On the evening of August 18, Baidu (NASDAQ: BIDU) released its second-quarter earnings report, showing total quarterly revenue of 31.3 billion yuan, with AI business revenue reaching 12.5 billion yuan, accounting for 50% of Baidu's core business income. With AI revenue exceeding the halfway mark for two consecutive quarters, the market must now confront a notable reality: global top-tier institutions and asset managers are undergoing a symbolic collective shift in their stance toward Baidu.
According to 13F filings for the second quarter of 2026, legendary Wall Street investor Stanley Druckenmiller's Duquesne Family Office made its first entry into Chinese stocks in two and a half years after liquidating its Alibaba position in Q4 2023, with 88,000 Baidu ADRs now representing its sole Chinese holding. This move is not an isolated incident: Morgan Stanley simultaneously established a new position of approximately 6.5 million Baidu ADRs; UBS increased its stake by 917,700 ADRs, up 43% quarter-over-quarter; and Bank of America boosted its position by 1.087 million ADRs, a 305% quarter-over-quarter increase. Meanwhile, David Tepper, the most influential voice in the hedge fund space, has taken an even more aggressive bullish stance on Baidu. His Appaloosa Management, amid an overall contraction of Chinese internet assets in Q2, uniquely added over 600,000 Baidu ADRs, a sequential increase of 87%.
From Druckenmiller's return to Chinese stocks after two and a half years with Baidu as his pick, to the synchronized accumulation by three global banking leaders — Morgan Stanley, UBS, and Bank of America — and Tepper's extreme selectivity, this multi-dimensional, cross-strategy, cross-capital resonance among institutions clearly signals a key message: Baidu's AI investment value is receiving collective recognition from the world's top asset managers and leading banks. Behind this lies the fact that Baidu's AI commercialization has entered a phase of strong delivery with scaled replication. The company's AI business revenue has exceeded the 50% threshold for two straight quarters, meaning that for every two yuan of Baidu's core business income, one yuan now comes from AI. Clearly, AI has become Baidu's true revenue foundation.
This marks the accelerated commercial realization of Baidu's decade-long "chip, cloud, model, application" full-stack layout, now functioning as a reusable underlying capability. As one of the few global leaders with a complete full-stack presence in the AI industry, Baidu has become a barometer for observing China's AI commercialization progress. Its long-term intrinsic value is poised for systematic revaluation as AI revenue share continues to climb, which is precisely the underlying logic driving global institutions to vote for Baidu with real capital.
Leading the AI Office Sector, "Scaled Replication" Strengthens AI Application Growth Foundation
By 2026, the AI industry's narrative has progressively shifted — from "whose model is stronger" to "who can turn AI into real revenue." Even OpenAI, standing at the pinnacle of the global AI wave, is no longer leisurely enjoying technological dividends but is under immense profitability pressure to accelerate commercialization. As the industry enters the second half focused on monetization, whoever can first close the "technology-product-commercial" loop holds the ticket to the next round of competition. AI office applications, as the super-scenario with the highest certainty for large model commercialization, have become the main battleground for tech giants.
In 2026, Tencent consolidated resources like the QClaw team into WorkBuddy, targeting a C-end office entry point built on the WeChat/WeCom ecosystem. Alibaba integrated QoderWork, Wukong, and MuleRun into "Qianwen Office," leveraging DingTalk as an enterprise-grade AI productivity platform. ByteDance merged its Feishu team into Doubao and moved sales operations to Volcano Engine, forming a "Doubao + Feishu + Volcano Engine" To B office agent system. As giants simultaneously upgrade AI office from "internal horse racing" to "group-level campaigns," Baidu has taken the lead in seizing the super entry point for AI To B, with multiple AI application products — including Baidu Dazi, Kuku AI, and Miaoda — performing strongly in the market.
In the AI office agent ranking released on August 17, Baidu Dazi's monthly active users surged 1063.79% month-over-month, ranking first in growth rate among AI office agents. On August 10, Baidu Dazi launched design suites and financial suites, expanding AI office scenarios to accelerate user growth. Kuku AI, since officially launching version 1.0 in April 2025, has rapidly iterated to version 4.0 within a year, now available across all terminals of domestic and international products including Wenku Pan and Oreate AI. Currently, Kuku AI's office MAU has surpassed 25 million, ranking first in the AI office sector. In the "Office Agent Workflow Assessment Report" released by the National Industrial Information Security Development Research Center on August 5, Baidu Wenku scored first, independently occupying the top tier. Additionally, according to Frost & Sullivan's "China AI-Native No-Code Application Generation Platform Market Research, 2026 H1," Baidu Miaoda leads the industry with a 33.4% market share, ranking first in both OPC/commercial monetization users and effective application creation users in platform penetration.
