In 2025, the development of Agentic AI has led to a paradigm shift in the field of artificial intelligence. Also known as agent-based AI, this stage occurs after Generative AI and before Physical AI. Agentic AI enables various technologies to automatically adapt and make decisions. Its key feature is perceiving and understanding context, solving problems through reasoning, and taking action after planning. However, the implementation of Agentic AI has increased the demand for computing power. Compared to traditional large language models, when the token generation speed is the same, the computing power requirement is more than 10 times higher. The current demand for computing power has just opened the door to a vast number of applications. High-density computing infrastructure poses continuous challenges regarding power supply, cooling, large-scale parallel computing, and scalability for AI infrastructure.
On September 18, 2026, Frost & Sullivan, also referred to as 'Frost & Sullivan', in collaboration with LeadLeo Research Institute, released the '2026 H1 China Full-Stack AI Cloud Service Market Report - Ready-to-Use Cloud Services' (hereinafter referred to as the 'Report'). This report focuses on the 'China Full-Stack AI Cloud Service' sector, with full-stack technical integration and ecosystem collaboration as its core research direction. The research period covers the entire year 2025 and H1 2026. The purpose of the study is to systematically analyze the key technologies, commercialization, competitive landscape, and future trends of full-stack AI cloud services. Through in-depth analysis, it aims to reveal the critical role of full-stack AI cloud services and help identify the strategic opportunity window for the deep integration of AI and cloud computing.
WeChat QR code scanning to access the report
The following is an excerpt from the white paper. For detailed content, please scan the QR code:
Analysis of the Full-Stack AI Cloud Service Market
01
Cloud computing is the solid foundation for AI development
In 2025, the development of Agentic AI has led to a paradigm shift in the field of artificial intelligence. Also known as agent-based AI, this stage occurs after Generative AI and before Physical AI. Agentic AI enables various technologies to automatically adapt and make decisions. A key feature of Agentic AI is its ability to perceive and understand context, solve problems through reasoning, and take action after making a plan. To support running high-demand AI workloads in the cloud, AI clouds must combine performance, advanced networking, flexibility, security, and scalability to meet customers' needs for reasoning and training of AI workloads and new Agentic applications.
Source: Literature search, Frost & Sullivan analysis
02
Out-of-the-box full-stack AI cloud services
With the development of cloud computing, cloud services have evolved from traditional computing and storage resource provisioning to higher-value AI cloud services. AI cloud services not only provide underlying computing power support but also integrate capabilities such as model training, inference platforms, and Model as a Service (MaaS), promoting the application of artificial intelligence technologies across various industries. Full-stack AI cloud services focus on ready-to-use functionality, full-stack integration, and flexible expansion, aiming to reduce the barriers to using AI.
Source: Literature search, Frost & Sullivan analysis
03
IaaS layer of full-stack AI cloud services
The role of IaaS layer is to provide computing power services and large-scale data storage centers for upper-level Paas and SaaS, offering underlying resources such as computing power infrastructure (like GPU/TPU clusters and distributed computing resources), storage, and networks, to meet the flexible expansion requirements for AI model training and inference. As a basic server, IaaS is typically delivered through combinations with other services.
IaaS products are usually AI cloud infrastructure of various manufacturers, such as AI heterogeneous computing platforms. As a cloud service model, users can remotely access virtualized hardware resources through a management platform and pay based on actual consumption. It provides flexibility, scalability, and economic benefits, allowing enterprises to quickly adapt to technological changes and demand fluctuations, thus focusing on core business rather than infrastructure management.
04
PaaS layer of full-stack AI cloud services
Frost & Sullivan Frost & Sullivan China LeadLeo Research Institute LeadLeo YUAN CAPITAL SULLIVAN TELE-TREND CLOUD TECHNOLOGY TradeGo MagnaTEC Automotive & Mobility Environmental Protection & Energy Saving Technology Logistics & Supply Chain MATERNAL AND INFANT Education & Training Real Estate & Property Catering & New Retailing Advanced Materials Healthcare & Life Sciences Semiconductor & Chip COMMERCIAL AVIATION Technology, Media and Telecom LANDSCAPING Big Data & AI Infrastructure Construction & Utilities Culture & Entertainment AGRICULTURE, FORESTRY ANIMAL HUSBANDRY AND FISHERY Food & Beverage Fintech SHIPPING AND PORTS Dual Carbon & New Energy Mining & Metals Public Sector Cross-Border E-commerce Trade Building Technology, Construction & Decoration Beauty & Fashion Smart Homes Digital Infrastructure Enterprise Services Consumer Electronics
As the data provider for AI applications, data owners are crucial for the development, governance, and assetization of initial metadata. In full-stack AI cloud services, the data governance platform ensures that data used for training and inference is accurate, complete, and consistent through functions such as metadata management, data quality monitoring, and data cleaning. High-quality data is a prerequisite for training high-performance and reliable AI models.
Source: Literature search, Frost & Sullivan analysis
Based on model development, scenario adaptation, inference deployment, and product application, full-stack AI cloud services are becoming platform-based and simplified, aiming to provide users with a more complete product construction solution.
05
MaaS layer of full-stack AI cloud services
In the cloud services of the AI era, the application layer has evolved from individual AI models and their capabilities into reusable services known as MaaS. The emergence of AI Agents has further transformed the nature of enterprise services on the application layer, shifting from "letting customers work independently" to "completing tasks for customers". The advent of new forms such as Serverless reasoning, multi-model routing engines, and intelligent routing pricing (Token billing) has also changed the current market paradigm.
06
Full-stack AI cloud services are building a new paradigm of Agentic Cloud
The Agent workload features "irregular flexibility, short lifecycle, and rapid growth followed by rapid decline," requiring the construction of a complete Agent runtime environment. This includes providing a lightweight and efficient sandbox execution environment to achieve task-level isolation, multi-Agent collaboration and orchestration capabilities, cross-task memory management, smooth data flow mechanisms, and comprehensive intelligent operation and maintenance capabilities.
Agentic Cloud differs from the previous generation "AI Native Cloud" which focused on model production iteration, but it is based on AI Native Cloud. AI Native Cloud focuses more on model production iteration, providing elastic and efficient computing power scheduling. Agentic Cloud targets the runtime of agents, with the core being to provide a full set of enabling capabilities such as sandbox, AI gateway, memory management, security protection, and orchestration governance, making the cloud the operating system for Agent operation.
07
Competitive landscape of full-stack AI cloud services
Current full-stack AI cloud providers mainly fall into two categories. The first category includes comprehensive cloud computing providers such as Alibaba Cloud and Baidu Smart Cloud, which have fully implemented AI strategies as HyperScalers. These providers rely on cloud computing as their foundation and possess large-scale infrastructure and cost control capabilities. The second category consists of Neo Cloud providers like SenseTime, which have been deeply integrated with AI from their inception, possessing an AI DNA, and have gradually developed into full-stack AI cloud providers.
