Upgrade Azure Local operating system to new version
Today, we’re sharing more details about the end of support for Azure Local, with OS version 25398.xxxx (23H2) on October 31, 2025. After this date, monthly security and quality updates stop, and Microsoft Support remains available only for upgrade assistance. Your billing continues, and your systems keep working, including registration and repair. There are several options to upgrade to Azure Local, with OS version 26100.xxxx (24H2) depending on which scenario applies to you. Scenario #1: You are on Azure Local solution, with OS version 25398.xxxx If you're already running the Azure Local solution, with OS version 25398.xxxx, there is no action required. You will automatically receive the upgrade to OS version 26100.xxxx via a solution update to 2509. Azure Local, version 23H2 and 24H2 release information - Azure Local | Microsoft Learn for the latest version of the diagram. If you are interested in upgrading to OS version 26100.xxxx before the 2509 release, there will be an opt-in process available in the future with production support. Scenario #2: You are on Azure Stack HCI and haven’t performed the solution upgrade yet Scenario #2a: You are still on Azure Stack HCI, version 22H2 With the 2505 release, a direct upgrade path from version 22H2 OS (20349.xxxx) to 24H2 OS (26100.xxxx) has been made available. To ensure a validated, consistent experience, we have reduced the process to using the downloadable media and PowerShell to install the upgrade. If you’re running Azure Stack HCI, version 22H2 OS, we recommend taking this direct upgrade path to the version 24H2 OS. Skipping the upgrade to the version 23H2 OS will be one less upgrade hop and will help reduce reboots and maintenance planning prior to the solution upgrade. After then, perform post-OS upgrade tasks and validate the solution upgrade readiness. Consult with your hardware vendor to determine if version 24H2 OS is supported before performing the direct upgrade path. The solution upgrade for systems on the 24H2 OS is not yet supported but will be available soon. Scenario #2b: You are on Azure Stack HCI, version 23H2 OS If you performed the upgrade from Azure Stack HCI, version 22H2 OS to version 23H2 OS (25398.xxxx), but haven’t applied the solution upgrade, then we recommend that you perform post-OS upgrade tasks, validate the solution upgrade readiness, and apply the solution upgrade. Diagram of Upgrade Paths Conclusion We invite you to identify which scenarios apply to you and take action to upgrade your systems. On behalf of the Azure Local team, we thank you for your continuous trust and feedback! Learn more To learn more, refer to the upgrade documentation. For known issues and remediation guidance, see the Azure Local Supportability GitHub repository.
Announcing the preview of Software Defined Networking (SDN) on Azure Local
Big news for Azure Local customers! Starting in Azure Local version 2506, we’re excited to announce the Public Preview of Software Defined Networking (SDN) on Azure Local using the Azure Arc resource bridge. This release introduces cloud-native networking capabilities for access control at the network layer, utilizing Network Security Groups (NSGs) on Azure Local. Key highlights in this release are: 1- Centralized network management: Manage Logical networks, network interfaces, and NSGs through the Azure control plane – whether your preference is the Azure Portal, Azure Command-Line Interface (CLI), or Azure Resource Manager templates. 2- Fine-grained traffic control: Safeguard your edge workloads with policy-driven access controls by applying inbound and outbound allow/deny rules on NSGs, just as you would in Azure. 3- Seamless hybrid consistency: Reduce operational friction and accelerate your IT staff’s ramp-up on advanced networking skills by using the same familiar tools and constructs across both Azure public cloud and Azure Local. Software Defined Networking (SDN) forms the backbone of delivering Azure-style networking on-premises. Whether you’re securing enterprise applications or extending cloud-scale agility to your on-premises infrastructure, Azure Local, combined with SDN enabled by Azure Arc, offers a unified and scalable solution. Try this feature today and let us know how it transforms your networking operations! What’s New in this Preview? Here’s what you can do today with SDN enabled by Azure Arc: ✅ Run SDN Network Controller as a Failover Cluster service — no VMs required! ✅ Deploy logical networks — use VLAN-backed networks in your datacenter that integrate with SDN enabled by Azure Arc. ✅ Attach VM Network Interfaces — assign static or DHCP IPs to VMs from logical networks. ✅ Apply NSGs - create, attach, and manage NSGs directly from Azure on your logical networks (VLANs in your datacenter) and/or on the VM network interface. This enables a generic rule set for VLANs, with a crisper rule set for individual Azure Local VM network interface using a complete 5-tuple control: source and destination IP, port, and protocol. ✅ Use Default Network Policies — apply baseline security policies during VM creation for your primary NIC. Select well-known inbound