Can Dynamics 365 Finance data be queried in Fabric using metadata mirroring?

Description / Problem Statement

Hello everyone,

I’m currently working on an analytics architecture involving Dynamics 365 Finance & Operations (F&O), Dataverse, and Microsoft Fabric, and I’d like to clarify whether a specific integration scenario is currently supported or on the roadmap.

Context

• Dynamics 365 F&O exposes its data through Dataverse, where F&O entities typically appear as virtual tables.
• When using Link to Microsoft Fabric (Fabric Link), only physical Dataverse tables with change tracking enabled are supported.
• As a result, to make F&O data available in Fabric today, it is usually necessary to:
  • Materialize F&O entities into physical Dataverse tables
  • Enable change tracking
  • Accept additional Dataverse storage consumption, since Dataverse generates Delta/Parquet replicas

This approach works, but it introduces:

• Extra Dataverse storage costs
• Data duplication (operational data being materialized purely for analytics)
• Additional operational complexity

What I’m trying to understand

Microsoft Fabric supports different mirroring approaches, including metadata mirroring (shortcut-based access), where:

• Data remains in the source system
• No physical copy is written into OneLake
• Fabric queries data through metadata/shortcuts only

Given this, my questions are:

1. Is it currently possible (or planned) to connect Dynamics 365 F&O tables to Fabric using a metadata-only mirroring approach, similar to how shortcuts work for other platforms, without materializing data into Dataverse storage or OneLake?

2. Has Fabric (or Dataverse) considered supporting F&O virtual tables directly, so they could be queried from Fabric without requiring physical Dataverse tables and Delta replication?

3. Some community discussions suggest that Dataverse-based replication might no longer be the optimal long-term approach for large F&O analytical workloads.

   • Is there any official guidance or roadmap indicating a shift toward a different integration pattern (e.g., direct metadata mirroring or another Fabric-native approach)?