ZStack Selected as a Representative Vendor in IDC’s Market Report on Large Model Inference Computing Power

Recently, the International Data Corporation (IDC) released the “China Large Model Inference Computing Power Market Analysis Report” (August 2025). The report analyzes the current status, structure, and market size of China’s large model inference computing power market and outlines key considerations for building computing power for large models and generative AI. ZStack, with its AI Infra platform ZStack AIOS, has been recognized as a representative vendor in the construction of China’s large model infrastructure, becoming a leading company in the field of AI infrastructure management and services.

The IDC report highlights that as generative AI applications rapidly penetrate various industries, enterprises have significantly increased their demands for computing power scheduling efficiency, data security, and model management. ZStack AIOS is a self-developed, productized, and standardized next-generation private AI Infra platform. With its all-in-one platform advantages, ZStack AIOS comprehensively supports enterprises in AI innovation across three key areas: computing power scheduling, large model training and inference, and AI application service development, significantly enhancing cost-effectiveness. It stands out due to the following core advantages:

1.Supports deep adaptation of various CPUs/GPUs and dynamic fine-grained partitioning of vGPUs, achieving intelligent computing power scheduling to help enterprises reduce costs and increase efficiency.

2.Supports deployment of models ranging from lightweight to ultra-large scale, meeting diverse scenario requirements.

3.Compatible with a wide range of mainstream open-source models, catering to different AI scenario needs, enabling continuous iteration of enterprise AI innovations.

4.Provides a visual fine-tuning workbench and a multi-dimensional model evaluation system to help enterprises build customized AI engines.

5.Ensures enterprise-level data security across three dimensions: data, access, and auditing, with a departmental data sandbox mechanism achieving dual isolation of model services and application data.

6.Supports localized low-code LLMOps applications, enabling rapid and deep penetration of AI business into vertical fields, allowing enterprises to easily incubate private AI experts.

ZStack AI Infra Platform AIOS Product Architecture Diagram

The IDC report indicates that the market size of China’s generative AI infrastructure will reach 35.89 billion yuan in 2024. Industries such as finance and government, due to high data sensitivity, are more inclined to build their own AI computing power centers. As an important platform for private deployment of large models in enterprises, ZStack AIOS has been implemented in various industries, including government, finance, transportation, education, and media, effectively helping numerous enterprises achieve private deployment of large models.

In the government sector, ZStack has collaborated with provincial laboratories to jointly launch a multimodal ubiquitous intelligent infrastructure platform, providing integrated cloud-edge collaboration, supporting intelligent Q&A, precise intent understanding, and situational awareness.

In the financial industry, ZStack serves a Shenzhen insurance company for scenarios such as claims review, intelligent customer service, and risk prediction.

In the education sector, ZStack assists a Shanghai university in building an Agent intelligent platform, enabling faculty and students to quickly create campus assistants through low-code methods and allowing students to creatively build various Agent applications.

In the transportation industry, ZStack supports Southwest Expressway with a multi-machine, multi-card setup to host the DeepSeek 671B quantized model version, supporting video image collection and data analysis decision-making.

In the media industry, ZStack assists a large new media company in establishing a generative AI media integration innovation platform for scenarios such as video generation, text-to-image, intelligent customer service, multimodal search, and ticket purchasing.

In the future, ZStack will continue to focus on the operation of intelligent computing centers, integrated training and inference platforms, and industry-specific solutions, working with ecosystem partners to promote the large-scale application of large models across various industries.