The short answer
Microsoft Fabric Warehouse represents the next generation of cloud data warehousing on Azure, modernizing the architecture established by Synapse Dedicated SQL Pool.
Synapse Dedicated Pool Microsoft Fabric Warehouse
---------------------------------------------------------------------
Proprietary Columnstore on disk Open Delta Lake Parquet in OneLake
Fixed 60-distribution hash sharding Serverless compute, no 60-slice split
DWU provisioning (DW100c-30000c) Fabric Capacity Units (CUs)
Manual data copies between engines Zero-copy shortcuts across OneLakeThe OneLake foundation
In Synapse Dedicated Pool, data is stored in a proprietary format across 60 fixed distributions. If a data science team wanted to read those tables using Apache Spark or Databricks, the data had to be exported or extracted via connectors.
In Microsoft Fabric, all warehouse tables are stored natively in OneLake as open-source Delta Lake Parquet files. When a SQL query writes to a Fabric Warehouse, it writes standard Delta Lake tables with ACID transaction logs.
Apache Spark notebooks, Power BI Direct Lake reports, and external engines can read those exact same files instantly without data duplication.
Elimination of distribution tuning
Synapse Dedicated Pool requires data engineers to carefully select HASH, ROUND_ROBIN, or REPLICATED distributions to prevent across 60 slices.
Fabric Warehouse completely decouples compute from storage using an autonomous serverless query engine. The system dynamically scales compute slots, compiles vectorized machine code, and optimizes join paths automatically without requiring manual distribution keys.
Fabric enables Direct Lake mode in Power BI, allowing dashboards to query millions of rows from OneLake storage directly with sub-second response times without importing data into memory.