The outstanding performance of multiple applications in the AI office arena highlights Baidu's leading edge in this track, with its AI applications also extending rapidly into B-end enterprise scenarios. Among these, the self-evolving agent Baidu Fumou has received trial applications from 3,000 enterprises since its launch, covering key industries such as ports, logistics, automotive, and finance. Furthermore, Baidu Yijing has served over 100,000 clients across more than 30 industries, with average live-stream conversion rates up 29%, delivery efficiency up 150%, and production costs down 63%. According to reports from Frost & Sullivan and IDC, Baidu Yijing has ranked first in China's AI digital human market share and overall product strength for two consecutive years.
Baidu's ability to produce hit products on both the C-end and B-end is no accident — it is the inevitable result of its "chip, cloud, model, application" full-stack capabilities being reused across customers, industries, and scenarios. From Kuku AI, Baidu Dazi, and Miaoda leading in office scenarios, to Fumou and Yijing accelerating B-end adoption, Baidu has clearly been the first to close the complete "technology-product-commercial" loop for AI applications. Investors should recognize that when the same full-stack architecture can be rapidly replicated from office scenarios to numerous other vertical domains, the "scaled replication" effect — where marginal costs decline while customer lifetime value increases — will continue to amplify, laying a solid foundation for future growth in Baidu's AI application business.
Full-Stack Capability Drives Multi-Layer Monetization, AI Commercialization Accelerates Toward Multi-Scenario Replication
Baidu's thoroughness in using full-stack AI capabilities to deeply reconstruct business lines across scenarios is rarely matched among domestic tech companies. The direct payoff of this strategic choice is a fundamental shift in its growth paradigm — the company is accelerating its transition from a traditional internet model to a high-value, sustainable growth model driven by an AI core. AI Cloud, Apollo Go, and Kunlun Chip — these multi-layer commercial realizations provide the most solid financial footnote to this transformation.
During the reporting period, Baidu's AI cloud infrastructure revenue reached 7.3 billion yuan, up 50% year-over-year, with GPU cloud revenue surging 283% year-over-year, accelerating from the previous quarter's 184% and marking four consecutive quarters of triple-digit growth. The sustained GPU cloud explosion fundamentally reflects expanding customer demand for AI training and inference computing power on public clouds. GPU public clouds, through unified resource pools, can serve multiple customer types simultaneously, with customer compute consumption continuously growing as model applications deepen. This indicates that Baidu's AI cloud revenue structure is shifting from one-time project delivery to standardized, sustainable service usage — upgrading the business model from "doing projects" to "providing scaled services."
This "scaled service" upgrade is directly validated in market share and customer structure. According to IDC data, as of mid-2025, Baidu Intelligent Cloud ranked first in the financial large model market, AI cloud gaming market, and embodied intelligence cloud market. In the first half of 2026, Baidu Intelligent Cloud led the industry with 1.385 billion yuan in winning bids, accounting for nearly 60% of total bids among five major cloud vendors, maintaining its leading position. Currently, Baidu Intelligent Cloud's new full-stack AI cloud has achieved scaled penetration in core industries. In finance, it covers 100% of systemically important banks and over 800 financial institutions. In automotive, it supports the delivery of over 20 million L2-level assisted driving vehicles. In AI hardware, it serves more than 1,000 clients. In embodied intelligence, it serves over 30 clients. Additionally, over 80% of central state-owned enterprises choose Baidu Intelligent Cloud, which has also helped State Grid implement intelligent agents at scale across more than 40 scenarios in the energy sector.
The dual leadership in market position and customer coverage breadth clearly demonstrates that Baidu's AI cloud services are expanding from internet customers to deep industrial scenarios, with the "scaled replication" effect rapidly amplifying — a core driver behind the explosive GPU cloud growth. Apollo Go, as Baidu's most technologically moated business segment, is accelerating cross-market scaled replication. Since Q2, Apollo Go's global operational footprint has visibly expanded. Apollo Go launched fully driverless commercial operations in Dubai, obtained Hong Kong's first fully driverless testing permits, and initiated in-vehicle safety officer-free testing on Airport Island, becoming the world's first platform to conduct fully driverless testing under a "right-hand drive, left-hand traffic" system. Simultaneously, Apollo Go started public road testing in London and Switzerland, and signed a strategic cooperation with Kazakhstan's TPH, marking the first entry of Chinese autonomous vehicles into the Central Asian market and filling a regional gap.