ports such as HTTP (while we block everything else for you), while still allowing outbound traffic. Or select an existing NSG you already have! SDN enabled by Azure Arc (Preview) vs. SDN managed by on-premises tools Choosing Your Path: Some SDN features like virtual networks (vNETs), Load Balancers (SLBs), and Gateways are not yet supported in the SDN enabled by Azure Arc (Preview). But good news: you’ve still got options. If your workloads need those features today, you can leverage SDN managed by on-premises tools: - SDN Express (PowerShell) - Windows Admin Center (WAC) The SDN managed by on-premises tools continues to provide full-stack SDN capabilities, including SLBs, Gateways, and VNET peering, while we actively work on bringing this additional value to complete SDN enabled by Azure Arc feature set. You must choose one of the modes of SDN management and cannot run in a hybrid management mode, mixing the two. Please read this important consideration section before getting started! Thank You to Our Community This milestone was only possible because of your input, your use cases, and your edge innovation. We're beyond excited to see what you build next with SDN enabled by Azure Arc. To try it out, head to the Azure Local documentation Let’s keep pushing the edge forward. Together!
EOL of Azure Linux 2.0 on Azure Kubernetes Service enabled by Azure Arc
Azure Linux 2.0 will reach its End of Life (EOL) in July 2025 Azure Linux 2.0 (formerly CBL-Mariner) will reach its official End of Life (EOL) on July 31, 2025. After this date, it will no longer receive updates, security patches, or support from the Azure Linux team. Starting with the Azure Local 2507 release, Azure Kubernetes Service enabled by Azure Arc will ship Azure Linux 3.0 images for all supported Kubernetes versions. This change applies to all AKS enabled by Azure Arc deployments, as we have used Azure Linux 2.0 as the base image in the past. To maintain security compliance and ensure continued support, all AKS Arc customers must plan on migrating to Azure Linux 3.0 at the earliest by upgrading their Azure Local instances to the 2507 release, when it is available. What's new in Azure Linux 3.0 Approximately every three years Azure Linux releases a new version of its operating system with upgrades to major components. Azure Linux 3.0 offers increased package availability and versions, an updated kernel, and improvements to performance, security, and tooling and developer experience. Some of the major components upgraded from Azure Linux 2.0 to 3.0 include: Component Azure Linux 3.0 Azure Linux 2.0 Release Notes Linux Kernel v6.6 (Latest LTS) V5.15 (Previous LTS) Linux 6.6 Containerd v1.7.13, but will also offer v2.0 once it becomes stable 1.6.26 Containerd Releases SystemD v255 V250 Systemd Releases OpenSSL v3.3.0 V1.1.1k OpenSSL 3.3 For more details on the key features and updates in Azure Linux 3.0 see the 3.0 GitHub release notes. Upgrading to Azure Linux 3.0 Once the Azure Local 2507 release is available, update to 2507 . Once your Azure Local instance has upgraded, you can then upgrade your Kubernetes clusters You can choose to the remain on the same Kubernetes version and provide the same version number in the aksarc upgarde command. Once the upgrade is completed, you should be able to check the kernel version on your Linux nodes. Kernel version v6.6 is the latest Azure Linux 3.0 version. Sample command kubectl --kubeconfig /path/to/aks-cluster-kubeconfig get nodes -o wide Sample output NAME STATUS ROLES AGE VERSION INTERNAL-IP EXTERNAL-IP OS-IMAGE KERNEL-VERSION CONTAINER-RUNTIME moc-lsbe393il9d Ready control-plane 3h14m 100.72.248.133 100.72.248.133 <none> CBL-Mariner/Linux 6.6.92.2 Containerd://1.6.26 moc-lzwagtkjah5 Ready control-plane 3h12m 00.72.248.134 v1.30.4 <none> CBL-Mariner/Linux 6.6.92.2 Containerd://1.6.26 FAQs Is Azure Linux same as Mariner? Yes, Mariner was rebranded to Azure Linux. We will slowly update our documentation and VM/container image tags to reflect this name change When did Azure Linux 3.0 GA? Azure Linux 3.0 became generally available in August 2024. When will Azure Linux 3.0 reach End of Life (EOL)? We currently support each major version for 3 years after it becomes generally available. Azure Linux 3.0 will reach EOL in Summer 2027. How to keep in touch with the AKS Arc team For updates, feedback, and feature requests related to AKS Arc: Ask questions & submit feedback via AKS Arc GitHub Issues Partners with support questions can reach out to aks-hci-talk@microsoft.comCloud infrastructure for disconnected environments enabled by Azure Arc
