential performance tempers the enthusiasm. AI Cloud Infrastructure revenue declined from RMB8.8 billion in the first quarter to RMB7.3 billion in the second quarter, a 17% quarterly reduction, even though the year-over-year comparison remained strong. Investors therefore have two valid signals to monitor at the same time: extraordinary growth against the previous year’s smaller base, and evidence that quarterly demand can still be uneven as large enterprise infrastructure projects move through procurement cycles.

Are Baidu’s enterprise AI applications becoming a meaningful second revenue engine?

AI Applications revenue reached RMB2.5 billion, increasing only 3% year over year, making this operation far smaller and slower growing than AI Cloud Infrastructure. Baidu nevertheless continued expanding the product portfolio, including the June launch of an enterprise edition of DuMate, its general-purpose productivity agent, while AI daily-active-user penetration across Baidu Wenku and Baidu Drive increased 27.4% year over year. The usage numbers indicate broader adoption of AI functionality, but revenue growth suggests that user engagement has not yet translated into the same commercial acceleration visible in cloud infrastructure.

This difference matters because infrastructure businesses and applications can have very different long-term economics. GPU cloud demand can produce large revenue quickly but requires substantial expenditure on servers, accelerators, data centres and networking equipment. Software agents and productivity applications may eventually carry higher incremental margins if they achieve scale, yet Baidu still needs to establish durable willingness to pay for these tools in a competitive Chinese AI application market.

AI-native Marketing Services generated another RMB2.6 billion and was approximately flat year over year. Together, AI Applications and AI-native marketing produced about RMB5.1 billion of quarterly revenue, giving Baidu more than one route to monetize artificial intelligence beyond raw computing capacity. Their combined growth trajectory remains less impressive than GPU Cloud, which increases the importance of proving that Baidu can build software revenue on top of the infrastructure investment rather than becoming disproportionately dependent on capital-intensive compute.

What do Baidu’s falling profits reveal about the cost of rebuilding the company around AI?

Baidu’s consolidated revenue fell only 4%, but profitability weakened much more sharply. Net income attributable to Baidu declined from RMB7.32 billion in the second quarter of 2025 to RMB2.32 billion, while diluted earnings per American depositary share fell from RMB20.35 to RMB5.74. Non-GAAP net income was RMB2.57 billion and non-GAAP diluted earnings per ADS was RMB7.22. Operating income was RMB3.02 billion with a 10% margin, while adjusted EBITDA reached RMB6.15 billion and carried a 20% margin.

Part of the year-over-year earnings decline came from below-the-line movements rather than operating deterioration alone. Total other income dropped to RMB184 million from RMB4.86 billion a year earlier, primarily because of lower fair-value gains on long-term investments and greater foreign-exchange losses. Operating income itself fell from RMB3.28 billion to RMB3.02 billion, a much smaller reduction than the headline net-income decline, which is important when separating underlying business performance from investment and currency effects.

The cost structure nevertheless shows the capital intensity of the AI transition. Cost of revenue increased 4% year over year to RMB19.1 billion despite lower consolidated revenue, with Baidu identifying increased AI Cloud costs as a principal driver. Research and development expenditure remained substantial at RMB4.6 billion, while management has indicated that continued investment in talent and infrastructure will be necessary if Baidu intends to regain stronger positioning around its Ernie models and compete across China’s AI ecosystem.

Baidu still has considerable financial capacity to fund that strategy. Total cash and investments stood at RMB283.1 billion, equivalent to approximately $41.7 billion, at June 30, and operating cash flow remained positive for a fourth consecutive quarter at RMB3.4 billion. The balance sheet therefore gives Baidu room to tolerate weaker near-term margins, but investors will eventually expect those investments to produce consolidated growth rather than merely offsetting disappearing legacy revenue.

Why did BIDU shares remain under pressure despite the artificial intelligence growth?

Baidu’s Nasdaq-listed American depositary shares closed at approximately $94.08 on August 21 after falling sharply following the earnings release. That left the stock about 9.3% below its August 14 close and approximately 13.1% below its July 21 close of $108.22. The shares remained substantially below their 52-week high of $165.30 and much closer to the lower end of the roughly $88 to $165 trading range, indicating that investors continue to place a heavy discount on the transition despite the growth of AI Cloud.

The negative reaction appears understandable when the earnings are viewed as a transition rather than an AI growth headline. Artificial intelligence already accounts for half of General Business revenue, but General Business itself contracted, profits declined and the largest AI infrastructure business fell sequentially from an exceptionally strong first quarter. The market is effectively demanding evidence that Baidu can move from substitution, where AI replaces declining advertising revenue, to expansion, where artificial intelligence generates net company-wide growth.

There are nevertheless several potential catalysts. GPU Cloud growth remains exceptionally high, Baidu has maintained substantial liquidity, and Apollo Go continues expanding internationally, with operations or testing across markets including Dubai, London, Hong Kong and Switzerland. The company is also seeking shareholder approval on August 26 for steps associated with converting its Hong Kong presence to a dual-primary listing, which management expects to complete during 2026 subject to the required approvals.

Baidu’s second quarter therefore shows a transformation that is already financially real but still economically unfinished. Artificial intelligence has reached the scale necessary to become the company’s core engine, yet the next phase is more demanding: AI revenue needs to grow fast enough, consistently enough and profitably enough to overcome legacy declines rather than simply running alongside them.