At Create 2026, Baidu’s co-founder and CEO Robin Li proposed replacing token consumption with “Daily Active Agents” (DAA) as the key metric for gauging the value of AI platforms. He declared, “For the first time, the stars of AI aren’t models — they’re applications.” Immediately afterward, Shen Dou, President of Baidu Intelligent Cloud, announced that Baidu Intelligent Cloud would evolve into a “new full-stack AI cloud tailored for large-scale agent applications.” Centered around two core infrastructures — “AI Infra” and “Agent Infra” — this initiative aims to build a comprehensive product ecosystem dubbed “chip-cloud-model-agent.” Specifically, Kunlun chips provide computing power; Baidu Cloud serves as the underlying infrastructure; ERNIE 5.1, the upgraded version of Baidu’s large language model, acts as the foundational capability layer; while Baidu’s super intelligent system suite — including DuMate, Miaoda 3.0, and Farmou — constitutes the upper-layer applications. Additionally, Miaoda 3.0’s enterprise edition was launched simultaneously; official figures reveal that since its debut, applications generated via this platform have served over 10 million users, generating a total economic value exceeding 5 billion RMB. Meanwhile, Kunlun P800 chips have seen widespread adoption, while ERNIE 5.1 achieves a training efficiency rate surpassing 97% across clusters containing 30,000 GPUs.
This strategic shift stems from clear financial considerations: In Q1 2026, Baidu’s overall revenue stood at roughly 32.075 billion RMB, marking a slight 1.2% year-on-year decline. Conversely, revenue from core AI-related services surpassed 13.6 billion RMB — a robust 49% increase — accounting for over half of Baidu’s total non-generic business earnings. Notably, revenue derived from intelligent cloud infrastructure alone hit approximately 8.8 billion RMB, reflecting a staggering 79% surge. According to analysis by Huxiu, Baidu’s current AI growth primarily relies on infrastructure development; the monetization potential of its application ecosystem remains somewhat uncertain. This reality underpins Baidu’s decision to prioritize DAA metrics: amid intensifying competition among similar large language models, framing success through “ecosystem vibrancy” better aligns with platform-based companies’ interests. Furthermore, such an approach dovetails perfectly with Baidu’s overarching strategies — safeguarding its search engine dominance while deepening penetration into B2B markets. Concurrently, Kunlun plans to list both on Hong Kong’s Stock Exchange and China’s STAR Market; this independent capitalization drive compels Baidu to swiftly craft a compelling new narrative for future growth.