Warehouse

A T-SQL data warehouse in Fabric, with full read and write, storing its tables in OneLake.

The Fabric Warehouse is a SQL data warehouse for teams who think in T-SQL. It stores its tables as Delta in OneLake, like a Lakehouse, but you load and change them with SQL.

Part of the Store layer of the ecosystem. Reviewed 13 November 2026 by Piyush Jain.

What Warehouse is

You create tables, load them and transform them with T-SQL: INSERT, UPDATE, MERGE, stored procedures and transactions that span several tables. There are no servers to size; compute comes from your Fabric capacity.

Because the tables are Delta in OneLake, a Lakehouse, a notebook or a Direct Lake semantic model can read them without a copy. You can also query Lakehouse tables from the Warehouse in the same workspace.

When to use it

Use it when

  • Your data team already works in SQL Server, Synapse or another SQL warehouse.
  • Loads need transactions across several tables, so a failed step leaves nothing half-written.
  • You are moving an existing SQL warehouse into Fabric and want the code to carry over.

Think twice when

  • Most of your data arrives as files, or your team works in Python and Spark. Use a Lakehouse.
  • You expect it to be a drop-in copy of SQL Server. Some T-SQL features are not supported, so check your code first.

The mistakes I see most

From opening other people's models. Each one is fixable.

  1. 1

    Choosing by habit, not by who loads the data

    The choice between Warehouse and Lakehouse matters less than people think, because both end in Delta tables. What matters is who will maintain the loads. Pick the one your team can support at 2am.

  2. 2

    Lifting and shifting every stored procedure

    Old warehouses carry years of procedures nobody owns. Moving them all costs capacity and keeps the mess. Move what the reports use, and retire the rest.

  3. 3

    Pointing reports at raw tables

    A semantic model built straight on wide staging tables is slow and fragile. Build a star schema, facts and dimensions, as the reporting layer, and point the model at that.

Questions people ask

Is a Fabric Warehouse the same as Azure Synapse dedicated SQL pools?
No. The Fabric Warehouse has no dedicated pool to size or pause; it uses your Fabric capacity and stores tables as open Delta files in OneLake. Most T-SQL carries over, but not all of it, so test before migrating.
Can Power BI use Direct Lake on a Warehouse?
Yes. A semantic model can read Warehouse tables in Direct Lake mode, the same way it reads Lakehouse tables.
Should I use a Lakehouse or a Warehouse?
If your team writes SQL and needs transactions, use a Warehouse. If it writes Spark or lands files, use a Lakehouse. Many setups use both: a Lakehouse for raw data and a Warehouse for the reporting layer.

Sources: Microsoft Learn: What is data warehousing in Microsoft Fabric?; Microsoft Learn: Star schema guidance. Opinions are my own.