1.0 The AI Harvest

As the global technology sector enters a new cycle defined by Artificial Intelligence, companies that have made long-term, strategic investments in the field are beginning to see their bets pay off. Baidu stands as a prime example of such a “long-term technology investor,” whose decade-long focus on AI is now yielding a significant harvest. The company’s H1 2025 financial report provides a clear headline metric: for the first time, revenue from its AI new business has surpassed 10 billion RMB, demonstrating that its foundational investments are translating into a powerful growth engine.

The core purpose of this document is to provide a comprehensive analysis of Baidu’s vertically integrated, “full-stack” AI strategy. We will argue that this vertical integration is not merely a technical choice but a strategic imperative designed to confer a decisive advantage in cost, efficiency, and speed of application deployment—the key battlegrounds in the next phase of AI competition. By controlling every layer from foundational silicon to user-facing applications, Baidu has constructed a formidable competitive advantage.

This analysis will examine the key business segments driving this transformation: Baidu AI Cloud, which has established a dominant position in the enterprise market; the Apollo Go autonomous driving service, which is scaling globally; and the suite of AI-native applications that are both revitalizing existing products and creating entirely new avenues for monetization. Ultimately, it is the foundational nature of Baidu’s comprehensive technology stack that serves as the core enabler of its success across these diverse verticals.

2.0 Baidu’s Full-Stack Architecture: The Four Layers of a Technology Moat

Baidu’s full-stack AI architecture is its central strategic pillar. By developing in-house capabilities across four critical layers—chips, framework, models, and applications—Baidu has established a level of vertical integration that serves as a key differentiator in a fiercely competitive market. This structure provides a powerful technological moat, enabling greater efficiency through deep optimization, enhanced control over the development roadmap, and the ability to innovate and deploy new features at a rapid pace.

2.1 The Foundation: KunlunXin Chips and PaddlePaddle Framework

At the base of Baidu’s architecture lie its proprietary hardware and deep learning framework, which provide the essential computational power and development ecosystem for its entire AI portfolio.

• KunlunXin (昆仑芯): Baidu’s self-developed AI chip is now in its third generation. The KunlunXin P800’s inference performance exceeds that of the NVIDIA A800, it boasts a formidable FP16 compute power of 345 TFLOPS, and Baidu has successfully deployed a massive 30,000-card cluster, providing a powerful hardware foundation for training and running large models.
• PaddlePaddle (飞桨): As China’s first self-developed, open-source deep learning platform, PaddlePaddle has cultivated a vast and mature ecosystem. As of November 2024, the platform has attracted an impressive 18.08 million developers and serves over 430,000 enterprises, making it the bedrock of AI development in the region.

2.2 The Brain: The Wenxin (ERNIE) Large Model Ecosystem

Building upon its foundational layers, Baidu has developed a sophisticated and rapidly evolving ecosystem of large models. The Wenxin (ERNIE) series has demonstrated impressive capabilities with the recent releases of Wenxin 4.5 (a native multi-modal model) and Wenxin X1 (a “deep thinking” model designed for complex reasoning). This rapid iteration has allowed Baidu to significantly close the performance gap with its global peers. According to the Stanford AI Index Report, the performance difference between top Chinese models, represented by Baidu, and top US models has narrowed dramatically from 17.5% in 2023 to just 0.3% in 2024.

2.3 Strategic Rationale: The Case for Closed-Source Models

In a market increasingly influenced by open-source alternatives, Baidu’s CEO, Robin Li, has taken a clear and pragmatic strategic stance, arguing that for most enterprises, relying on open-source models constitutes an “IQ tax.” He posits that closed-source models from leading technology companies offer a superior value proposition for three primary reasons:

1. Superior Performance: Closed-source models developed by top-tier firms consistently outperform their open-source counterparts in both capability and reliability.
2. Lower Total Cost of Ownership: Enterprises using closed-source models benefit from a shared, highly optimized infrastructure managed by the provider. This allows them to avoid the significant hidden costs associated with self-hosting, managing compute resources, and fine-tuning open-source models, resulting in a lower total cost.
3. Enhanced Efficiency and Specialization: Providers of closed-source models can offer a range of smaller, distilled models specifically tailored for different tasks. This allows businesses to choose the most efficient and cost-effective tool for a given job—whether it requires maximum power or maximum speed—rather than relying on a single, large, general-purpose open-source model.