Organizations in highly regulated industries such as government, defense, financial services, healthcare, and energy often operate under strict security and compliance requirements and across distributed locations, some with limited or no connectivity to public cloud. Leveraging advanced capabilities, including AI, in the face of this complexity can be time-consuming and resource intensive. Azure Local, enabled by Azure Arc, offers simplicity. Azure Local’s distributed infrastructure extends cloud services and security across distributed locations, including customer-owned on-premises environments. Through Azure Arc, customers benefit from a single management experience and full operational control that is consistent from cloud to edge. Available in preview to pre-qualified customers, Azure Local with disconnected operations extends these capabilities even further – enabling organizations to deploy, manage, and operate cloud-native infrastructure and services in completely disconnected or air-gapped networks. What is disconnected operations? Disconnected operations is an add-on capability of Azure Local, delivered as a virtual appliance, that enables the deployment and lifecycle management of your Azure Local infrastructure and Arc-enabled services, without any dependency on a continuous cloud connection. Key Benefits Consistent Azure Experience: You can operate your disconnected environment using the same tools you already know - Azure Portal, Azure CLI and ARM Templates extended through a local control plane. Built-in Azure Services: Through Azure Arc, you can deploy, update, and manage Azure services such as Azure Local VMs, Azure Kubernetes Service (AKS), etc. Data Residency and Control: You can govern and keep data within your organization's physical and legal jurisdiction to meet data residency, operational autonomy, and technological isolation requirements. Key Use Cases Azure Local with disconnected operations unlocks a range of impactful use cases for regulated industries: Government and Defense: Running sensitive government workloads and classified data more securely in air-gapped and tactical environments with familiar Azure management and operations. Manufacturing: Deploying and managing mission-critical applications like industrial process automation and control systems for real-time optimizations in more highly secure environments with zero connectivity. Financial Services: Enhanced protection of sensitive financial data with real time data analytics and decision making, while ensuring compliance with strict regulations in isolated networks. Healthcare: Running critical workloads with a need for real-time processing, storing and managing sensitive patient data with the increased levels of privacy and security in disconnected environments Energy: Operating critical infrastructure in isolated environments, such as electrical production and distribution facilities, oil rigs, or remote pipelines. Here is an example of how disconnected operations for Azure Local can provide mission critical emergency response and recovery efforts by providing essential services when critical infrastructure and networks are unavailable. Core Features and capabilities Simplified Deployment and Management Download and deploy the disconnected operations virtual appliance on Azure Local Premier Solutions through a streamlined user interface. Create and manage Azure Local instances using the local control plane, with the same tooling experience as Azure. Offline Updates The monthly update package includes all the essential components: the appliance, Azure Local software, AKS, and Arc-enabled service agents. You can update and manage the entire Azure Local instance using the local control plane without an internet connection. Monitoring Integration You can monitor your Azure Local instances and VMs using external monitoring solutions like SCOM by installing custom management packs and monitor AKS Clusters through 3 rd party open-source solutions like Prometheus and Grafana. Run Mission-Critical Workloads – Anytime, Anywhere Azure Local VMs You can run VMs with flexible sizing, support for custom VM images, and high availability through storage replication and automatic failover – all managed through the local Azure interface. AI & Containers with AKS You can use disconnected AI containers with Azure Kubernetes Service (AKS) on Azure Local to deploy and manage AI applications in disconnected scenarios where data residency and operational autonomy is required. AKS enables the deployment and management of containerized applications such as AI agents and models, deep learning frameworks, and related tools, which can be leveraged for inferencing, fine-tuning, and training in isolated networks. AKS also automates resource scaling, allowing for the dynamic addition and removal of container instances to more efficiently utilize hardware resources, including GPUs, which are critical for AI workloads. This provides consistent Azure experience in managing Kubernetes clusters and AI workloads with the same tooling and processes in connected environments. Get Started: Resources and Next Steps Microsoft is excited to announce the upcoming preview of Disconnected Operations for Azure Local in Q3 ‘CY25 for both Commercial and Government Cloud customers. To Learn more, please visit Disconnected operations for Azure Local overview (preview) - Azure Local Ready to participate? Get Qualified! or contact your Microsoft account team. Please also check out this session at Microsoft Build https://build.microsoft.com/en-US/sessions/BRK195 by Mark Russinovich, one of the most influential minds in cloud computing. His insights into the latest Azure innovations, the future of cloud architecture and computing, is a must-watch event!
