Period: last 7 days · Items: 12 · Source: Azure official updates RSS + Azure Blog
Today’s issue really reads as two clear themes. One is that Retirement notices that are easy to quietly put off are now accumulating in earnest. The other is that data and developer features you can apply directly in real-world work—such as SQL, PostgreSQL, and agentic AI—are reaching GA. In particular, for items that directly affect assets already in operation, such as Batch, Storage, Synapse Link, and VPN Client, you’ll feel much more comfortable if you start building your inventory now.
5 items
What you get: HPC and batch workload operations teams can identify in advance which Batch pools are medium- to long-term replacement candidates. That enables lower-risk VM series transition planning than a sudden migration.
What to do now: If you use Azure Batch, extract your current VM SKU inventory by pool and first classify whether Av2/F/Fs/Fsv2/G/Gs/Lsv2 are in use.
Why now: Although the deadline is in 2028, Batch workloads require preparation time because performance validation, quota review, and cost comparison all need to be considered together. Early PoCs are especially important for operational stability in research, HPC, and large-scale batch environments.
Source: https://azure.microsoft.com/updates?id=564774
What you get: Batch operations teams can prioritize replacement targets that are on the earlier timeline of May 2028. It also creates an opportunity to standardize Batch pools through cost and performance reassessment by VM series.
What to do now: Identify jobs and pools using D/Ds/Dv2/Dsv2/Ls families in Azure Batch pools first, and set a schedule for alternative SKU review and performance testing.
Why now: Even among Batch retirements, this notice has the earlier deadline. For operations teams, this is no longer a matter of “replace it someday,” but of putting a phased transition plan in place now.
Source: https://azure.microsoft.com/updates?id=564772
What you get: Storage customers can decide whether to keep carrying older account types or standardize on modern account types. In particular, this provides a strong rationale to redefine new deployment standards around GPv2.
What to do now: Storage operations teams should build an inventory of GPv1 and legacy blob accounts across all subscriptions, and check whether new deployment templates (ARM/Bicep/Terraform) still create those account types.
Why now: Storage standards tend to persist for a long time once established. Cleaning up the new-account creation path early helps prevent technical debt from growing further.
Source: https://azure.microsoft.com/updates?id=564441
What you get: Network operations teams can reduce uncertainty such as, “Do we need to change all our single-VM NAT rules too?” By narrowing the impact scope around VMSS, you can identify actual migration targets more precisely.
What to do now: If you use VMSS, create an inventory of whether Inbound NAT Pools are used in your Azure Load Balancer configuration and review replacement configurations.
Why now: More than the retirement itself, this is an update in which the scope of impact has changed, so if you rely on earlier assumptions, you may either overreact or miss something. It is time to realign operations design documents and IaC templates with the current scope.
Source: https://azure.microsoft.com/updates?id=565482
What you get: Data platform teams can clarify the direction of their Cosmos DB analytics integration architecture sooner. New services can consider alternatives instead of Synapse Link, while existing environments can secure a transition roadmap before 2029.
What to do now: If you use Cosmos DB NoSQL with Synapse Link, first assess the impact on newly planned accounts and begin evaluating migration feasibility to Fabric Link for Azure Cosmos DB.
Why now: The date when new account restrictions begin—March 31, 2026—is relatively close. Data integration structures take time to change, so it is important to make decisions now by separating new and existing workloads.
Source: https://azure.microsoft.com/updates?id=558560
1 item
What you get: Network operations teams and Linux user organizations can identify their current dependency on the Preview client early and review alternatives without operational disruption. This is also a good opportunity to clean up unofficial usage patterns from a security standards perspective.
What to do now: If you have a significant Linux user base, investigate whether Azure VPN Client for Linux (Preview) is in use and add alternative access method review to your network operations standard work items.
Why now: Preview features are easy to adopt with a “let’s just use it for now” mindset, but when they retire, they can create operational gaps. Because this area touches both user endpoints and access policy, advance preparation is important.
Source: https://azure.microsoft.com/updates?id=565393
4 items
What you get: Data and AI teams gain an option for designing agentic applications while retaining control over production data. The practical value is that you can add agent architecture to existing SQL-based assets instead of replacing them.
What to do now: If your team operates SQL, PostgreSQL, or Cosmos DB, choose one internal PoC based on SQL MCP Server and start by evaluating agent integration in a read-only scenario.
Why now: Agent adoption is accelerating, but the real bottleneck is controlling data access. GA means this is no longer just for experimentation—it is time to evaluate it as an operational pattern.
Source: https://azure.microsoft.com/updates?id=564734
CREATE LOGIN ... FROM EXTERNAL PROVIDER in the virtual master database.What you get: DBAs and security teams can design authentication and authorization models in Azure SQL Database more consistently around Entra. You can modernize account management standards while reducing the gap with existing SQL operational practices.
What to do now: Azure SQL Database operations teams should review Entra-based login policies and validate CREATE LOGIN ... FROM EXTERNAL PROVIDER scenarios in a test database.
Why now: In enterprise environments, standardizing database authentication ties directly to security audits and operational automation. Now that this is GA, it is a good time to move from local or individual accounts toward Entra-centric management.
Source: https://azure.microsoft.com/updates?id=565154
What you get: DBAs and data platform teams gain a new tool that can improve productivity in routine diagnostics, tuning, and operations tasks. It can reduce repetitive analysis work and lower the barrier to entry for operational knowledge.
What to do now: DBA teams using SSMS should enable Copilot Agent mode in a non-production environment and run a PoC starting with limited tasks such as performance analysis or query review.
Why now: Database operations is an area with significant room for productivity gains, but incorrect use also carries risks of false positives or overcorrection. If you establish team standards and guardrails during Preview, it will be easier to scale later.
Source: https://azure.microsoft.com/updates?id=562637
What you get: Development teams can get started faster without having to search through scattered PostgreSQL on Azure resources. It is also a useful starting point for creating internal standard architectures or training content.
What to do now: If your team is preparing a new PostgreSQL-based app or AI-integrated feature, review the PostgreSQL Hub’s samples and accelerators and identify candidates for internal standard templates.
Why now: PostgreSQL is increasingly becoming a default choice for many apps on Azure. This GA hub reduces learning costs and can directly accelerate early-stage design for new projects.
Source: https://azure.microsoft.com/updates?id=562084
2 items
What you get: IT leaders and architects gain insight into how to realign AI roadmaps around “business connectivity and measurable outcomes” rather than simply “number of PoCs.” It helps prompt a fresh look at the operating model before technology selection.
What to do now: AI leadership teams should inventory their current AI initiatives and schedule a review meeting to assess whether each one is actually connected to real business systems, data, and KPIs.
Why now: As the number of AI projects grows, individual pilots tend to multiply while actual business impact becomes less clear. This is the moment to revisit operating and governance models alongside the technology itself.
Source: https://azure.microsoft.com/en-us/blog/3-things-leaders-need-to-know-from-microsoft-build-2026/
What you get: AI platform teams can understand the market and technology trend of expanding model options available within the Microsoft ecosystem. This is useful input for multi-model strategy and agent quality comparison.
What to do now: Teams evaluating Foundry or Copilot expansion should review the official documentation, available regions, and security requirements for Claude Fable 5, and add it to their model comparison criteria.
Why now: In generative AI, the key question is becoming less about the model itself and more about how it is operated on which platform. Changes in model options quickly become changes in architecture options.
Source: https://azure.microsoft.com/en-us/blog/claude-fable-5-is-now-available-in-microsoft-foundry-powering-the-next-era-of-autonomous-agents/
CREATE LOGIN ... FROM EXTERNAL PROVIDER