This full-stack architecture, from custom silicon to a strategic preference for optimized, closed-source models, provides the technological and economic foundation for Baidu’s AI businesses. The next section will explore how this carefully constructed stack translates into tangible financial results.

3.0 Financial Performance: The Engine of AI Transformation

Baidu’s financial results for the first half of 2025 paint a clear picture of a company in the midst of a successful strategic transformation. The data reveals contrasting trends between its legacy business and its burgeoning AI ventures, highlighting a decisive shift in the company’s growth drivers. While the traditional online marketing segment faces headwinds, the AI-powered businesses are demonstrating robust, double-digit growth, underscoring the success of Baidu’s long-term pivot to AI.

Baidu H1 2025 Financial Highlights

A closer analysis of these figures confirms the strategic pivot. The AI New Business segment, which falls under the category of non-online marketing revenue, has become the primary growth engine for the company. Its 34% year-over-year expansion in the first half of the year stands in stark contrast to the pressures on the core advertising business.

This strong H1 performance was anchored by the impressive Q2 results of Baidu AI Cloud, which grew 27% YoY to 6.5 billion RMB. This single division’s quarterly result makes it the clear workhorse of Baidu’s entire AI pivot, constituting the majority of the “over 10 billion RMB” in AI New Business revenue for the first half of the year. The financial results are no longer just about search and ads; they are a clear indicator of a new Baidu powered by cloud, autonomous driving, and AI-native applications. This robust financial performance is not monolithic; it is driven by distinct, high-growth verticals, each leveraging the full-stack architecture in unique ways.

4.0 Analysis of Key AI Business Verticals

Baidu’s full-stack strategy materializes in three distinct and high-growth business verticals that are successfully converting technological prowess into market leadership and commercial success. This section will evaluate the market position, performance, and strategic initiatives of Baidu AI Cloud, the Apollo Go autonomous driving platform, and the company’s rapidly evolving consumer application ecosystem.

4.1 Baidu AI Cloud: Dominating the Enterprise Market

Baidu AI Cloud has firmly established itself as the leader in China’s enterprise AI sector. According to a recent report from the International Data Corporation (IDC), Baidu AI Cloud holds a 24.6% market share, ranking first in China’s AI public cloud market for six consecutive years. This dominance is particularly pronounced in key industries; for instance, it holds the #1 position in the financial industry for providing generative AI solutions. This market leadership is translating into significant commercial wins. In the first half of 2025 alone, Baidu Intelligent Cloud secured the most large model-related project bids in China, leading in both the number of projects (48) and their total value (510 million RMB).

4.2 Apollo Go: Accelerating Global Autonomous Mobility

The “萝卜快跑” (Apollo Go) Robotaxi service has scaled rapidly to become a global leader in autonomous mobility. Its operational scope and aggressive international expansion plans underscore Baidu’s ambition to define the future of transportation.

• Global Scale: Apollo Go is the world’s largest Robotaxi platform by operational scale, having completed over 14 million cumulative rides as of August 2025.
• Quarterly Performance (Q2 2025): The service delivered over 2.2 million rides in the second quarter, marking a remarkable 148% year-over-year increase.
• International Expansion: Baidu has forged pivotal strategic partnerships with global mobility giants. The collaboration with Uber will deploy vehicles in Asia and the Middle East, while the partnership with Lyft will focus on Europe, starting with Germany and the UK in 2026.

4.3 Consumer Ecosystem: AI-Native and Enhanced Applications

Baidu is pursuing a dual strategy for its consumer-facing products: creating entirely new AI-native applications while simultaneously retrofitting its existing products, which have a massive user base, with advanced AI capabilities.