Deploy a Kubernetes Application Programmatically Using Terraform and CLI
In our previous blog post, we explored Kubernetes apps benefits along with an introduction into how to programmatically deploy Kubernetes Apps. Today we will cover deploying a Kubernetes application programmatically using Terraform and CLI. These deployment methods can streamline your workflow and automate repetitive tasks. Deploying your Kubernetes Application using Terraform This walkthrough assumes you have previous knowledge of Terraform. For additional information and guidance on using Terraform to provision a cluster, please refer here. Prerequisites Before we begin, ensure you have the following: Terraform Azure CLI Sample Location You can find the Terraform sample we will be using at this location: Terraform Sample Prepare the Environment First, initialize Terraform in the current directory where you have copied the k8s-extension-install sample by running the following command: terraform init In the directory, you will find two example tfvars files. These files can be used to deploy the application with different configurations: azure-vote-without-config.tfvars - Deploy the application with the default configuration for azure-vote. azure-vote-with-config.tfvars - Deploy/update the application with a custom configuration for azure-vote. Before you test run the sample tfvars files, update the following in the tfvars files: cluster_name - The name of the AKS cluster. resource_group_name - The name of the resource group where the AKS cluster is located. subscription_id - The subscription ID where the AKS cluster is located. Deploy the Application To deploy the application with the default configuration for azure-vote, run: terraform apply -var-file="azure-vote-without-config.tfvars" To deploy or update the application with a custom configuration for azure-vote, use: terraform apply -var-file="azure-vote-with-config.tfvars" Conclusion And that's it! You've successfully deployed your Kubernetes application programmatically using Terraform. This process can drastically reduce the time and effort involved in managing and scaling your applications. By using Terraform, you can ensure that your deployment is consistent and repeatable, making it easier to maintain your infrastructure as code. Deploying a Kubernetes Application from Azure CLI Deploying a Kubernetes application using Azure CLI can seem daunting, but we’re here to make it simple and accessible. Follow these steps, and you’ll have your azure-vote application up and running in no time! Prerequisites Before we get started, ensure you have the following: Azure CLI installed on your machine Deploying the Sample Azure-Vote Application from the Marketplace Step 1: Log in to Azure Open your terminal and log in to your Azure account by running: az login Step 2: Set Your Subscription Specify the subscription you want to use with: az account set --subscription Step 3: Deploy the Azure-Vote Application Now, deploy the azure-vote application to your Kubernetes cluster with the following command: az k8s-extension create --name azure-vote --scope cluster ` --cluster-name <clusterName> --resource-group <resourceGroupName> --cluster-type managedClusters ` --extension-type commercialMarketplaceServices.AzureVote ` --plan-name azure-vote-paid ` --plan-product azure-vote-final-1 ` --plan-publisher microsoft_commercial_marketplace_services ` --configuration-settings title=VoteAnimal value1=Cats value2=Dogs Updating Configuration Settings If you want to update the configuration settings of the azure-vote application, you can do so easily. Use the following command to change the configuration settings: az k8s-extension update --name azure-vote ` --cluster-name <clusterName> --resource-group <resourceGroupName> --cluster-type managedClusters ` --configuration-settings value1=Elephant value2=Horse And there you have it! By following these steps, you can deploy and update the azure-vote application on your Kubernetes cluster using Azure CLI. Conclusion Deploying Kubernetes applications using Azure CLI is a powerful way to manage and scale your applications. The process described above helps ensure your deployments are consistent and repeatable, simplifying the maintenance of your infrastructure as code.😄Introducing Azure Local: cloud infrastructure for distributed locations enabled by Azure Arc
Today at Microsoft Ignite 2024 we're introducing Azure Local, cloud-connected infrastructure that can be deployed at your physical locations and under your operational control. With Azure Local, you can run the foundational Azure compute, networking, storage, and application services locally on hardware from your preferred vendor, providing flexibility to meet your requirements and budget.Announcing general availability of workload orchestration: simplifying edge deployments at scale
