当前全球智能体正迎来爆发式增长，身为其核心基础设施的Agentic AI开发平台正进入功能快速丰富期。先进平台不仅要能支持极致的工作负载扩展与弹 性、知识管理与上下文工程、工作流编排、无代码开发，还要能满足精细化计费等更多功能。基于此，Omdia对亚太主流厂商的Agentic AI开发平台从七大 维度进行详细且严格的评估。其中，阿里云在模型支持、上下文工程、多智能体框架、运营与生命周期管理、开源与社区等5大核心能力获得最高评级，是 入选厂商中获最高评级数仅次于AWS的第二领先厂商。 报告强调，阿里云深耕全球云计算领域多年，已确立其作为全栈AI服务供应商的顶级地位。在智能体开发方面，阿里云通过提供全面的企业级功能套件， 在一众厂商中脱颖而出，成为企业构建和部署自有AI智能体的首选。尤其是基于阿里云函数计算 FC 构建的一站式Agentic AI基础设施平台——AgentRun， 以其卓越的性能优化、极高的性价比以及企业级的安全保障，受到研究机构的高度评价，可使企业平均TCO降低60%，让开发者可以专注于Agent的核心业 务逻辑创新，无需自建和管理底层基础设施。 2月12日，国际市场研究机构Omdia发布《20 ...

Investment Rating - The report does not explicitly state an investment rating for the industry Core Insights - The report emphasizes the transition from traditional application development to Serverless AI application development, highlighting the need for a new infrastructure paradigm that supports AI agents and their unique requirements [10][27] - It identifies the importance of dynamic elasticity and task-driven orchestration in AI-native architectures, which allows for efficient resource allocation and management [19][24] - The report discusses the advantages of Serverless AI runtimes, including reduced operational burdens, cost efficiency, and enhanced developer focus on business innovation rather than infrastructure [26][34] Summary by Sections 01 Enterprise-level AI Application Development Runtime Selection - AI-native paradigms demand new infrastructure requirements, focusing on agent-centric services rather than traditional user-centric models [13][15] - The infrastructure must support state persistence and low-latency access, enabling agents to maintain memory and personality [17] - Embracing uncertainty is crucial, with infrastructure designed to lower risks associated with non-deterministic outputs from large language models (LLMs) [21] - The transition from traditional architectures to AI-native architectures is necessary for effective application development [26] 02 Key Technologies of Serverless AI Runtime - Serverless platforms provide heterogeneous computing capabilities, integrating various programming languages and event-driven architectures [33][34] - The report highlights the importance of security isolation and automatic disaster recovery in Serverless AI runtimes [38][42] - Serverless GPU services are emphasized for their rapid cold start capabilities and efficient resource utilization, significantly reducing costs [43][49] 03 Customer Cases – Serverless + AI Simplifying Application Development - The report presents internal case studies from Alibaba, showcasing successful implementations of Serverless runtimes in building models and AI tools [87][90] - It illustrates how Serverless AI runtimes have become core to Alibaba Cloud's AI-native applications, enhancing performance and reducing operational costs [90][92] - The case studies demonstrate the ability to handle high concurrency and low latency requirements in real-time AI applications [93][99]