#我看好的AIAgent AI Agent Explosion! Who Are the Biggest Winners? These Five Major Tracks Welcome a Historic Opportunity

An infrastructure construction wave led by AI Agents is sweeping across the industry, ushering in a new growth cycle in computing power, networks, and security.
2025 is dubbed the "AI Agent Year Zero" by industry insiders, as a profound technological revolution is quietly unfolding. As tech giants like Alibaba and ByteDance upgrade their AI Agent platforms, shifting from "chatting" chatbots to "doing things" intelligent assistants, the entire industry chain is facing unprecedented reshaping.
Data shows that China's AI Agent industry market size is expected to grow from 147.3 billion yuan in 2024 to 3 trillion yuan in 2028, with an astonishing compound annual growth rate. In this technological revolution, which tracks will benefit the most? Let’s find out.

01, Cloud Computing and Intelligent Computing Centers:
Exponential surge in computing demand—cloud computing and intelligent computing centers are the core beneficiaries of the AI Agent explosion.
Unlike traditional single large model inference, AI Agents require multiple calls to large models for complex task decomposition and planning, leading to exponential growth in computing power demand.
Take ByteDance’s "Kouzi" 2.0 platform as an example. Its new Agent Plan feature allows AI to develop long-term plans and execute them continuously, meaning a single AI Agent task may consume dozens of times more computing resources than traditional chat. This demand has prompted major cloud service providers to adjust their pricing strategies in 2025.
Foxconn’s 2025 performance forecast shows that the company's cloud service provider AI server revenue increased over 3 times year-on-year, and revenue from 800G+ high-speed switches increased by as much as 13 times. This data confirms the explosive growth in AI computing power demand.
Lenovo Group’s ISG revenue has achieved over 60% consecutive growth, further indicating that the computing power market has entered a high prosperity cycle. As AI Agent tasks become increasingly complex, the growth in computing power demand is far from reaching its ceiling.

02, CDN and Edge Computing:
Low latency response becomes a necessity—AI Agent’s "action" capability demands extremely low network latency.
When Agents need real-time tool calls or access to the latest data, traditional cloud architectures struggle to meet low-latency requirements, creating huge opportunities for CDN and edge computing.
Traditional CDNs mainly handle content distribution, but the demands of the AI Agent era have fundamentally changed: a single Agent task may involve thousands of data access requests, each requiring real-time responses. This high-frequency interaction mode causes explosive growth in CDN usage.
Take Alibaba’s Qianwen App as an example. When a user says, "Help me order a coffee," the Agent needs to call multiple tools in real-time—location services, merchant recommendations, order generation, payment interfaces, etc.—each relying on low-latency responses from edge nodes. This task-based interaction is becoming the norm for AI Agents.
Another advantage of edge computing is localized data processing. To address privacy and security challenges brought by AI Agents, deploying computing tasks at edge nodes is an ideal solution. This also explains why many telecom operators and CDN providers are actively building edge computing platforms.

03, High-Speed Networks and Optical Communications:
The foundation for multimodal interaction of Agents—AI Agents’ multimodal interaction and real-time inference capabilities—require comprehensive upgrades in network bandwidth, latency, and stability.
Optical module manufacturers are direct beneficiaries of this trend. As intra-data center traffic surges, demand for high-speed optical modules continues to rise. 800G and above high-speed optical modules are becoming standard in data centers, and the industrialization of 1.6T optical modules is accelerating.
In the network equipment sector, the proliferation of AI Agents is accelerating the "East Data West Computation" project. Due to limited energy resources in eastern regions and abundant clean energy in western regions, which are more suitable for large-scale data center construction, efficient network connectivity is key to achieving nationwide computing power scheduling.
The evolution of 5G-A and future 6G technologies also opens up possibilities for mobile scenarios of AI Agents. China Mobile’s Lingxi Intelligent Agent 2.0 has demonstrated the ability to provide intelligent services in mobile environments, supported by high-speed networks.

04, Computing Chips and Servers:
Hardware infrastructure is experiencing an upgrade wave—since inference accounts for over two-thirds of total computing power consumption in AI Agents, demand for AI chips and servers continues to grow.
Unlike training, inference needs of AI Agents are more dispersed and real-time, driving diversification in the inference chip market. Besides traditional GPU giants, many domestic chip companies are actively deploying in this field, launching inference chip solutions suitable for different scenarios.
Servers, as the direct carriers of computing power, are undergoing a new round of technological transformation. Lenovo’s Wanquan Heterogeneous Intelligent Computing Platform supports multiple AI chip architectures and can automatically perform AI calculations and deploy models or inference services. This flexibility is an ideal solution to meet the diverse needs of AI Agents.
Breakthroughs in green computing power technology also make large-scale deployment of AI Agent infrastructure possible. Lenovo’s Poseidon liquid cooling solution can reduce PUE below 1.1, and operating at full capacity can reduce carbon emissions by approximately 3,179 tons annually. Such low-power solutions are becoming new industry standards.

05, Network Security and AI Security:
New challenges in autonomous execution—AI Agents’ autonomous operation and cross-system capabilities bring new security challenges. When Agents can proactively call tools and perform complex tasks, security auditing and permission control become essential features.
On one hand, AI Agents need access to multiple systems to complete tasks, expanding potential attack surfaces; on the other hand, the transparency of the Agent’s autonomous decision-making process is necessary to ensure behaviors meet expectations. These needs have given rise to a new generation of AI security markets.
As the capabilities of AI Agents expand, establishing corresponding governance systems is crucial. "Development and governance are not opposing forces; the key is to strengthen bottom-line standards and guide artificial intelligence to evolve in a beneficial, safe, and fair direction."
As AI Agents move from concept to implementation, the five major tracks of cloud computing, CDN, optical communications, computing chips, and cybersecurity are already seeing clear growth opportunities.

The next explosion point may be at the edge—when AI PC and smartphones deeply integrate Agent capabilities, the combination of edge computing and edge-side computing power will open new growth space.
The true revolution of AI Agents lies in reshaping the human-machine interaction paradigm—from "humans adapting to machines" to "machines proactively serving." This shift will drive the entire ICT industry upgrade and create a new market worth trillions of yuan.