When a company operates several seemingly disparate businesses that are “folded” into the same financial report, there is often a “time lag” in how capital perceives these ventures.

Amazon was long viewed as a low-margin retail company until AWS was recognized as a standalone entity. NVIDIA was once confined to the graphics chip cycle until the explosion in AI computing demand revealed that it held not just graphics cards, but the “shovels” of a new era.

Capital assesses a company primarily through core metrics and growth expectations, without delving deeply into each individual business. This leads to a common phenomenon where “business development outpaces market understanding.”

Baidu finds itself at such a pivotal juncture.

During the Q1 2026 earnings release, Robin Li stated that AI had become Baidu’s core growth driver.

On August 18, Baidu released its Q2 2026 financial report, revealing total revenue of RMB 31.3 billion, with general business revenue at RMB 25.2 billion. AI business revenue accounted for over 50%, surpassing the halfway mark for the second consecutive quarter.

From a fundamental perspective, Baidu’s AI cloud infrastructure, AI applications, and AI-native marketing services have formed three distinct revenue streams. Meanwhile, Kunlunxin and Apollo Go have established significant technical and operational assets in chips and autonomous driving, respectively.

Notably, the traditional valuation framework based solely on search advertising can no longer encapsulate Baidu’s fully formed AI ecosystem and the ongoing transformations.

On July 28, Baidu announced its intention to switch its Hong Kong listing status from secondary to primary and called for a special shareholder meeting on August 26 to seek approval.

On August 19, Baidu stated that its board had approved the conversion of its Hong Kong listing from secondary to dual-primary status, with the process expected to take effect within the year. The company is preparing for inclusion in the Southbound Trading link under the Stock Connect program.

In light of the latest financial results, it is clear that Baidu’s push for a dual-primary listing in Hong Kong is not merely about adding another listing status—it is about finding new pricing benchmarks for its mature AI assets.

Currently, AI business revenue accounts for over half of Baidu’s total, and its full-stack capabilities have transformed into multiple commercial revenue streams, bolstering the company’s business confidence. Additionally, its organizational agility and delivery capabilities have enabled synergies across chips, cloud, models, and platforms, achieving technological confidence through customer procurement, large-scale delivery, and real-world operations.

Baidu deserves—and must undergo—a re-evaluation.

As AI-related revenue, growth, and assets are progressively integrated into financial reports, Baidu has both the incentive and the confidence to enter more open capital markets, where it can be reassessed by diverse investors using new benchmarks.

Baidu’s AI assets have reached a level of clarity that warrants renewed research and pricing.

I. AI Assets, Unfolding from Obscurity

For a long time, the capital markets have been most familiar with Baidu’s search business.

Search advertising provides stable cash flow and has established a relatively fixed valuation framework. Meanwhile, autonomous driving, AI chips, deep learning frameworks, and large models require continuous investment, with their values difficult to reflect directly in early-stage revenue and profits.

When these businesses are “folded” into financial reports, market interpretations often lag. After all, financial statements record current results, while technological investments bet on future possibilities.

In hindsight, search served as the gateway, and advertising as the monetization method. During the golden age of PC internet, this framework was sufficiently precise. However, since then, Baidu has channeled substantial profits into two “invisible” ventures with no immediate returns: autonomous driving (Apollo) and AI infrastructure chips (later known as Kunlunxin).

Robin Li once succinctly explained Baidu’s approach to technological investment: “When we have RMB 1, we invest it in technology; when we have RMB 100 million, we still invest it in technology; when we have RMB 10 billion, we continue to invest it in technology.”

Before the AI industry exploded, this choice primarily manifested as investment. Search advertising generated continuous revenue, while AI existed mostly in the form of R&D expenses, talent acquisition, and long-term projects.

Take Kunlunxin as an example. In 2011, to improve computing efficiency for its search business, Baidu began developing FPGA accelerators internally. Years of large-scale FPGA deployment and application equipped Baidu’s team with extensive experience in AI computing hardware, laying the foundation for transitioning from programmable FPGAs to dedicated AI chips. In 2018, the first-generation Kunlun chip was released, followed by independent financing in 2021.

