Announcing new compute SKUs, faster storage, immutable backup and fleet management capabilities for Azure Database for PostgreSQL flexible server
The Postgres on Azure team is announcing a new set of capabilities for Azure Database for PostgreSQL flexible server at Ignite 2026. Our priorities are guided by what matters most to enterprises running PostgreSQL in production: more power for demanding workloads, higher resiliency for critical data, and better value at every Compute tier.
This week we're announcing new compute SKUs (including our first Arm-based option), faster storage, immutable backups, more flexible disaster recovery options, and new fleet management features that make it easier to run Postgres at scale. Here's a quick tour of what's new.
Feature Highlights
- New Compute SKUs
- Enhanced Storage capabilities
- Resiliency & recovery
- Engine & extensions
- Networking
- Fleet management
New Compute SKUs
Each new generation of processor brings new performance characteristics which, depending on your workload, deliver improved price-performance. This week we’re announcing support for new compute SKUs that offer more power and better value at every level, from burstable v2 for your dev/test workloads and lower cost “diskless” SKUs to Azure’s latest high-powered Arm CPUs.
Burstable v2
Burstable v2 Compute SKUs are based on AMD's third Generation EPYC™ 7763v processor, bringing better performance for Azure Database for PostgreSQL dev/test workloads with more powerful CPUs, higher network bandwidth and storage throughput.
In internal testing, Basv2 delivered up to 2.8x higher throughput and 64% lower average latency than v1. This means work will be completed faster at the same busting utilization percentage, providing better value as well as performance compared to v1.
Find out more: Basv2 series, Compute options in flexible server.
Cobalt 200 Arm-based SKU
Azure Database for PostgreSQL now supports the Azure Cobalt 200 VM SKU, using Microsoft's custom Arm-based processor built around the Arm Neoverse V3 Compute Subsystems. Cobalt 200 is built for cloud-native workloads and gives you a new price-performance option for Postgres. Initial testing has shown outstanding performance compared to similar x86-64 SKUs, particularly for in-memory analytical workloads.
Find out more: Azure Cobalt 200 VMs Now Available.
Intel/AMD v5 & v6 “diskless” SKUs
Compared to Intel & AMD v4 Compute SKUs, v5 and v6 deliver more transactions and use less CPU to do it. If you’re running your PostgreSQL workload on a v3 or v4 SKU, you will improve performance with better value, especially for CPU-constrained workloads, by upgrading your Compute SKU to v5 or later (see our recent blog post for some examples: Better Price Performance: Azure Postgres V3 & V5 Compute Compared).
This week, we’re announcing new Intel & AMD v5 and v6 diskless SKUs, which are provided without a temporary disk at a lower price point.
Traditional Compute SKUs include a local temp disk. The temp disk is used for temporary files created when sorts, joins and index builds exceed available memory.
How does this affect PostgreSQL workloads? Temp disk SKUs are still recommended for larger analytical workloads that generate temporary files, but if you’re running an OLTP workload, or an analytical workload that fits in memory, the new diskless SKUs perform as well as their temp disk counterparts at a lower cost. What happens if temporary files need to be generated for larger workloads? The "diskless" SKUs will still work well, with temp files written to attached storage and sharing IO with data operations.
Our initial testing with 10GB and 30GB analytical workloads for example showed each CPU generation performance better, with the lower cost diskless SKUs performance as well that the SKUs with temp disks.
Find out more: Azure VM sizes with no local temporary disk
Confidential compute v6 SKU
Azure Database for PostgreSQL now supports the latest generation of Confidential Compute SKUs, delivering enhanced security and high performance for sensitive workloads. Powered by Intel TDX and AMD SEV-SNP, confidential computing helps protect data while it’s actively being processed within a hardware-backed trusted execution environment, complementing protection for data at rest and in transit.
Intel-based SKUs are initially available in 5 regions and AMD-based SKUs in 23 regions, with availability continuing to expand.
Learn more: Azure confidential computing in Azure Database for PostgreSQL flexible server
Coming soon: Intel/AMD v7 SKUs
Another set of SKUs to look forward are the v7 SKUs, based on Intel’s Xeon® 6 6973PC (Granite Rapids) and AMD's 5th Generation EPYC™ 9005 processors. Scale up to 372 vCPUs and 2.8 TiB memory the most demanding analytic workloads with very large dataset. We’ll be making these SKUs available for Azure Database for PostgreSQL in the next few weeks.
Enhanced Storage capabilities
This week, we’re dramatically increasing the performance of flexible server storage options, as well as improving the usability of SSDv2.
Ultra Disk
Very large high-traffic transactional systems, high-ingest logging and telemetry platforms and other extremely latency-sensitive applications can benefit from storage that supports high IOPS. For the most demanding I/O-intensive workloads, Azure Ultra Disk is now generally available for Azure Database for PostgreSQL. Ultra Disk offers up to 400K IOPS with 10 GB/s throughput and consistently low latency (with support for up to 64TB storage).
Learn more: Ultra Disk storage option in Azure Database for PostgreSQL flexible server
SSDv2 auto-growth
As data size grows DBAs need to monitor storage usage to avoid disk-full scenarios. With storage auto-grow, additional disk storage is automatically added when available storage drops below pre-defined thresholds, reducing the risk of running out. This feature is now generally available for high-performance SSDv2 storage.