We’re excited to announce the General Availability of workload orchestration, a new Azure Arc capability that simplifies how enterprises deploy and manage Kubernetes-based applications across distributed edge environments. Organizations across industries, such as manufacturing, retail, healthcare, face challenges in managing varied site-specific configurations. Traditional methods often require duplicating app variants—an error-prone, costly, and hard-to-scale approach. Workload orchestration solves this with a centralized, template-driven model: define configurations once, deploy them across all sites, and allow local teams to adjust within guardrails. This ensures consistency, improves speed, reduces errors, and scales with your CI/CD workflows—whether you’re supporting 200+ factories, offline retail clusters, or regionally-compliant hospital apps. Fig 1.0: Workload orchestration – Key features Key benefits of workload orchestration include: Solution Configuration & Template Reuse Define solutions, environments, and multiple hierarchy levels using reusable templates. Key-value stores and schema-driven inputs allow flexible configurations, validations with role-based access to maintain control. Context-Aware Deployments Automatically generate deployable artifacts based on selected environments (Dev, QA, Prod) and push changes safely through a git ops flow — enabling controlled rollouts and staged testing across multiple environments. Deploying at Scale in Constrained Environments Deploy workloads across edge and cloud environments with built-in dependency management and preloading of container images (a.k.a Staging) to minimize downtime during narrow maintenance windows. Bulk Deployment and Git Ops-Based Rollouts Execute large-scale deployments — including shared or dependent applications — across multiple sites using Git-based CI/CD pipelines that validate configurations and enforce policy compliance before rollout. End to End Observability K8 diagnostics in workload orchestration provide full-stack observability by capturing container logs, Kubernetes events, system logs, and deployment errors—integrated with Azure Monitor and Open Telemetry pipelines for proactive troubleshooting across edge and cloud environments. Who Is It For? Workload orchestration supports two primary user personas: IT Admins and DevOps Engineers: Responsible for initial setup and application configuration via CLI. OT Operators: Use the portal for day-to-day activities like monitoring deployments and adjusting configurations. Resources for You to Get Started You can start using workload orchestration by visiting the Azure Arc portal and following the documentation. We encourage you to try it with a small application deployed to a few edge sites. Create a template, define parameters like site name or configuration toggles, and run a deployment. As you grow more comfortable, expand to more sites or complex applications.Announcing the Public Preview of the Azure Arc gateway!
The wait is over, we are thrilled to introduce the Public Preview of the Azure Arc gateway for Arc-enabled Servers, and Arc-enabled Kubernetes! They reduce the number of required endpoints for customers to configure their Enterprise proxy when setting up for using Azure Arc services. How Does it Work? Arc gateway introduces two new components: Arc gateway – An Azure Resource with a single, unique endpoint that will handle the incoming traffic to Azure from on-prem Arc workloads. This endpoint is to be configured in customer’s enterprise proxies. Azure Arc Proxy – A component of the Arc connected machine agent that routes all Agent and extension traffic to its destination in Azure via an Arc gateway Resource. The Arc Proxy is installed on every Arc-enabled Resource within the core Arc agent. Arc gateway on Arc-enabled Servers Architecture Arc gateway on Arc-enabled Kubernetes Architecture How do I Deploy Arc gateway? At a high level, there are three steps: create an Arc gateway Resource. Get the Arc gateway URL, and configure your Enterprise proxy Either onboard your Servers/K8s clusters using the gateway resource info or update the existing Arc Server/K8s resource with the created gateway resource info. For Arc enabled Servers, you can find Arc gateway details & instructions in the Public Preview documentation, and the Arc gateway for Arc-enabled Servers Jumpstart Episode. For Arc-enabled Kubernetes, more details are available in the Public Preview Documentation. Arc gateway Endpoint Coverage, Illustrated by the Azure Monitoring Scenario For the Arc gateway public preview, we have focused on covering primarily Service Endpoints for Azure control plane traffic. Most of the data plane endpoints are not yet covered by Arc gateway. I’d like to use the Azure monitoring on Arc-enabled Servers scenario to illustrate the Endpoints covered by the Public Preview release. Below is a comparison of the list of endpoints customers must open access to in their enterprise proxy with and