Azure Weekly Update - 11th September 2026

Azure Weekly Update (11 Sept 2026) — Key Tech Highlights

Compute / VM & Storage

• VM OS disk caching with “disk 4 caching”
  • Instead of caching only writes locally (with reads from remote storage), Azure will cache the entire OS disk image onto local storage.
  • Benefits:
    • Higher reliability (reduced dependency on remote storage for OS reads)
    • Improved performance / lower latency during VM operation
    • Faster provisioning behavior since caching happens asynchronously after VM start (VM boots using remote initially, then caching proceeds)
  • Intended workloads: AI, data, and stateless services
  • Availability: across many VM SKUs

Azure Functions (Flex Consumption)

• TLS/certificate support expanded
  • Up to three customer certificates (public + private) per application.
  • Certificate sources:
    • Import from Azure Key Vault
    • Or issue via App Service manage certificates

Storage Security / SAS (User-Delegation)

• End-to-end TLS encryption for storage is mentioned as newly available/added (context indicates prior discussion).
• Userbound delegated SAS now GA
  • Moves signing away from account keys to signing bound to Microsoft Entra identity (user/service principal/managed identity).
  • Allows restricting a user-delegated SAS to be usable by specific identities.
  • Maximum validity: 7 days

Developer Tooling / CLI & App Lifecycle

• Azure Developer CLI (azd)
  • Streamlines going from local application code → fully running Azure solution, including:
    • infrastructure provisioning
    • app packaging
    • deployment
    • CI/CD setup
• AZD extension framework
  • Lets you add custom capabilities such as tools, MCP servers, security checks, and custom commands
  • Microsoft Foundry uses this model to expose capabilities via “AZD AI”

Testing / Performance Testing

• Azure Load Testing with Playwright workspaces (app testing capability)
  • Available in:
    • Australia East
    • Japan East
    • Switzerland North
  • Goal: run Playwright tests on cloud-hosted infrastructure as part of end-to-end app testing

AI Copilot / Debugging

• Azure Copilot new “troubleshooting agent”
  • “Genetic help”-style investigation agent focused initially on:
    • Azure Compute: VMs
    • Azure Kubernetes Service (AKS)
  • Example investigations:
    • why a VM restarts unexpectedly
    • why an AKS deployment fails

GitHub Copilot Research (Preview)

• GitHub Copilot “HydraFusion” (research preview)
  • Access via /experimental in Copilot.
  • Described as model routing to balance performance, cost, and latency using three patterns:
    1. Single: choose one model
    2. Cascade: draft with one model, then a quality gate accepts/escalates to a stronger model
    3. Critique: one model drafts, another model family reviews; the drafting model can revise once
  • Overall goal: best output with efficient model usage

---

Main speaker(s) / source(s)

• Primary source: The video host/presenter (no specific name provided in the subtitles).
• Mentioned external sources/products:
  • Microsoft (Azure updates, Foundry, AZD AI)
  • GitHub Copilot (HydraFusion)
  • Azure services: VMs, Functions Flex Consumption, Storage SAS, Load Testing/Playwright, Copilot