沙利文发布《开箱即用的AI云服务——2025中国全栈AI云服务市场报告》
Frost & Sullivan Releases the " Turnkey AI Cloud Solutions—— 2025 China’s Full-Stack AI Cloud Services Market Report"
2026年，中国AI日均Token调用量突破140万亿，两年增长超1000倍。算力的消耗单位已从“GPU小时”演进为“Token量”，全栈AI云厂商的竞争焦点兼具“算力规模”与“每Token成本效率”，云服务的定价体系与价值主张正在逐步重构。以云计算资源为基础底座、大模型能力为核心驱动、智能体服务为基本单元，具备理解用户意图并自主完成复杂任务能力的新一代云计算服务体系。云的主要消费者正在从人变为智能体，标志着云服务从提供“资源”和“算力”转向输出“能力”和“智力”。 Anthropic MCP、Google A2A、IBM ACP三大协议的相继发布与普及，将多智能体协作从技术实验推向企业生产环境。全球57%企业已在生产环境部署多步工作流AI Agent。全栈AI云厂商竞争进化为：谁能提供最兼容、最低摩擦的Agent互联基础设施，谁就赢得开发者生态。AI云服务的竞争已进入“生态战争”阶段：企业不再从零开发Agent，而是在云市场上选购，云厂商通过不同层级服务的垂直整合，构建IaaS+PaaS+MaaS三位一体模式打造场景化竞争力。
本报告聚焦“中国全栈AI云服务”领域，以全栈式技术整合与生态协同为核心研究方向，研究周期覆盖2025年全年。研究旨在系统梳理全栈AI云服务的关键技术、商业化、竞争格局及未来趋势。通过深度分析，揭示全栈AI云服务的关键作用，助力把握AI与云计算深度融合的战略机遇窗口。
In 2026, China’s daily AI Token call volume exceeded 140 trillion, representing a more than 1,000-fold increase within two years. The unit for measuring computing power consumption has shifted from "GPU hours" to "Token volume". Full-stack AI cloud vendors now compete on two fronts: computing scale and cost efficiency per Token, driving a fundamental restructuring of cloud service pricing frameworks and value propositions.
A new-generation cloud computing ecosystem has taken shape, built on cloud computing resources as the underlying infrastructure, large model capabilities as core drivers, and intelligent agents as basic functional units. This system can interpret user intent and autonomously execute sophisticated multi-step tasks.
The primary consumers of cloud services are shifting from human users to AI agents, marking a paradigm shift where cloud providers no longer merely supply raw resources and computing power, but deliver functional capabilities and machine intelligence.
The successive rollout and mainstream adoption of three major protocols — Anthropic’s Model Context Protocol (MCP), Google’s Agent-to-Agent Protocol (A2A), and IBM’s Agent Communication Protocol (ACP) — have moved multi-agent collaboration out of experimental labs and into enterprise production environments. Globally, 57% of enterprises have deployed multi-step AI agent workflows in production.
Competition among full-stack AI cloud vendors has pivoted to a new battlefield: whichever delivers the most compatible, lowest-friction inter-agent connectivity infrastructure will capture dominant developer ecosystem share. Competition in AI cloud services has entered the "ecosystem warfare" era. Instead of building agents from scratch, enterprises source ready-made agents via cloud marketplaces. Cloud vendors build scenario-specific competitiveness through vertical integration across service layers, establishing an integrated IaaS+PaaS+MaaS tripartite architecture.
This report focuses on China’s full-stack AI cloud service sector, with core research themes covering full-stack technology integration and ecosystem synergy, and its research scope spans the full calendar year of 2025. The research systematically sorts out core technologies, commercialization models, competitive landscape and future trends of full-stack AI cloud services. Through in-depth analysis, it reveals the pivotal value of full-stack AI cloud services and helps stakeholders seize strategic opportunities arising from deep integration between AI and cloud computing.
点击以下链接，查看节选版报告：
沙利文 李女士
E-mail: livia.li@frostchina.com
