Azure Update 10th April 2026

Key Concepts

• AKS CNI Overlay: A networking configuration for Azure Kubernetes Service that separates pod IP space from node IP space.
• MCP (Model Context Protocol): A standard for AI applications to interact with external tools and resources (knowledge).
• Network Security Perimeter (NSP): A feature allowing multiple PaaS services to be grouped within a single security boundary.
• PG Bouncer: A connection pooler for PostgreSQL that improves scalability by managing idle and short-lived connections.
• Azure Foundry Local: A lightweight runtime for executing optimized AI models directly on local devices using the ONNX runtime.

Azure Kubernetes Service (AKS) Updates

• CNI Overlay Expansion: The ability to expand the pod CIDR (Classless Inter-Domain Routing) range is now Generally Available (GA). Users can expand an existing range (e.g., from /18 to /16) for IPv4 on Linux nodes. Note: Shrinking or adding non-contiguous ranges is not supported.
• HTTP Proxy Disablement: Users can now disable the HTTP proxy for existing clusters. This action triggers a node re-image; therefore, the use of disruption budgets is recommended to maintain workload availability.
• Observability Improvements: AKS now integrates data from Azure Monitor workspaces (powered by Prometheus) into namespace and workload views, providing deeper insights into resource utilization and troubleshooting trends.
• GPU Support: Azure Red Hat OpenShift clusters now support VM SKUs powered by Nvidia H100 and H200 GPUs, enabling high-performance computing and AI-driven container workloads.

Azure Functions & AI Integration

• MCP Resource Triggers: Azure Functions now support MCP resource triggers in GA. This allows functions to act as MCP servers that provide not only "tools" (actions) but also "resources" (static or dynamic knowledge) to Large Language Models (LLMs).

Networking & Security

• Network Watcher Rule Impact Analysis: A new preview feature that allows administrators to simulate the impact of security admin rule changes before they are applied, preventing potential misconfigurations.
• Service Bus in NSP: Azure Service Bus can now be included in a Network Security Perimeter, allowing it to securely communicate with other services within the same perimeter (e.g., accessing Azure Key Vault for customer-managed keys).

Storage & Database

• Azure Migrate: Now supports assessments for Azure Files (SMB and NFS) in preview. It evaluates on-premises file shares and recommends the appropriate Azure Files SKU based on resiliency, performance, and regional requirements.
• PostgreSQL Maintenance: Maintenance notifications are now consolidated by region rather than by individual server, reducing notification bloat.
• PG Bouncer 1.25.1: This version is now GA, offering improved performance, stability, security, and protocol handling for connection pooling.

AI Models & Foundry

• New AI Models: Microsoft released three new models in preview:
  • Speech Transcription: Supports 25 languages with 50% reduced GPU requirements.
  • Speech Generation: High-fidelity model capable of generating 60 seconds of speech in 1 second.
  • Maya Image 2: A high-capability text-to-image model.
• GLUE 4.2: A new general-purpose LLM added to the Azure Foundry catalog, specifically optimized for reasoning-intensive problem solving.
• Foundry Local: Now GA, this 20MB runtime enables local inferencing on devices using hardware acceleration (GPUs/NPUs) via the ONNX runtime. It utilizes a curated catalog of models optimized for local execution, supporting a hybrid cloud-to-edge architecture.

Synthesis

The updates for April 10th reflect a strong focus on hybrid AI deployment and operational efficiency. By enabling local inferencing via Foundry Local and expanding MCP capabilities in Azure Functions, Microsoft is lowering the barrier for integrating AI into edge and cloud applications. Simultaneously, infrastructure management is being streamlined through consolidated database notifications, improved AKS observability, and proactive network security analysis, allowing for more robust and scalable cloud environments.