This business line grew organically from internal demand. Initially, it served to optimize costs and efficiency for Baidu’s search and AI computing. As large models drove explosive demand for computing power, chip and cluster capabilities began to find a broader external market.

The Wenxin large model provided a unified outlet for Baidu’s accumulated expertise in models, frameworks, computing power, and data, while Baidu Intelligent Cloud delivered these capabilities to enterprise clients. What was once recorded as R&D expenditure began transforming into revenue from model usage, enterprise procurement, and cloud services.

In Q2 this year, Baidu’s AI cloud infrastructure revenue reached RMB 7.3 billion, up 50% year-on-year. GPU cloud revenue surged by 283%, accelerating from the previous quarter’s 184% growth and maintaining triple-digit expansion for four consecutive quarters.

Enterprise procurement of GPU cloud represents the most direct form of monetary endorsement. It indicates that computing demand driven by large models is moving beyond technical demonstrations into enterprise budgets and production systems.

Meanwhile, Baidu’s AI application revenue reached RMB 2.5 billion in Q2, while AI-native marketing services generated RMB 2.6 billion. AI revenue is no longer concentrated in a single model or product but is diversifying across infrastructure, applications, and commercial services.

Andy Grove once observed that true strategic inflection points often begin before noticeable changes in financial results. Baidu’s long-term technological investments and their current value realization serve as a fresh illustration of this principle.

Beyond the financial reports, many of Baidu’s rapidly accelerating businesses are entering their harvest phase.

According to Baidu, since 2025, Kunlunxin has deployed multiple 10,000-card P800 clusters across industries such as internet, finance, energy, and manufacturing, achieving compatibility with mainstream models like Wenxin, DeepSeek, GLM, and MiniMax. The Information, citing insiders, reported that Tencent has also become a Kunlunxin customer.

When a company’s technological capabilities serve not only internal operations but also the procurement systems of other leading tech firms, it transitions from an internal cost center to an independent market asset.

Recently, this asset received standalone valuation: Morningstar estimates Kunlunxin’s value at HKD 400–500 billion, while JPMorgan projects an independent valuation of USD 40–49 billion, with Baidu’s attributable share at USD 27–34 billion.

Another illustrative business is autonomous driving. In Q1, Apollo Go provided 3.2 million fully driverless rides, up over 120% year-on-year, with a weekly peak of over 350,000 rides in March and a cumulative total exceeding 22 million. By Q2 2026, Apollo Go covered 28 cities globally, accumulating over 350 million autonomous kilometers, including over 240 million kilometers of fully driverless operation.

In July, Apollo Go secured its first fully driverless testing license in Hong Kong and launched public road tests in London in partnership with Uber and Lyft’s Freenow. It subsequently entered Kazakhstan and became the first and only provider of fully driverless services in Dubai through a self-operated platform.

These developments suggest that Apollo Go is moving beyond “pilot” status toward large-scale commercialization.

“Smart money” movements provide corroborating evidence. According to foreign media, Stanley Druckenmiller’s family office established a position of 88,000 Baidu ADRs in Q2—its first return to U.S.-traded Chinese stocks in years, having publicly stated in 2024 no intention to invest in Chinese equities. Meanwhile, David Tepper’s Appaloosa added 602,900 Baidu shares while divesting from JD.com, Pinduoduo, and reducing its Alibaba stake.

With AI-related revenue, growth, and assets now reflected in financial reports, and Baidu’s AI business supported not by a single product or client but by a diversified portfolio, the company has gained the confidence to engage with more open capital markets.

II. Inherent Execution DNA

Chips and autonomous driving represent just two facets of Baidu’s full-stack AI strategy. The significance of a full-stack approach lies not in the quantity of businesses but in the reusability of a unified technological capability across different layers, enabling mutual commercial validation.

Baidu refers to its full-stack AI layout (strategic layout) as “chip-cloud-model-platform,” leveraging shared capabilities from Kunlunxin, Baidu Intelligent Cloud, and the Wenxin large model to create diverse commercial touchpoints in both digital and physical realms.

