Import Models to DirectLake Semantic Models

Our organization is a small startup currently running 10 Power BI semantic models, each built on top of on-premise source tables.

We recently migrated those source tables to a Fabric Lakehouse and want to simplify our architecture by:

• Creating one master semantic model on top of the Lakehouse

• Repointing the 10 existing reports to this new semantic model

Key Constraints

• We cannot afford significant downtime, since these reports are actively used.

• We want to avoid rebuilding the reports if possible.

• We want to avoid creating additional Lakehouses because we already have a single centralized Lakehouse.

Current Challenge

We explored the Semantic Link Labs migration script, but it requires:

1. Creating a Power Query template

2. Building a Dataflow Gen2

3. Loading the data into a new Lakehouse per model

This would effectively result in 10 additional Lakehouses, which contradicts our goal of maintaining a single centralized Lakehouse architecture.

What We Are Looking For

We are trying to determine the best migration approach that allows us to:

• Move from 10 existing semantic models → 1 master semantic model

• Repoint existing reports to the new model

• Minimize downtime

• Avoid duplicating Lakehouses

• Keep the architecture clean and maintainable

Questions

1. Is there a recommended approach in Microsoft Fabric / Power BI to consolidate multiple semantic models into a single master semantic model backed by a Lakehouse?

2. Can we rebind or repoint reports to the new semantic model without rebuilding them?

3. Is there a way to reuse the existing Lakehouse tables directly without using Dataflow Gen2 or creating additional Lakehouses?

4. Are there best practices or tools (e.g., Tabular Editor, XMLA endpoints, dataset rebinding) that would support this migration with minimal disruption?