IDC: AI Applications Use Exponential Fission, New Cloud Vendors Reconstruct Agentic Infrastructure

The Zhitong Finance App learned that the International Data Corporation (IDC) published an article on October 13 stating that the strong adoption of generative AI and agents has greatly boosted the growth of AI infrastructure. In the Agentic era, the once relatively linear technology stack has evolved into a dynamic, connected ecosystem. This change has not only expanded the role of established companies, but also stimulated many enterprises to enter the AI infrastructure market across borders. In the third quarter of 2025, IDC studied the changing AI infrastructure market in the Asia-Pacific region and the new challenges it brings. It also focused on investigating how typical vendors in the entire ecosystem can adapt to these new requirements, and recently released the “AI Native Cloud/New Cloud Vendors Reconstruct Agentic Infrastructure” report.

Market pattern of China's big model platform in 2024

Driven by big models, there are currently thousands of AI application companies around the world. These applications include basic large-scale model-based general assistants, as well as model-based AI search, social networking, audio and video generation, etc. According to data from IDC Global Tracking, the number of global consumers using these generative AI apps is expected to exceed 5.7 billion by the end of 2029, far exceeding 3.1 billion by the end of 2024. The compound growth rate of the number of users is expected to reach 13.2% during the forecast period. Overall paying users are expected to grow at a CAGR of 32.3%, while non-paying subscribers will grow at a CAGR of 12.0%. In the Asia-Pacific market, consumer adoption of GenAI is also showing a clear and accelerating trend.

The enterprise market is also expanding generative AI to various departments. Whether it is a consumer-level AI application or an enterprise-level AI application, large-scale implementation is inseparable from various AI application development platforms, tools and ecosystems, as well as the infrastructure that supports AI applications. In particular, training for different application scenarios and tuning of various large models require the guarantee of high computational efficiency and high reliability infrastructure.

Dedicated AI infrastructure accelerates the implementation of AI applications

At a time when the AI infrastructure market is exploding, more and more new entrants are beginning to pour into this market. As shown in the figure below,

Public cloud providers provide global-scale and flexible integrated AI services to accelerate enterprise AI adoption; edge and hybrid cloud providers bring AI closer to data sources to support real-time processing and localized deployment in key scenarios;

AI native cloud providers focus on high-performance GPU clusters and cost-effective infrastructure to meet training and inference needs;

Data center and managed service providers are committed to creating safe, reliable, and high-performance operating environments;

AI hardware and platform providers are advancing the development of specialized chips and systems to power the next generation of AI performance. Additionally, more and more telecom companies are entering this field, using their extensive network infrastructure to support edge AI deployments, enhance connectivity for AI workloads, and provide integrated AI-driven services across distributed environments.

IDC's advice to AI infrastructure buyers

As GenAI moves from an experimental phase to enterprise-grade deployment, the need for infrastructure is now more than just computing power. Safety, resilience, vendor support and cost efficiency are key factors affecting successful implementation. IDC recommends that users consider the following factors when selecting models:

1 Priority should be given to selecting dedicated AI infrastructure close to the geographical area: to handle latency-sensitive workloads and sovereign workloads. Look for providers that deploy high-performance computing or custom chips in the region to ensure low latency AI performance, compliance with data sovereignty laws, and scalable GenAI and LLM training/inference capabilities.

2 Choose vendors carefully: Ensure flexible deployment models to support hybrid, public, and private proprietary AI environments. Businesses moving from PoC to production should prioritize partners that provide modular, customizable computing services (such as GPU clusters, on-demand or hybrid orchestration engines) that can be dynamically scaled and cost controlled according to business needs and data governance requirements. In addition, priority should be given to choosing manufacturers with stable supply chains and sufficient resources.

3 Strong security and compliance are always at the core: Choose suppliers that can demonstrate strong enterprise-grade security practices, certifications, and compliance with relevant data protection regulations. Verify that features such as workload isolation, zero trust access, and encryption meet your industry and regional needs to protect sensitive AI data.

Typical vendor recommendations: In the Asia-Pacific market, large full-stack cloud vendors such as Amazon Cloud Technology and Google Cloud have won the favor of many AI application companies. For new cloud vendors/AI native cloud vendors, this report focuses on IDC research and recommends CoreWeave and GMI Cloud.