What was announced
On November 17, 2026, Microsoft announced a limited preview of cross-database querying for Azure SQL Database Hyperscale. The Azure Updates entry is tagged "In preview," but the Azure SQL engineering blog says this is a limited preview. You request access, and the preview team confirms eligibility and enables the feature once your configuration is approved. Treat it as a gated test capability, not something every Azure customer can switch on today.
The timing lines up with Ignite. Microsoft's Ignite database roundup describes cross-database querying for Hyperscale, now in limited preview, as letting developers use familiar T-SQL three-part names to query across databases in the same elastic pool from one connection, while each database enforces its own permissions. A separate SQL dev blog frames the value for migrations. It says cross-database queries help customers unblock SQL Server migrations by reducing the query rewrites and application redesign needed for applications with cross-database dependencies.
How it works
You connect to one database and refer to objects in other participating databases as database.schema.object. Microsoft's example connects to SalesDB and joins the local dbo.Orders table to InventoryDB.dbo.Inventory on product ID. The application gets one result set. The database prefix names the remote object and does not switch your connection's current database.
Microsoft stresses three points:
- It is native three-part-name access, not external-table access.
- It needs no extra network setup.
- Normal SQL connectivity and access controls still apply.
You can use the cross-database references directly or inside views, stored procedures and functions. You can also combine cross-database reads with data modifications in T-SQL workflows.
The hard boundaries
This is where the fine print matters.
- Hyperscale elastic pools only. All participating databases must be in the same pool on the same logical server. Microsoft's announcement doesn't extend the preview to standalone Hyperscale databases, other service tiers, other pools or other servers.
- Databases must stay put. They must remain online and in the pool. Moving one out of the pool breaks any query that uses it.
- Writes are limited to one database. Write support is still under development in the preview. The public post doesn't spell out the supported write patterns, which Microsoft shares during onboarding.
- No savepoints or nested transactions.
Microsoft's write example shows the shape of the limit. Connected to InventoryDB, it reads an order from SalesDB.dbo.Orders and calls a local procedure that updates InventoryDB's own inventory table, all inside one transaction. The read crosses databases and the only write is local. It is not a multi-database write, so don't read it as distributed transaction support.
Getting into the preview
Microsoft lays out the path like this:
- On a non-production Azure SQL logical server, create a Hyperscale elastic pool with two or more test databases.
- Request access through the sign-up form. The team will confirm eligibility, enable the feature, and send an onboarding guide with setup steps and the full list of preview limitations.
- Create a Microsoft Entra login, map users in each participating database, and grant the permissions your workload needs across them.
- Connect to any database in the pool and query the others using three-part names.
The identity step is easy to underestimate. Cross-database access isn't automatic. Each database still enforces its own permissions, so a query can fail simply because a user mapping is missing in one participant.
How it differs from elastic query
Azure SQL Database already has a cross-database option, elastic query, which Microsoft Learn also labels as preview. Microsoft's documentation describes elastic query as read-only querying of remote databases, and an option for SQL Server customers migrating applications that use three- and four-part names or linked servers. It is built on external data sources and external tables.
The new feature is a different mechanism. It uses native three-part names, avoids external tables, and has its own eligibility rules and constraints. It also allows a limited form of writing, which elastic query does not. Elastic query is supported on all service tiers, while the new capability is Hyperscale elastic pools only. The new feature is not a drop-in replacement for every existing option.
The change also lands against a long-standing caveat. A July 2026 Redmondmag column on Hyperscale listed no cross-database queries using three- or four-part naming as one of its two big compatibility exceptions, the other being CLR functions. This preview addresses the first of those, within the limits above. It does not close the CLR gap, and Microsoft has not claimed it brings full SQL Server parity.
Who should test it
Good candidates:
- Teams migrating SQL Server applications whose queries use three-part names across databases.
- Applications that fetch from several databases and join the results in code.
- Teams that keep synchronized copies of data only so cross-database operations are possible.
Microsoft says removing those workarounds may reduce application compute, duplicate storage and integration-pipeline costs. Treat that as a possible benefit, not a promise. The announcement gives no benchmarks, performance thresholds or quantified savings, and it gives no general availability date or pricing change.
Practical advice
- Test in non-production, as Microsoft requires, and use realistic query shapes. Join behavior and data volume across databases haven't been publicly characterized.
- Audit your code for cross-database writes. If a workflow modifies more than one database in a transaction, expect to refactor it.
- Check for savepoints and nested transactions in stored procedures, since both are unsupported.
- Plan around pool membership. Any operational process that moves databases between pools could break queries.
- Get the onboarding guide before designing anything. It holds the full limitation list, which the public post doesn't.
Bottom line
Hyperscale cross-database querying removes one of the more annoying reasons to rewrite an application before moving it to Azure SQL. It remains a gated preview limited to Hyperscale elastic pools on a single logical server, with single-database writes and no savepoints or nested transactions. It is a worthwhile test for migration teams and not yet a basis for production plans.
References
- (In preview) Public Preview: Azure SQL DB Hyperscale Cross Database Querying Azure Updates · 2026-11-17T22:35:29Z
- Why Azure SQL Database Hyperscale Is Not Just for Massive Workloads -- Redmondmag.com redmondmag.com
- Azure SQL Hyperscale Cross-database querying: now in limited preview | Microsoft Community Hub techcommunity.microsoft.com