On July 17, 2026, the 10th World Artificial Intelligence Conference (WAIC) officially opened, marking a significant milestone in China’s ongoing AI development over the past decade. The event showcased a high-caliber lineup, jointly organized by ten government ministries and the Shanghai Municipal Government, with renowned Turing Award laureate Yao Qizhi serving as the academic chair. For the first time, a dedicated OPC (one-person company) display zone was featured, highlighting a full industry chain—from chips to applications—demonstrating China’s fast-evolving AI ecosystem.
This wave of large-language model innovation has brought transformative changes across China’s AI industry. Reflecting on the past few years, when ChatGPT’s popularity surged, China was still catching up, with only Baidu’s Wenxin model making a notable appearance. By 2024 and 2025, domestic models like DeepSeek, Zhitu AI, and Minimax began gaining prominence in industry rankings, signaling rapid progress on the application front.
Since late 2025, the global AI landscape has been dominated by the rise of intelligent agents, exemplified by OpenClaw. This shift signaled a new era in AI competition, transforming China’s industry from chasing parameter counts in massive models to breakthroughs in self-developed computing chips, and now into a period where intelligent agents are actively deployed across diverse sectors.
China’s AI revolution over the past three years has not hinged on a single technological breakthrough but on the deep coupling of models, computing power, engineering, and real-world scenarios. This comprehensive integration has fostered an extensive and robust AI industry chain that reached a dramatic bloom along the Huangpu River in the summer of 2026.
Amidst this landscape, Baidu has exhibited a distinctive transformation. Its exhibition featured an entire ecosystem stack for intelligent agents, encompassing foundational components like Kunlun chips and Tianci supernodes, Baidu Cloud’s platform, large-model capabilities, and practical applications. Such full-stack in-house development is rare at WAIC—most companies tend to showcase isolated technological advances or futuristic features.
In recent years, Baidu’s AI strategy accelerated, with notable financial results. In the first quarter of 2026, Baidu’s total revenue hit 32.1 billion yuan, with AI segments contributing 13.6 billion yuan—nearly half of the company’s revenue. AI cloud income surged 79% to reach 8.8 billion yuan, with GPU cloud revenue soaring by 184%. This rapid growth underscores Baidu’s shift from experimentation to real profitability, diverging from industry trends that often prioritize showcasing potential over actual financial returns.
The transition from traditional large models to intelligent agents signifies a fundamental change in AI competitiveness. While large models remain valuable, the era now emphasizes AI’s utility as a productivity tool—agents that, given a clear objective, can break down tasks, execute, debug, and deliver solutions independently. Unlike chatbots that are essentially advanced search tools, intelligent agents are envisioned as digital labor forces capable of autonomous operation.
This shift is reflected in the operations of AI giants. For instance, OpenAI recently integrated coding agents (Codex) into its ChatGPT platform, moving beyond simple chatbots towards more autonomous “agents” capable of executing complex tasks. Meanwhile, domestic models are following suit, prioritizing agent deployment over mere model capabilities, emphasizing practical implementation.
This evolution fundamentally alters AI’s measurement metrics. The focus transitions from token consumption—traditionally used to gauge model size and cost—to token output, emphasizing efficiency and real-world results. Baidu’s CEO Robin Li proposed the concept of Daily Active Agents (DAA) at the 2026 Create AI Developers Conference—an indicator that measures active intelligent agents engaging with business scenarios daily. Li predicts that the global DAA could surpass 10 billion, signifying a shift toward efficiency and usability rather than just consumption.
Baidu’s strategic focus on DAA aligns with its long-standing belief that intelligent agents are the core medium for deploying large models in real-world settings. Instead of measuring by token usage, Baidu emphasizes active engagement frequency and the active scale of intelligent agents—a move from processing power to tangible effect.
To support this, Baidu has invested heavily in full-stack infrastructure. Upgrades to its Agent Infrastructure include the Token Factory for optimizing token costs and Harness Engineering, facilitating enterprise and individual developer deployment of smart agents. Their cloud platform demonstrates impressive efficiency—reducing token consumption by about 23% while maintaining a 95% success rate across enterprise use cases.
This comprehensive, integrated approach is Baidu’s unique advantage. Building from chips to applications, Baidu’s architecture emphasizes the entire stack—from chip design (Kunlun chips) to software frameworks and industry-specific deployment—reducing technical disjointedness, lowering costs, and enhancing capabilities. Notably, the company’s self-developed Kunlun chips have been scaled to hundreds of thousands of units, and their cloud infrastructure has achieved significant optimizations in training costs and inference efficiency.
In the AI hardware sector, Baidu’s complete internally developed ecosystem provides a robust competitive edge. From initial chip design through to full-stack deployment, the company’s system-wide strategy aims for cost-effective, high-performance AI solutions that are difficult for competitors to replicate quickly. Industry insiders see a three-year window—equal to typical hardware development cycles—to establish a dominant, fully integrated AI stack.
Market-wise, Baidu’s efforts are already paying off. In the first half of 2026, Baidu’s AI-related project wins exceeded 1.386 billion yuan, surpassing other cloud providers by multiples. More impressively, AI now accounts for over half of Baidu’s total revenue, signaling its transition from an emerging technology to a core business driver—what industry analysts call the “first growth curve” of the era of intelligent agents.
Behind these figures lies a deep strategic shift—one that prioritizes efficiency and deployment success over mere token consumption. Baidu’s DAA metric embodies a fundamental industry evolution: from measuring AI by input costs to valuing tangible, revenue-generating outputs.
Li Jing, Baidu’s SVP and head of Baidu Cloud, emphasizes that intelligent agents are the main vehicle for operationalizing large models. The key is not just building powerful chips or models but ensuring they work seamlessly across the entire infrastructure—hardware, algorithms, cloud, and enterprise applications. Baidu’s full-stack architecture, with standardized optimized interfaces at every level, aims to reduce deployment costs and improve productivity.
This holistic approach extends to the supply chain. Baidu has built a comprehensive hardware and software ecosystem, which includes its own AI chips, cooling solutions, and specialized infrastructure—all aligned at the instruction set level—leading to smoother integration and higher efficiency.
Looking ahead, Baidu foresees the adoption of this full-stack model as critical. Industry timelines estimate that developing a fully in-house AI chip—from design to production—takes about two to three years. For companies building this from scratch, that timeframe represents a strategic window to gain a competitive advantage. Baidu’s early investments and continuous improvement position it favorably to establish an unassailable leadership position once the market fully matures.
Ultimately, the biggest story from WAIC 2026 may be the industry’s shift in measurement and strategy—adopting a new “ruler” calibrated for results rather than inputs. Baidu’s rapid ascent, evidenced by revenue and deployment metrics, demonstrates how changing the metrics can effectively reposition a company at the forefront of the AI revolution.
While it remains to be seen whether Baidu’s approach will become the industry standard, their early exploration and validation of this new valuation framework—centered on active, results-oriented AI—set a compelling precedent for others to follow. The future of AI, after all, hinges not just on technological prowess but on tangible, market-driven outcomes, and Baidu appears well on its way to leading the charge.
