SQL got its own stage at FabCon Europe in Barcelona. SQL Server on Azure Local and DiskANN vector indexes reached general availability, with Hyperscale and Database Hub changes alongside. What each one means if you run SQL Server today.
This year the Barcelona conference changed its name to the European Microsoft Fabric + SQL Community Conference, and SQL got its own stage. It ran from 28 September to 1 October 2026. Most coverage focused on Fabric and Copilot, but the SQL announcements matter to any organisation still running SQL Server, which in the mid-market is most of them.
Here is what was announced, what is ready for production, and what each change means in practice.
The short version
- Generally available: SQL Server on Azure Local, including disconnected operations.
- Generally available: DiskANN vector indexes in Azure SQL Hyperscale, Azure SQL Managed Instance and SQL database in Fabric.
- Generally available: larger Hyperscale compute, up to 192 vCores.
- Preview: serverless auto-pause for Hyperscale.
- Preview: Database Hub in Microsoft Fabric.
- Coming soon: database agents for SQL and PostgreSQL on Azure, and higher-density Hyperscale elastic pools of up to 50 databases.
SQL Server on Azure Local
Azure Local is Microsoft's hardware for running Azure services in your own building. SQL Server on Azure Local is now generally available, so you can run SQL Server on infrastructure you own while managing it through Azure. Disconnected operations cover sites that cannot or must not connect to the cloud.
This is for organisations that need their data on site: manufacturing plants, regulated industries, or anyone with data sovereignty rules. It is not a reason to keep databases on-premises if nothing requires it. Before planning around it, check the hardware requirements, your licensing agreements and the support model, particularly for disconnected sites.
DiskANN vector indexes
Vector indexes let a database search by meaning rather than exact match, which is what most AI assistants need to find the right documents or records before answering. DiskANN is the indexing method Microsoft uses, and it is now generally available in Hyperscale, Managed Instance and SQL database in Fabric.
The practical benefit is keeping AI search next to your relational data, under the same security and backup, instead of adding a separate vector database. For many mid-market AI projects, that is one less system to run.
Hyperscale changes
Serverless auto-pause, in preview, pauses a Hyperscale database when nobody is using it and stops compute charges until it is needed again. That suits development, test and reporting databases that sit idle overnight. Larger compute options of up to 192 vCores are generally available for heavy workloads, and Microsoft announced higher-density elastic pools of up to 50 databases per pool.
Database Hub
Database Hub, now in preview after early access in March, gives a single view in Fabric of your databases across SQL Server, Azure SQL, PostgreSQL, Cosmos DB and SQL database in Fabric. Microsoft also announced database agents for SQL and PostgreSQL, due in preview soon, to help with monitoring and tuning. Like other early previews, keep it away from anything production depends on until it is generally available.
What this means if you run SQL Server today
- If your SQL Server versions are near end of support, these announcements widen your options: Azure SQL, Hyperscale, SQL Server on Azure Local, or SQL database in Fabric.
- If you want your SQL data in Fabric for reporting, you rarely need to move the database. Mirroring replicates it into OneLake without a pipeline.
- If you are planning AI search over your own data, look at vector indexes in the database you already run before buying another product.
Choosing between these is a platform decision, not a feature decision. Our data platform and warehouse work starts with what your data needs to do, then picks the place it should live. If you would like that conversation, book a free audit.
Related reading
- FabCon Europe 2026: what to adopt now, what to watch, and what to leave alone
- Mirroring in Microsoft Fabric: replicating a database without building a pipeline
- Lakehouse SQL endpoint or Fabric warehouse: which to use and when
- Microsoft Fabric implementation
Kiran Sahiba
Analytics Solutions Consultant
Part of the Hopton Analytics team, delivering governed analytics programmes for UK mid-market organisations.