To date, Apollo Go has expanded to 28 cities globally, with cumulative autonomous driving mileage exceeding 350 million kilometers, including over 240 million kilometers of fully driverless operation. The ability to deploy the same autonomous driving technology in parallel across markets with vastly different regulations, road conditions, and driving habits shows that Baidu is replicating far more than a vehicle fleet — it is replicating a combination of technology, operations, and compliance capabilities. Each new scenario provides fresh training material for model iteration, and this data accumulation is an asset that latecomers will find difficult to replicate quickly, serving as a key pillar for Apollo Go to cement its market leadership. Additionally, Kunlun Chip is in a similarly rapid phase of scaled replication. On the hardware front, the Tianchi Supernode 1.0 has entered mass production, the 2.0's 256-card version is about to launch, and a next-generation 512/1024-card supernode is expected next year, with a single node capable of supporting trillion-parameter model training.
On the delivery front, the P800 30,000-card cluster has been lit up, and since 2025, multiple 10,000-card clusters have been delivered, covering internet, finance, energy, manufacturing, and other industries, serving over a hundred clients including Tencent, with delivery scales ranging from dozens to tens of thousands of cards. On the software front, Kunlun Chip has completed adaptation for mainstream models including ERNIE, DeepSeek, GLM, and Minimax, with inference efficiency improved by 50%. Scaled clusters, customer base, and software ecosystem together form a complete closed loop for the commercial validation of Kunlun Chip's self-developed silicon. From the AI cloud GPU explosion to Apollo Go's global operations and Kunlun Chip's scaled compute delivery, Baidu is converting its "chip, cloud, model, application" full-stack capabilities into commercial outcomes across cloud services, mobility, and computing power. As this foundation delivers simultaneously across multiple tracks, Baidu's AI commercialization has clearly moved from single-point validation to multi-scenario replication.
Valuation Framework Shift Becomes Market Consensus, Two Catalysts Accelerate Value Recovery
As Baidu's AI business revenue share consistently exceeds half, a practical question confronts the capital markets: how should this company — with scarce AI full-stack capabilities and AI commercialization leading across multiple tracks — be properly valued? In reality, Baidu's true value being undervalued is a "well-worn topic," with the core factor being that the market has historically applied traditional PE valuation methods to Baidu as a whole, causing the value of high-growth businesses like AI Cloud, Kunlun Chip, and Apollo Go to be heavily obscured by the search business. However, as AI commercialization continues to deliver and AI core business revenue share breaks through the 50% threshold, the consensus that Baidu's pricing should shift from traditional PE to a SOTP (Sum-of-the-Parts) framework is gradually forming — this is the core driver pushing Baidu's valuation center upward.
Nearly ten brokerages have recently explicitly adopted SOTP/segment valuation for Baidu, including top investment banks such as JPMorgan, UBS, Macquarie, CICC, and BOCOM International. Clearly, the need to switch Baidu's valuation framework has taken root in institutional circles. JPMorgan stated that under SOTP valuation, Baidu's AI business alone is worth $169 per ADS, implying a fair value of $230 per ADS for Baidu's US-listed shares — roughly double the current price. Building on this framework shift, two developments are poised to accelerate Baidu's value recovery. First, Baidu is advancing its dual primary listing conversion in Hong Kong. On August 19, Baidu Group CFO He Haijian stated that the board had previously approved converting its Hong Kong listing status from secondary to dual primary, with the conversion expected to take effect within this year, and the group is preparing for inclusion in the Southbound Stock Connect following the conversion. If it catches the September 2026 Stock Connect adjustment window, more mainland capital familiar with China's internet, domestic computing power, and AI industry will directly participate in pricing Baidu, potentially leading Southbound funds to rediscover the true value of its full-stack AI business.
Second, Kunlun Chip's Hong Kong IPO is actively progressing, and its listing would provide a direct catalyst for Baidu's revaluation. As the core asset of Baidu's full-stack AI infrastructure layer, Kunlun Chip's listing would offer a clear valuation anchor for its true worth. According to publicly available market information, global leading investment research firm Morningstar estimates Kunlun Chip's valuation in the range of $51.04 billion to $63.8 billion, while JPMorgan's March research report assigns Kunlun Chip an independent valuation of $40 billion to $49 billion. This implies substantial revaluation potential for Baidu over the medium to long term. At the current juncture, this mismatch between "undervaluation" and "AI high growth" presents an attractive entry window for long-term capital. With Q2 earnings now delivered, Baidu's current valuation largely reflects the market's interim assessment of its traditional business, while the long-term value of AI Cloud, Apollo Go, and Kunlun Chip remains to be fully released. As expectations for September Stock Connect inclusion heat up and the potential value of AI applications, Kunlun Chip, and Apollo Go undergoes market reassessment, the reshaping of Baidu's valuation anchor could begin at any time — and investors patiently awaiting catalyst delivery may reap substantial excess returns.
Comments