ZStack Named a Representative Vendor for AI Infrastructure in IDC’s AI Agent Report

Recently, International Data Corporation (IDC) released its report, “IDC Market Glance: China AI Agent Market Overview, 3Q25” (hereinafter referred to as “the report”). ZStack was recognized as a representative vendor in two key segments—“Model Development Service Platforms” and “AI Infrastructure Software”—for the strong toolchain support provided by its AI infrastructure platform, ZStack AIOS, for AI Agent development.

The report focuses on the latest developments in China’s AI Agent market in Q3 2025. It highlights that AI Agents are evolving beyond the concept of “large model applications,” forming a technology framework centered on cognitive generalization, closed-loop action, and memory evolution. This shift is propelling enterprise AI applications from pilots to large-scale deployments.

Within this progression:

Model development service platforms take on complex task decomposition, systems integration, and full lifecycle management, becoming the core vehicle for scaling AI Agent deployments.AI infrastructure software orchestrates compute resources, cloud–edge collaboration, and high-performance networking—capabilities whose performance directly determines an agent’s efficiency and reliability in handling complex tasks.

Together, these layers form an essential foundation for the commercialization and scaled rollout of AI Agents.

ZStack AIOS is a self-developed, productized, and standardized next-generation private AI infrastructure platform. With its all-in-one architecture, ZStack AIOS accelerates enterprise AI innovation across three dimensions—compute orchestration, large-model training and inference, and AI application service development—delivering significant cost-efficiency gains. ZStack stood out in IDC’s report thanks to the following strengths:

1.Deep optimization for diverse CPU/GPU stacks and dynamic, fine-grained vGPU partitioning, enabling intelligent compute scheduling to reduce costs and improve efficiency.

2.Support for deployments ranging from lightweight models to ultra-large-scale full-parameter models, covering diverse use cases.

3.Compatibility with a broad set of mainstream open-source models to meet varied AI scenarios and drive continuous enterprise AI iteration.

4.A visual fine-tuning workbench and a multidimensional model evaluation system to help enterprises build bespoke AI engines.

5.Enterprise-grade security across data, access, and auditing, plus department-level data sandboxing to ensure dual isolation of model services and application data.

6.Localized, low-code LLMOps for rapid verticalization of AI services, enabling enterprises to incubate private AI experts with ease.

As a key platform for private deployment of enterprise large models, ZStack AIOS has been adopted across government, finance, transportation, education and research, and media, helping many organizations successfully implement private large-model solutions. Looking ahead, ZStack will continue to leverage its integrated “platform + infrastructure” capabilities to help more customers move from “usable” to “truly useful” AI Agent deployments at scale, accelerating the market’s journey into a value-realization phase.