Beyond business confidence, Baidu possesses substantial technological confidence as it moves toward more open capital markets. This confidence stems from its strong full-stack execution capabilities, evident in three dimensions: customer procurement, large-scale delivery, and real-world operations.

In customer procurement, the market has voted with its wallet. According to Smart Hyperparameter statistics, in the first half of 2026, five major domestic AI cloud providers secured approximately RMB 2.044 billion in large model-related bids, with Baidu Intelligent Cloud leading at RMB 1.385 billion (nearly 60%), maintaining its top position from 2025.

IDC reports show that Baidu Intelligent Cloud ranked first in China’s financial industry generative AI market with a 16.6% share, held a 51% share in AI gaming cloud (exceeding the combined total of second to fifth place), and led the embodied AI cloud market with a 29.55% share.

With the advent of the Agent era, model usage has shifted from simple Q&A to complex tasks, driving increased demand for inference capabilities. Enterprises now require not just individual chips but stable clusters, cloud services, and engineering systems capable of running large models.

Kunlunxin has thus become a critical pillar of Baidu’s full-stack delivery capabilities.

The Tianchi 256-card supercomputing node, built on Kunlunxin, delivers 25% higher throughput and 50% greater inference efficiency than its predecessor. A single Tianchi 512 supernode can support training for trillion-parameter models.

With Kunlunxin’s proven delivery capabilities for 10,000-card clusters, its business naturally extends into external production environments. To date, its customers span internet, finance, energy, and manufacturing sectors. Over 800 AI applications at China Merchants Bank now run stably on Kunlunxin P800.

Beyond computing power, Baidu Intelligent Cloud has achieved large-scale delivery in core industries. According to Baidu, over 1,000 AI hardware companies have integrated with Baidu Intelligent Cloud, which serves systemically important banks and more than 800 financial institutions. In state-owned enterprises, automotive, and embodied AI sectors, cloud, models, and computing power are delivered as a comprehensive package to clients.

AI must exit the lab; its technological value ultimately manifests in real-world scenarios, workflows, and road conditions.

In this regard, Baidu’s increasingly robust AI “portfolio” provides ample real-world service cases to strongly support its technological confidence.

For example, Baidu Dazi achieved 8.97 million monthly visits in July, up 845% month-on-month, ranking among the fastest-growing AI Agent products globally during the period. Miaoda secured a 33.4% share in China’s AI-native no-code application generation platform market.

Additionally, Baidu Yijing serves over 100,000 clients across 30+ industries, Famu has been trialed by over 3,000 enterprises, and Apollo Go operates at scale in 28 cities worldwide.

This diverse product matrix underscores Baidu’s strong technological and execution DNA, enabled by its full-stack organizational capabilities. Technological reserves do not automatically translate into business success; full-stack execution requires aligning dispersed technologies under a unified goal. GenFlow—the cross-platform AI agent jointly launched by Baidu Wenku and Baidu Netdisk in 2025—exemplifies this.

On August 14, during AI Day, GenFlow unveiled its Chinese name, “Kuku AI.”

Since its launch, Kuku AI has rapidly integrated technologies, products, traffic, and commercialization resources previously scattered across different businesses, evolving from GenFlow 1.0 to Kuku AI 4.0 within a year. It ranked first in the National Industrial Information Security Development Research Center’s Office Agent workflow evaluation, securing its position in the top tier.

To date, Kuku AI’s office MAU exceeds 25 million.

This organizational agility in rapidly mobilizing resources toward new objectives has significantly accelerated Baidu AI’s execution speed, forming another cornerstone of its technological confidence and underpinning its decision to pursue dual-primary listing.

III. China’s AI Needs New Pricing Benchmarks

When a company’s business structure evolves, its capital market valuation framework must adapt accordingly.