without Arc gateway for this common scenario. As displayed, Arc gateway cuts the list of required endpoints nearly in half and removes the need for customers to allow wildcard endpoints in their on-prem environment. Endpoints required without Arc gateway (17) Endpoints required with Arc gateway (8) Arc-enabled Servers Endpoints aka.ms download.microsoft.com packages.microsoft.com login.microsoftonline.com *.login.microsoftonline.com pas.windows.net management.azure.com *.his.arc.azure.com *.guestconfiguration.azure.com azgn*.servicebus.windows.net *.blob.core.windows.net dc.services.visualstudio.com Azure Monitor Endpoints global.handler.control.monitor.azure.com <virtual-machine-region-name>.handler.control.monitor.azure.com <log-analytics-workspace-id>.ods.opinsights.azure.com <virtual-machine-region-name>.monitoring.azure.com <data-collection-endpoint>.<virtual-machine-region-name>.ingest.monitor.azure.com Arc-enabled Servers Endpoints <URL Prefix>.gw.arc.azure.com management.azure.com login.microsoftonline.com gbl.his.arc.azure.com <region>.his.arc.azure.com packages.microsoft.com Azure Monitor Endpoints <log-analytics-workspace-id>.ods.opinsights.azure.com <data-collection-endpoint>.<virtual-machine-region-name>.ingest.monitor.azure.com We're continuing to expand the endpoint coverage and further reduce the number of endpoints required to be configured through customers' Enterprise proxies. I’d like to invite you to try out the Arc gateway Public Preview release and share any questions, comments or feedback and requests to the Public Preview Contact Form.Preview of Arc enabled SQL Server in US Government Virginia
Introduction We are excited to announce that Azure Arc-enabled SQL Server on Windows is now in public preview for the US Government Virginia region. With Azure Arc-enabled SQL Server, U.S. government agencies and organizations can manage SQL Server instances outside of Azure from the Azure Government portal, in a secure and compliant manner. Arc-enabled SQL Server resources in US Gov Virginia can be onboarded and viewed in the Azure Government portal just like any Azure resource, giving you a single pane of glass to monitor and organize your SQL Server estate in the Gov cloud. Preview features of Azure Arc-Enabled SQL Server Currently, in the US Government Virginia region, SQL Server registration provides the following features: Connect (onboard) a SQL Server instance to Azure Arc. SQL Server inventory which includes the following capabilities in the Azure portal: View the SQL Server instance as an Azure resource. View databases as an Azure resource. View the properties for each server. For example, you can view the version, edition, and database for each instance. All other features, including Extended Security Updates (ESU), are not currently available. How to Onboard Your SQL Server Onboarding a SQL Server to Azure Arc in the Government cloud is a two-step process that you can initiate from the Azure (US Gov) portal. Step 1: Connect hybrid machines with Azure Arc-enabled servers Step 2: Connect your SQL Server to Azure Arc on a server already enabled by Azure Arc Limitations The following SQL Server features are not currently available in any US Government region: Failover cluster instance (FCI) Availability group (AG) SQL Server services like SSIS, SSRS, or Power BI Report Server Future Plans and Roadmap This public preview is a major first step in bringing Azure Arc’s hybrid data management to Azure Government, and more enhancements are on the way. We will be enabling features like Arc-based billing (PAYG) and ESU purchasing along with feature parity with public cloud in future. After US Gov Virginia, we will expand to other US Gov regions starting with US Gov Arizona. Conclusion The availability of Azure Arc-enabled SQL Server in the US Gov Virginia region marks an important milestone for hybrid data management in Government. If you’re an Azure Government user managing SQL Server instances, we invite you to try out this public preview. And please, share your feedback with us through the community forum or your Microsoft representatives. Learn More: SQL Server enabled by Azure Arc in US Government Preview SQL Server enabled by Azure ArcEmpowering the Physical World with AI
