Azure SQL Managed Instance is depicted in a secure cloud setup spanning two availability zones, with performance and resource controls.
Azure SQL Managed Instance picked up three Next-gen General Purpose improvements in mid-November 2026. Microsoft's Azure Updates feed lists them as generally available. They cover transaction-log throughput, storage sizing and memory sizing. Each one targets a different bottleneck, so each is worth checking separately against your own instances.

Azure SQL Managed Instance is depicted in a secure cloud setup spanning two availability zones, with performance and resource controls. What Microsoft announced​

The mid-November item for Azure SQL Managed Instance lists three changes:

  • Higher log throughput. Next-gen General Purpose gets log rates of up to 220 MiB/s and higher per-vCore throughput. Microsoft says this helps transaction-intensive workloads.
  • Better storage-to-vCore ratios and finer-grained storage scaling. The stated aim is less need to overprovision vCores just to get enough storage for storage-heavy workloads.
  • Flexible memory for zone-redundant Next-gen General Purpose. Memory can be right-sized independently of vCores, so you don't have to buy compute you don't need.

Microsoft's own wording on the log change is "up to 220 MiB/s". That is a ceiling, not a promise for every instance.

Log throughput: the headline number​

An Azure SQL tech community post on the Next-gen General Purpose update gives the before and after:

  • On Standard-series hardware, maximum log throughput rises from 120 MiB/s to 220 MiB/s, an improvement of more than 80%.
  • On Premium-series hardware, the maximum rises from 192 MiB/s to 220 MiB/s.
  • The 40-vCore Premium-series configuration now delivers up to 220 MiB/s, compared with 192 MiB/s before.

The post also says the largest configurations gain the most. Microsoft's resource-limits page, last updated 2026-11-17, now says log write throughput has a separate instance-level limit of 6 MiB/s per vCore, up to 220 MiB/s per instance, across all Next-gen General Purpose hardware configurations. That same page states that increasing IOPS doesn't increase this log write throughput limit.

The earlier published limits were different. The previous Next-gen General Purpose documentation listed 4.5 MiB/s per vCore, with a 120 MiB/s maximum on Standard-series and 192 MiB/s on Premium-series. The new per-vCore rate of 6 MiB/s is therefore also an increase.

Who benefits: write-heavy OLTP, bulk loads and ETL that were hitting the instance log-rate ceiling. If your workload is read-heavy, or is limited by data IOPS or CPU, expect little change. A bigger log ceiling won't fix a different bottleneck.

Storage and vCore sizing​

Microsoft hasn't published the new storage-to-vCore ratios or the scaling increments in the material I reviewed. I won't guess at them. The stated rationale is that storage-hungry databases shouldn't force you to buy more vCores than you need.

Some existing limits still apply:

  • Microsoft's documentation says 32 TB is the maximum only for the Next-gen General Purpose tier, versus 16 TB for classic General Purpose.
  • Storage must be specified in multiples of 32 GB.
  • You pay for the maximum storage size you configure.

Next-gen General Purpose was already designed so that storage and IOPS can be resized independently of compute, typically finishing within five minutes. The November change refines the ratios and increments on top of that. It doesn't introduce independent storage scaling.

Flexible memory for zone-redundant instances​

Flexible memory lets you change the memory allocated to an instance without changing the number of vCores. It first reached general availability for Next-gen General Purpose in November 2025, alongside a memory slider with up to 49 memory configurations per instance.

Zone redundancy is the newer piece. The Azure Updates feed lists a separate item, "Zone-redundant Next-gen GP tier in Azure SQL Managed Instance", as Launched. The memory item in this update extends the right-sizing to those zone-redundant instances.

Some Microsoft documentation pages still describe this as preview. The Next-gen General Purpose usage page and the local and zone redundancy page both say that zone redundancy for Next-gen General Purpose is in preview, along with flexible memory for zone-redundant instances on Premium-series hardware. The resource-limits page, updated 2026-11-17, instead says flexible memory is generally available on Premium-series hardware for Next-gen General Purpose and Business Critical instances, with either local or zone redundancy.

The documentation is evidently catching up with the launch, so treat the Azure portal and the newest resource-limits page as the authority. That same page notes that zone-redundant Next-gen General Purpose instances require Azure Elastic SAN zone-redundant storage, which isn't available in some multi-zone regions. Check regional availability before you plan a zone-redundant deployment.

Some practical details from the resource-limits page:

  • You can change memory for new and existing instances, using the Azure portal or the REST API.
  • The change applies to all databases in the instance and performs a failover as the final step. Schedule it with that in mind.
  • You pay for extra memory by the GB per hour. Billable memory is total memory minus default memory.
  • In the portal, use the Compute + storage pane. In the REST API, use the properties.memorySizeInGB value.

What to check on your instances​

  1. Log write throughput. Compare your instance's log write rate against the ceiling for its vCore count (6 MiB/s per vCore, capped at 220 MiB/s). If you are regularly near the old limit, the new one may help.
  2. Storage versus compute. If you added vCores only to unlock storage, review the new ratios and scaling steps in the current documentation before resizing.
  3. Memory on zone-redundant instances. If you chose a larger vCore count mainly for RAM, compare that against a flexible-memory configuration. Weigh the per-GB memory charge against the cost of extra vCores.
  4. Benchmark before committing. Test a representative write-heavy workload. The new ceiling is a maximum, and real gains depend on where your bottleneck actually is.

Caveats​

  • The announcement doesn't give new ratios, scaling increments, memory ranges, per-region availability or any pricing change. Check the portal and documentation for those.
  • It describes no migration or restart procedure for the log and storage changes, so I won't invent one.
  • Pricing for Next-gen General Purpose was originally described as the same as the current General Purpose tier for compute and storage. That is background from the 2024 introduction, not a statement about this November update.

Bottom line​

The log-throughput change is the most concrete one. Standard-series instances gain the most, from 120 to 220 MiB/s, while Premium-series moves from 192 to 220 MiB/s. The storage and memory changes are about not paying for resources you don't use. If you run Next-gen General Purpose, revisit your sizing and retest your heaviest write workloads, and use the portal rather than the older documentation to confirm what applies in your region.

 

References

  1. (Launched) Generally Available: Azure SQL Managed Instance updates for mid-November 2026 Azure Updates 2026-11-17T20:00:32Z
  2. Resource Limits - Azure SQL Managed Instance | Microsoft Learn learn.microsoft.com
  3. Azure SQL Managed Instance gets a boost: New capabilities in Next-gen General Purpose techcommunity.microsoft.com