AI's Reshaping of Cloud Architecture: From Vertical Stacks to Agentic Services

An Omdia report, "Global AI Cloud Stack 2026: Rethinking Cloud in the Agentic AI Era," published on July 9, 2026, reveals a significant transformation in cloud architectures driven by artificial intelligence. The report identifies that the rise of large language models (LLMs) since 2023 is causing a breakdown of established vertical layered architectures. This shift is leading to a restructuring of the traditional cloud stack into new functional components. Omdia has redefined the AI Cloud technical architecture into three distinct layers: AI Cloud Infrastructure (AI Cloud Infra), Model-as-a-Service (MaaS), and Agent-as-a-Service (AaaS). The report also notes that a single model API call can now encompass functions across multiple layers of cloud-native architecture, integrating previously independent layers into the model itself. This report is a wake-up call for cloud architects, DevOps engineers, and IT leaders. The traditional understanding of cloud services (IaaS, PaaS, SaaS) is becoming insufficient to describe the AI-native era. The emergence of MaaS and AaaS signifies a profound shift in how applications are built, deployed, and consumed. For practitioners, this means a need to re-evaluate existing architectural patterns, skill sets, and procurement strategies. The collapse of vertical architectures implies that the lines between infrastructure, platform, and application logic are blurring, requiring a more holistic and AI-centric approach to system design. The financial implications are also significant, with a move towards token-based billing models, which demands new cost optimization and management strategies. The evolution described by Omdia is a natural progression of the cloud-native movement, accelerated by the unprecedented capabilities of AI. While cloud-native principles emphasized microservices, containers, and APIs for modularity and scalability, AI is now pushing this further by embedding intelligence directly into the core fabric of cloud services. This trend aligns with the broader industry focus on "AI-first" development, where AI is not merely an add-on but a foundational element of software and infrastructure. The increasing sophistication of LLMs and generative AI has necessitated a more integrated approach, where the model itself becomes a central architectural component, abstracting away underlying complexities and offering higher-level services. This also reflects the growing demand for specialized AI infrastructure and platforms that can efficiently handle the compute and data requirements of modern AI workloads, leading to the growth of the AI Cloud Infra market, valued at $65.74 billion in 2025. Practitioners should begin by understanding the implications of MaaS and AaaS for their current and future projects. This involves exploring how model APIs can consolidate functionalities previously spread across multiple services and evaluating the potential for agentic workflows to automate complex tasks. Cloud architects must start designing with these new layers in mind, considering how to integrate MaaS and AaaS effectively while managing data governance, security, and cost in an AI-centric environment. Furthermore, the report's emphasis on token-based billing necessitates a proactive approach to FinOps, where teams need to develop new methods for tracking and optimizing AI consumption. Investing in training for AI-specific cloud architectural patterns and understanding the nuances of AI Cloud Infra will be critical for staying ahead in this rapidly evolving landscape. The shift also implies a greater reliance on specialized AI cloud providers and a need for robust strategies to manage multi-cloud or hybrid environments that incorporate these new AI-native services.