Unlocking AI at the Edge with Azure Arc The integration of AI into the physical environment is revolutionizing ways we interact with and navigate the world around us. By embedding intelligence into edge devices, AI is not just processing data—it is defining how machines perceive, reason, and act autonomously in real-world scenarios. AI at the edge is transforming how we interact with our environment, driven by critical factors such as data sensitivity, local regulations, compliance, low latency requirements, limited network connectivity, and cost considerations. Added to this, the emergence of new, powerful agentic AI capabilities enables autonomous and adaptive real-time operations, making AI an indispensable tool in reshaping the physical world. Customers’ Use Cases By embedding AI into edge operations, industries are unlocking transformative efficiencies and innovations. In manufacturing, edge-powered AI enables real-time quality control and predictive maintenance, minimizing downtime and maximizing productivity. In retail, AI enhances customer experiences with personalized recommendations and streamlined inventory management. Similarly, finance leverages AI's capabilities for robust fraud detection and advanced risk management. Moreover, sectors like government and defense are increasingly adopting edge AI for safety-critical applications, enabling autonomous, real-time surveillance and response solutions that are both efficient and resilient. These advancements are paving the way for scalable, adaptive solutions that meet the unique demands of diverse operational environments. Azure’s Adaptive Cloud Approach enabling AI from cloud to edge Building on the promise to unify cloud and edge, Azure’s adaptive cloud approach is empowering teams to develop and scale AI workloads seamlessly across diverse environments. By enabling a unified suite of services tailored for modern AI applications, whether deployed in public clouds or distributed locations, Azure Arc enables streamlined operations with enhanced security and resilience. Central to extending AI services to the edge is our commitment to adaptive, scalable, and efficient solutions tailored to diverse operational needs. Azure Arc plays a key role in this vision by facilitating seamless deployment and management of AI workloads across various environments. This week, we’re excited to share that a subset of Microsoft Azure AI Foundry models, such as Phi and Mistral have been rigorously validated to run on Azure Local enabled by Azure Arc. Our investments are reflected in two primary areas: Foundational tools for MLOps and developer frameworks, which empower teams to build robust AI applications Intuitive, end-to-end low-code experiences designed for data analysts and solution developers. These low-code tools prioritize user-friendly interfaces and rapid deployment, enabling the creation of solutions with just a few clicks. This dual focus ensures enterprises can fully harness the potential of edge AI while maintaining flexibility and operational efficiency. Image 1: This high-level diagram illustrates our vision for the cloud to edge AI workloads, enabled by Azure Arc. Some components (agents and integration with AI Foundry and Foundry Local) are still under development, while others are more advanced and have been released to the market. Build 2025: New Capabilities and Releases This strategic vision is now being realized through a wave of new capabilities unveiled at Build 2025. These innovations are designed to accelerate edge AI adoption and simplify the developer experience—making it easier than ever to build, deploy, and manage intelligent applications across hybrid environments. Announcements related to developer Building blocks: Kubernetes AI Toolchain Orchestrator (KAITO), enabled by Azure Arc (public preview) Foundry Local (public preview) for Windows apps to be deployed on any client device read more here. Workload orchestration (public preview) Application development tools for Kubernetes enabled by Arc (public preview) Refer to this blog to read more: https://aka.ms/AdaptiveCloudBuild2025 Announcements related to End-to-end experiences: Edge RAG, enabled by Azure Arc is now available in public preview. Azure AI Video Indexer for recorded files, enabled by Arc is generally available since April 2025. Azure AI Video Indexer for live video analysis, enabled by Arc is available in private preview, for limited set of customers Customer scenarios: enabling search and retrieval for on-premises data on Azure Local Edge RAG targets customers who have data that needs to stay on premises due to data gravity, security and compliance, or latency requirements. We have observed significant and consistent interest from highly regulated sectors. These entities are exploring the use of RAG capabilities in disconnected environments through Azure Local. DataON is a hybrid cloud computing company for enterprises of all sizes, with a focus on educational institutions and local government agencies. Recently, they have worked with the their customers to successfully deploy our RAG solution on CPU and GPU clusters and begin testing with sample end-customer data. “DataON has been actively exploring how Edge RAG can enhance our Microsoft Azure Local solutions by providing more efficient data retrieval and decision-making capabilities. It’s exciting to be part of the private preview program and see firsthand how Edge RAG is shaping the future of data-driven insights.” Howard Lo | VP, Sales & Marketing | DataON This capability brings generative AI and RAG to on-premises