Cathay Securities and Haitong Securities: Google has clearly defined its 2C Agent strategy for the first time, and the development of Agents has become a consensus among major companies

According to Zhitong Finance APP, Guotai Junan released a research report stating that Google has brought a comprehensive upgrade to AI this time. In addition to upgrading the foundational models, Google has also revolutionarily integrated AI into its core business of search, which represents its high confidence in AI. The 3P 2CAgent strategy further confirms the bank's view: "Agents have become a consensus among major companies." Google's ecological advantages are expected to bring about a truly ultimate personalized intelligent assistant, maintaining an "overweight" rating for the computer sector. Guotai Junan's main points are as follows: Google's AI "Full Package" Comprehensive Upgrade According to Tencent Technology's official WeChat account, on May 21, at the Google 2025 I/O conference, Google released a large number of AI-related new products. In terms of foundational models, the conference announced: Gemini 2.5 Flash new version: millisecond-level inference, sparse MoE routing, stronger performance; Gemini 2.5 Pro + Deep Think: "multi-chain self-review" inference mode, deeper thinking; Imagen 4: new image model, 4K quality, 1.2s text-to-image, zero distortion in text layout; Veo 3 + Flow application: the first video model supporting native dialogue and synchronized sound generation, supporting 1080p, 60 frames output. Flow can be used for editing; Stitch: natural language directly converted into front-end UI. Veo 3 is particularly noteworthy as it generates both visuals and sound simultaneously with a single model for the first time, achieving a dimensionality reduction attack on other AI video models that rely on "video + post-dubbing." However, the Ultra subscription plan, costing up to $249/month, is required to unlock the full functionality of Veo 3 and the complete version of Flow editing, hindering its rapid promotion. Comprehensively Introducing AIMode, Google Revolutionarily Upgrades the Search System On the search side, the conference announced: AIMode: a more advanced AI search mode that can handle complex questions and continue to ask follow-up questions; Shopping Graph 2.0: AI search shopping recommendation engine, containing 500 billion SKUs, supporting virtual fitting; Deep Search: similar to OpenAI's deep research function, capable of providing more complex and systematic information; Search Live: allows for Q&A and search through a real-time camera, object labeling. It is evident that Google is betting on a bold proposition: AI can not only enhance search but will fundamentally reshape the essence of search. It must provide sufficiently valuable direct answers while not completely eliminating the user's motivation to click on ads and third-party websites. It must find a delicate balance between "AI answers everything" and "retaining the search advertising ecosystem." The bank believes that Google has revolutionarily upgraded its search system this time, which will advance alongside the Gemini App in the future. Google Clearly Defines the 2CAgent 3P Strategy for the First Time, Ecological Advantages May Fully Unleash Regarding Agents, Google released its version of Computer Use / Operator—Project Mariner. In addition to completing some basic tasks, it can also help users purchase baseball game tickets or buy groceries online; users only need to chat with Google's AI Agent, and it will visit websites and take action At this press conference, Google clarified its to CAgent strategy, summarized as 3P: Personal, Proactive, Powerful. Personalization relies on Gmail, Docs, Maps, and the private memory of Gemini and Search within the Google ecosystem. Proactivity is built on personalized data, predicting what users may need to be reminded of or tasks they might need to do based on their schedules and past habits. Power is supported by top models like Deep Think and Gemini 2.5 Pro. The firm believes that Google's greatest advantage is its vast ecosystem to build personalized information for users, making individual predictions possible, which is making the implementation of 2CAgent feasible. Risk Warning: AI technology development may fall short of expectations, and AI commercialization may not meet expectations