CITIC Securities: The development of Agents has become a consensus among major manufacturers, and new regulations stimulate the vitality of the mergers and acquisitions market

According to the Zhitong Finance APP, Guotai Junan released a research report stating that ByteDance and Google have launched related new products, accelerating the arrival of the first year of large-scale commercialization of AI Agents. The firm indicated that the iteration of AI technology and the implementation of Agents are steadily advancing among major companies, and AI commercialization continues to develop, further supporting their judgment that 2025 will be the first year for large-scale commercialization of AI Agents. The new rules from the China Securities Regulatory Commission (CSRC) are expected to accelerate the integration of segmented fields driven by the demand for digital transformation in the computer sector. The firm maintains an "overweight" rating for the computer sector.

Guotai Junan's main points are as follows:

ByteDance upgrades multiple models and launches the all-scenario intelligent agent Data Agent

On May 13, ByteDance's Volcano Engine released the Doubao Video Generation Model Seedance 1.0 lite, which supports text-to-video and image-to-video generation, with video generation durations of 5s and 10s, and resolutions of 480P and 720P; the Doubao 1.5 Visual Deep Thinking Model, which has only 20B activation parameters but possesses strong multimodal understanding and reasoning capabilities, ranking in the top tier in video understanding, visual reasoning, and GUI Agent capabilities. Additionally, ByteDance upgraded the Doubao Music Model. Meanwhile, Data Agent made its official debut, and Trae integrated the Doubao Deep Thinking Model with a complete upgrade. Volcano Engine is helping enterprises connect the application chain from business to intelligent agents with a more powerful model matrix and a richer set of intelligent tools. The firm believes that the iteration of AI technology and the implementation of Agents are steadily advancing among major companies, and AI commercialization continues to develop.

CSRC modifies the "Administrative Measures for Major Asset Restructuring of Listed Companies," deepening the reform of the listed company merger and acquisition market

The main contents of this modification include: first, establishing a phased payment mechanism for restructuring share consideration, such as extending the validity period of the registration decision for issuing shares to purchase assets in phases to 48 months; second, increasing tolerance for changes in financial conditions, industry competition, and related party transactions; third, introducing a simplified review procedure for restructuring, clarifying that restructuring transactions eligible for the simplified review procedure do not require review by the M&A Committee of the stock exchange, with the CSRC making a registration decision within 5 working days; fourth, improving lock-up period rules to support mergers between listed companies; fifth, encouraging private equity funds to participate in the mergers and acquisitions of listed companies. The firm believes that this move by the CSRC is expected to further stimulate the vitality of the merger and acquisition market and accelerate the integration of segmented fields driven by the demand for digital transformation in the computer sector.

Google releases the general intelligent agent AlphaEvolve

According to the official WeChat account of China Science Daily, on May 14, Google DeepMind launched the general scientific AI system AlphaEvolve, which can autonomously generate and improve algorithm codes, successfully solving significant problems in mathematics and computer science. DeepMind defines AlphaEvolve as an "agent," with its uniqueness focusing on the core aspects of the research process. After users input questions, evaluation criteria, and initial proposals, the LLM generates hundreds to thousands of improved proposals, which are then filtered by the "evaluator" algorithm based on preset indicators of the optimal solution In addition to solving mathematical problems, AlphaEvolve has been used to tackle real-world challenges—improving the design of the next-generation tensor processing units (computing chips designed for AI) and finding more efficient ways to utilize global computing resources, saving a total of 0.7% of resources. The bank believes that the implementation of Agents has basically become a consensus among major tech companies, and the development speed is expected to accelerate further in the future.

Risk Warning: The commercial implementation of AI may fall short of expectations, and policy implementation may also fall short of expectations