data. Edge RAG was validated on AKS running on Azure Local. Based on DataON and other customer feedback, we have expanded the version to include new features: Model Updates: Ability to use any model compatible with OpenAI Inferencing standard APIs Multi-lingual support: 100+ common languages for document ingestion and question-answer sessions Multi-modal support: Support for image ingestion & retrieval during question-answer sessions Search Types: Support for Text, Vector, Hybrid Text & Hybrid Text+Image searches Ingestion Scale-out: Integration with KEDA for fully parallelized, high-throughput ingestion pipeline Evaluation Workflow with RAG Metrics: Integrated workflow with built-in or customer-provided sample dataset Read more about Edge RAG in this blog: https://aka.ms/AzureEdgeAISearchenabledbyArc. AI Workloads for Disconnected Operations In fully disconnected (air-gapped or non-internet) environments, such as those often found in government and defense sectors, technologies like RAG, can be deployed on-premises or in secure private clouds. This is currently available with limited access. Use Cases: Video analysis: Automatically analyzes video and audio content to extract metadata such as objects and scenes. Use cases include live video and analysis, mission debriefing and training, and modern safety. Models consumption: A central repository for securely managing, sharing, and deploying AI/ML models. Use cases: model governance, rapid deployment of mission-specific models, and inter-agency collaboration. Retrieval-Augmented Generation (RAG): Combines LLMs with a document retrieval system to generate accurate, context-aware responses based on internal knowledge bases. Use cases include field briefings, legal and policy compliance, and cybersecurity incident response. Transforming Industries with AI: Real-World Stories from the Edge Across industries, organizations are embracing AI to solve complex challenges, enhance operations, and deliver better outcomes. From healthcare to manufacturing, retail to energy, and even national security, Azure AI solutions are powering innovation at scale. In the manufacturing sector, a global company sought to optimize production and reduce costly downtime. Azure AI Video Indexer monitored video feeds from production lines to catch defects early, while custom predictive maintenance models from the Model Catalog helped prevent equipment failures. RAG provided real-time insights into operations, empowering managers to make smarter decisions by asking questions. These tools collectively boosted efficiency, minimized downtime, and improved product quality. At Airports, Azure AI helped enhance passenger experience and safety. From monitoring queue lengths and tracking vehicles to detecting falls and identifying restricted area breaches, the combination of Azure Local, Video Indexer, Azure IoT for Operations, and custom AI created a smarter, safer airport environment. Retailers, too, are reaping the benefits. A major retail chain used Azure AI to understand in-store customer behavior through video analytics, optimize inventory with demand forecasting models, and personalize shopping experiences using RAG. These innovations led to better customer engagement, streamlined inventory management, and increased sales. In Healthcare, a leading provider operating multiple hospitals and clinics nationwide faced the daunting task of analyzing massive volumes of patient data—from medical records and imaging to real-time feeds from wearable devices. With strict privacy regulations in play, they turned to Azure AI. Using Azure AI Video Indexer, they analyzed imaging data like X-rays and MRIs to detect anomalies. The Model Catalog enabled predictive analytics to identify high-risk patients and forecast readmissions. Meanwhile, Retrieval-Augmented Generation (RAG) gave doctors instant access to patient histories and relevant medical literature. The result? More accurate diagnoses, better patient care, and full regulatory compliance. These stories highlight how Azure Arc enabled AI workloads are not just a set of tools—they are a catalyst for transformation. Whether it’s saving lives, improving safety, or driving business growth, the impact is real, measurable, and growing every day. Learn More Whether you are tuning in online or joining us in person, we wish you a fun and exciting Build 2025! The advancements in AI at the edge are set to revolutionize how we build, deploy, and manage applications, providing greater speed, agility, and security for businesses around the world. Recommended Build Sessions: Breakout session (BRK188): Power your AI apps across cloud and edge with Azure Arc Breakout session (BRK183): Improving App Health with Health Modeling and Chaos Engineering Breakout session (BRK 195): Inside Azure innovations with Mark Russinovich Breakout session (BRK 168): AI and Agent Observability in Azure AI Foundry and Azure Monitor