Learn more: Configure storage autogrow in Azure Database for PostgreSQL flexible server
Resiliency & recovery
Immutable point-in-time restore (PITR) backups - Preview
Ransomware and accidental deletion are top concerns for data teams and often form a part of regulatory compliance requirements for resiliency. With PITR Immutable backups, now in preview, Azure Database for PostgreSQL helps protect your point-in-time recovery backups from being modified or deleted during their retention period by isolating them from the server's lifecycle. Even if the server is deleted, protected recovery points remain available throughout their configured retention period, providing an additional layer of protection for disaster recovery and resiliency.
Learn more: Azure Database for PostgreSQL backup
Long-term daily backups of any size – Preview
The latest version of Azure Backup for PostgreSQL (v2) brings physical snapshot-based backups to long-term retention (LTR), supporting backups up to 64 TB (up from 1 TB) and daily backups (up from weekly). With this preview, you can retain any size backups for up to 10 years, providing greater scale and flexibility for long-term data protection and compliance needs.
Learn more: Long-term retention
Chaos Studio zone-down scenario
You only know how effective your disaster recovery strategy is when something breaks, and if there’s a problem, you don’t want to find out during an outage.
Azure Chaos Studio allows you to test disaster scenarios such as “zone-down” by simulating an outage – breaking things in a controlled fashion. The Azure Chaos Studio zone-down scenario for Azure Database for PostgreSQL is now generally available. Test workload resilience by triggering forced failover on healthy, high-availability-enabled servers and assessing application recovery.
Learn more: Chaos Studio documentation.
Cross-subscription read replicas
You can now create read replicas in a different Azure subscription from the primary server. This provides more flexibility in disaster recovery architecture design, workload isolation and organizational boundaries, such as keeping DR environments in a separate subscription.
Engine & extensions
PostgreSQL 19 on day zero, with major version upgrade
Azure Database for PostgreSQL flexible server supports PostgreSQL 19 on the same day as the community release. You can try the new release on a managed service without waiting. During Preview, PostgreSQL 19 is available in East Asia.
PostgreSQL 19 engine enhancements available in preview on Azure Database for PostgreSQL flexible server
Existing servers get a clear path forward. With major version upgrade (MVU), you can move eligible flexible server instances to PostgreSQL 19 in place. There's no dump-and-restore and no rebuilding your environment.
1st day support also covers the extensions you use most. Key extensions, including pgvector, azure_ai, DiskANN, PostGIS, and pg_duckdb, are available alongside PostgreSQL 19. You get the new engine and the extension ecosystem together, not one at a time.
Learn more: PostgreSQL 19 on Azure
Citus CDC support in Azure Database for PostgreSQL elastic clusters - Preview
Scenarios like real-time analytics and event sourcing require live changes to flow from databases to dashboards and downstream processing. This can be challenging for distributed tables.
Change Data Capture (CDC) enables real-time streaming of data updates to downstream systems using PostgreSQL's logical decoding framework. Citus CDC enables this streaming when data is sharded across multiple nodes by intercepting shard-level WAL changes and mapping them back to the original distributed table name, while normalizing shard names. This feature is now available in preview with elastic clusters in Azure Database for PostgreSQL flexible server.
Learn more: Citus Change Data Capture (CDC)
Networking
VNet to private endpoint network reconfiguration
Production workloads need a secure networking environment, and Azure Database for PostgreSQL offers private network configuration options based on VNet integration where the database server runs in a virtual network you control, and Private Link where your database server can be reached by a private endpoint in a designated subnet. Previously it was difficult to switch from VNet integration to Private Link, which was challenging if for example you needed to change your secure networking architecture to meet new security requirements. You can now migrate directly existing VNet-integrated servers to private endpoint networking with the general availability of VNet to private endpoint network reconfiguration.
Learn more: Change from virtual network integration to a private endpoint-capable network configuration
Fleet management
Database Hub: one place for your whole database estate
Database Hub is generally available. It gives you a single, tenant-wide view of your databases across Azure PostgreSQL, Azure SQL, and Azure Cosmos DB. You can see what you have, how healthy it is, and where it's exposed.
- Inventory: See every database across engines and subscriptions in one place.
- Health: Spot problems across your fleet without checking each server separately.
- Security posture: Find gaps early, such as auditing that isn't enabled or customer-managed key (CMK) warnings.
With GA, Database Hub is ready for your database server estate. Whether you run a handful of Postgres servers or thousands of instances across multiple engines, you can manage your entire estate of reacting server by server.
Try now: Database Hub
Wrapping up
From ARM-based compute and Ultra Disk to immutable backups and fleet-level management, this week’s announcements are about giving you more power, more resilience and better value from Postgres on Azure.
For more PostgreSQL and MySQL and announcements this week, including general availability of Azure Horizon DB and the latest AI innovations check out these blog posts:
Decisions, decisions: Your guide to PostgreSQL at Microsoft Ignite 2026
HorizonDB on Azure and Fabric
Postgres Built for AI: Introducing the Latest AI Innovations in Azure HorizonDB
Complete Oracle to PostgreSQL Migration: Move Your Schema, Data, and Application Code
If you're at Ignite, do drop by to meet our team at the PostgreSQL & MySQL on Azure station at the Hub. I hope to see you there!