Take Baidu as an example: Baidu Intelligent Cloud should be evaluated based on enterprise clients, computing revenue, and service scale; Kunlunxin by orders, cluster delivery, and model compatibility; AI applications by user base, retention, and monetization; and Apollo Go by city coverage, operational mileage, and commercialization progress.

When these assets are “folded” into a single financial report under the label of a “search company,” the differences between businesses can lead to valuation discounts. As Aswath Damodaran notes in *The Dark Side of Valuation*, this phenomenon is known as a “conglomerate discount”—where the combined valuation of a diversified company is often lower than the sum of its parts.

This is not unique to Baidu but a shared challenge for China’s AI assets.

Since the rise of large models, China’s AI industry has developed rich business forms across large models, AI clouds, chips, intelligent vehicles, and various applications. However, from a capital market perspective, some startups can quickly establish valuations based on a clear AI narrative, while comprehensive AI companies like Baidu—with cloud, chips, models, applications, and autonomous driving—face greater pricing complexity.

The reason is straightforward. Startups can be priced based on model capabilities, user growth, or future market potential, offering investors a relatively pure growth trajectory. In contrast, diversified tech giants are still primarily valued using the current revenue and profits of their mature businesses as anchors.

The paradox for China’s AI industry is that the more complete a company’s AI layout (strategic layout), the more business layers require identification, making it more susceptible to being “folded” under traditional labels.

Against this backdrop, Hong Kong stocks have emerged as a crucial window for recalibrating China’s AI valuation framework.

The Hong Kong market connects global capital with Chinese investors and hosts a growing number of Chinese internet, semiconductor, new energy vehicle, and AI application companies. Compared to a singular overseas tech stock comparison system, Hong Kong offers a more diversified evaluation framework for China’s AI assets, allowing investors more familiar with China’s tech industry, enterprise clients, and application scenarios to participate in pricing.

Of course, merely changing a business’s listing status without altering its fundamental operations offers only a partial solution. Investor structure determines which valuation framework a market uses, but a company’s business and organizational capabilities ultimately determine whether capital accepts that framework.

Currently, Baidu has been supported by two fundamental aspects: First, the business confidence brought by over half of its revenue coming from AI and the formation of multiple commercial revenue streams. Second, the technological confidence derived from the interconnectedness of 'chip, cloud, model, and agent' technologies, which has been validated through procurement, delivery, and operational inspections.

As for Baidu's dual-primary listing, which proudly steps into a more open capital market, its significance lies not only in enabling more investors to witness the company's qualitative transformation but also in demonstrating a value confidence that Baidu can undergo reevaluation and repricing. Baidu is preparing for its inclusion in the Southbound Trading link, aiming to open a channel for mainland capital to directly invest in Baidu and pave the way for reshaping its valuation.

In the past, Baidu's AI investments were long recorded as costs. Today, they are beginning to appear in the company's financial statements in the form of revenue, orders, clients, and assets.

This indicates that AI has become the core driver of Baidu's growth and the enhancement of its corporate value. Taking the initiative to join the Southbound Trading link reflects Baidu's assessment of the maturity of its AI business and its choice to embrace the long-term value of AI in China, signaling the arrival of a new AI cycle.

More importantly, Baidu's transformation is a microcosm of the evolution of China's AI industry.

Over the past few years, the global market has observed China's AI with a focus on model capabilities, technological catch-up, and investment scale. As AI cloud services, domestic chips, intelligent agents, autonomous driving, and other businesses enter a phase of large-scale implementation, it signifies that China's AI is transitioning from a phase of technological breakthroughs to one of asset formation and value discovery.

Looking around the capital market, the business development often outpaces market understanding, a time lag that every company traversing different cycles must experience.

Amazon once waited for the market to recognize AWS, and NVIDIA also waited for AI computing power to redefine itself. Today, Baidu's past AI investments are beginning to manifest as revenue and assets, necessitating an update in market understanding.

The market will eventually catch up with the pace of business development. As a long-term investor and trailblazer in China's AI wave, Baidu, along with China's AI industry, which is undergoing revaluation, will move from the technological stage to the value center of global capital, embracing a new round of challenges.