• Baidu Search: The company’s flagship product recently underwent its “biggest redesign in a decade.” This AI-centric overhaul has had a dramatic impact on user experience. As of July 2025, 64% of mobile search results are now AI-generated, a significant jump from 35% in April of the same year.
• Baidu Wenku (文库): This document-sharing platform has become a prime example of successful AI monetization. By integrating AI-powered document creation, summarization, and generation features, Baidu Wenku has attracted approximately 26 million paying users, demonstrating strong consumer willingness to pay for value-added AI services.
• AI Agents & Digital Humans: Baidu is seeing significant commercial traction with its digital human technology. The “慧播星” (Huiboxing) platform powered a digital human livestream for influencer Luo Yonghao that generated 55 million GMV. This success is reflected in the segment’s financials, with digital human-generated revenue growing 55% sequentially in Q2 2025. Furthermore, Baidu is empowering a new generation of creators with its “秒哒” (Miaoda) no-code tool, which has already been used to create over 200,000 AI applications.
  The impressive performance across these three core verticals showcases how Baidu’s integrated technology stack is driving tangible business outcomes. This positions the company not just as a technology developer, but as a formidable competitor in the global race to commercialize AI.

5.0 Competitive Positioning and Strategic Outlook

Baidu’s strategy is a direct, calculated response to the dueling market dynamics of the global AI race, particularly the technological and economic interplay between the United States and China. The company’s approach reflects its deep technological assets, its understanding of the unique characteristics of the Chinese market, and a pragmatic vision for the future of AI commercialization.

5.1 Navigating the China-U.S. AI Dichotomy

The era of clear U.S. dominance in large model performance appears to be closing. According to the Stanford AI Index Report, the performance gap between the top models from the U.S. and China has become negligible, narrowing to just 0.3%. However, the two markets are evolving with distinctly different business models. The U.S. market is largely characterized by high-cost, closed-source models from a few key players. In contrast, the Chinese market is currently defined by aggressive price wars and the rapid proliferation of powerful open-source alternatives, creating a highly competitive and cost-sensitive environment.

5.2 The Patent Moat: A Quantifiable Strategic Asset

A key component of Baidu’s long-term competitive advantage lies in its extensive intellectual property portfolio. The 2025 Baidu AI patent white paper quantifies this strategic asset, revealing a commanding lead across multiple critical domains. This “patent moat” not only protects its innovations but also serves as a foundation for future commercialization and partnerships.

5.3 Outlook: Winning Through Application and Efficiency

Baidu’s CEO, Robin Li, views the current AI price war in China not as a threat, but as an inevitable and ultimately beneficial phase. He believes this period of intense competition will accelerate market adoption and quickly weed out inefficient players. His strategic outlook is built on two core beliefs:

1. The ultimate winner in this environment will not be the company with the marginally better model, but the one with the most efficient technology stack and the lowest operational costs, enabling it to outlast competitors while still delivering high-quality services.
2. The primary battleground for AI is rapidly shifting. With foundational model performance reaching a point of near-parity, the next wave of competition will be fought and won through the creation of a massive ecosystem of AI applications.

This perspective frames Baidu’s strategy as one focused on endurance and ecosystem development, positioning the company to thrive in a market where efficiency and practical application are paramount.

6.0 Conclusion: A Pragmatic Path in the AI Revolution

This analysis reveals a company successfully navigating a profound strategic pivot. Baidu’s decade-long investment in a vertically integrated, full-stack AI architecture is no longer a future bet but the primary engine of its present-day growth. The company has effectively shifted its business model, moving its center of gravity from the traditional advertising market to a diversified and robust portfolio of AI-driven services spanning cloud computing, autonomous mobility, and a rapidly expanding consumer ecosystem. Financial results and market share data confirm that this transformation is well underway, with AI-powered divisions delivering strong double-digit growth and establishing clear market leadership.

Looking ahead, Baidu’s trajectory is guided by a pragmatic focus on commercial application and operational efficiency. CEO Robin Li’s public statements indicate a strategic choice to prioritize the development of tangible AI applications and cost-effective model architectures over a pure-play pursuit of Artificial General Intelligence (AGI), which he believes remains a decade or more away. This approach is particularly well-suited to the current market dynamics in China, where intense price competition rewards efficiency and a vast user base creates immense opportunities for application-level innovation.

This application-centric strategy is a calculated response to market realities: in a world where foundational model performance is becoming commoditized, the winning strategy shifts to the application layer and operational efficiency. Baidu’s full-stack architecture is not just a technical asset, but the essential economic engine that allows it to win this new race—a race defined by cost-effective application deployment, not just raw model capability. Fortified by a dominant patent portfolio, this approach positions Baidu as a definitive leader in the current, practical, and commercially vital phase of the AI industrial revolution.