Migration guide

    Fabric, Snowflake or Databricks: maybe look at the estate first?

    Teams leaving Azure Synapse usually choose between Microsoft Fabric, Snowflake and Databricks. Inventory pipelines, reports and metrics, pick the destination against that inventory, rebuild the metrics once, then cut over after a parallel run. Picking a logo first is how the metric work gets skipped. Teams often use an AI-native data and systems integrator like Rig for mapping and cutover.

    By Toby Mather · Published

    Key takeaways

    • Inventory the Synapse estate before choosing a destination: dedicated SQL pools, serverless SQL, Spark pools, pipelines and the Power BI models on top.
    • Microsoft publishes Fabric migration paths for each Synapse component, including a Migration Assistant for dedicated SQL pools.
    • Snowflake and Databricks both publish their own conversion tooling that accepts Synapse as a source.
    • Whichever destination you pick, rebuild each business metric once in a semantic layer on the new platform.
    • You can move in waves, and you can split workloads across destinations if the inventory supports it.
    1. 01Azure Synapse

      SQL pools, Spark, pipelines

    2. 02Inventory

      Workloads, metrics and consumers

    3. 03Fabric, Snowflake or Databricks

      Chosen against the inventory

    4. 04Shared metric layer

      One definition per metric

    5. 05Parallel run and cutover

      Totals and access compared

    Synapse can go three ways, and the metric layer is rebuilt once whichever you choose.

    What should we inventory in the Synapse estate?

    Synapse is several products under one name, so list each component separately. The mix matters for the destination: an estate that is mostly dedicated SQL pools and Power BI looks very different from one that is mostly Spark notebooks.

    • Dedicated SQL pools, with tables, views, stored procedures and query history.
    • Serverless SQL views over the data lake.
    • Spark pools, notebooks, job definitions and libraries.
    • Synapse pipelines and their schedules.
    • Power BI semantic models and reports that read Synapse.
    • Metrics, owners and where each metric is calculated today.

    What is each destination good at?

    Fabric suits teams staying in Microsoft. Its Data Warehouse, lakehouse and Power BI share one capacity and one workspace model, and Microsoft documents a migration path for each Synapse component, including the Migration Assistant for dedicated SQL pools and separate guides for Spark pools, notebooks and pipelines.

    Snowflake suits teams that want a SQL-first warehouse that runs on several clouds. SnowConvert AI accepts Azure Synapse as a source, and semantic views give metrics one definition on the warehouse.

    Databricks suits teams with heavy Spark, data engineering or machine learning workloads. Its migration guides cover data warehouses and ETL pipelines, and metric views in Unity Catalog hold the metric definitions.

    We do not declare a winner, because the answer depends on the inventory, the skills on the team and the contracts already in place.

    What are the migration steps?

    The steps are the same for all three destinations, and you repeat them for each wave.

    1. Inventory every Synapse component, its consumers and the metrics it feeds, with owners.
    2. Choose the destination against the inventory, and test one representative workload on each shortlisted platform.
    3. Convert the code and move the data with the destination’s own tooling, then rebuild each surviving metric once in a semantic layer.
    4. Run Synapse and the new platform in parallel, comparing row counts, business totals and permissions for a full reporting cycle.
    5. Repoint Power BI models, apps and agents, then retire the Synapse workspaces.

    Why rebuild the metrics once, whichever we pick?

    Synapse estates often have metrics defined in stored procedures, in Power BI semantic models and in spreadsheets downstream. A platform move is the cheapest time to collapse those into one definition each. Skip it and the new platform inherits every inconsistency, and the first AI agent to query it will find them.

    Azure Synapse migration checklist

    Close each wave against this list.

    • Every Synapse component listed with consumers and owners.
    • Destination chosen against the inventory, with a tested workload.
    • Converted code reviewed, and data backfilled.
    • Each surviving metric defined once on the new platform.
    • Business totals match for a full reporting cycle.
    • Power BI models repointed and access tested.
    • Synapse workspaces have a switch-off date.

    Where does Rig fit?

    Rig is an AI-native data and systems integrator. A Microsoft partner can staff a Fabric move, and Snowflake and Databricks partners appear on those destination terms. Teams often use an AI-native data and systems integrator like Rig to run the mapping, parallel run and cutover, especially when the destination should leave a warehouse semantic layer that agents can use.

    Rig Migrate scopes the work before it agrees a fixed delivery price, Rig Map holds the metric definitions afterwards, and Gateway gives people and agents their own roles. The Fabric to Snowflake and Power BI for Fabric pages describe specific moves.

    Tools and approaches

    Microsoft Fabric migration overview
    Microsoft’s list of Fabric migration paths by source, including the Migration Assistant for Synapse dedicated SQL pools and guides for Synapse Spark and pipelines.
    Microsoft: Synapse dedicated SQL pools to Fabric planning
    Microsoft’s planning guide for moving dedicated SQL pool workloads to Fabric Data Warehouse.
    Snowflake SnowConvert AI
    Translates SQL from sources including Azure Synapse into Snowflake SQL.
    Databricks migration documentation
    Databricks’s migration guides by workload type, with Genie Migrate for assessment and code conversion.

    Common questions

    How do you migrate off Azure Synapse?

    Inventory each Synapse component and its consumers, choose a destination against that inventory, convert code and move data with the destination’s tooling, rebuild each metric once, run both in parallel, then repoint reports and retire Synapse.

    Should we go to Fabric, Snowflake or Databricks?

    It depends on the estate. Fabric suits teams staying in Microsoft, Snowflake suits SQL-first warehouse workloads across clouds, and Databricks suits heavy Spark and data engineering. Test a representative workload on each shortlisted platform.

    Is Fabric the default if we already pay for Microsoft?

    It is the shortest path for many Microsoft shops, and Microsoft documents a route for each Synapse component. Check it against your workloads and costs before treating it as the default.

    Do we have to move in one go?

    No. Move in waves by subject area or component, and run each wave in parallel before cutover.

    Who runs a Synapse migration?

    An in-house team can do this with the vendor’s own tools, a systems integrator can staff the programme, or an AI-native data and systems integrator like Rig can run the mapping and cutover.

    Sources

    Vendor documentation checked on 20 November 2026. Confirm current features and terms with the vendor.

    About the author

    Toby Mather, Co-founder and CEO, Rig. Toby co-founded Rig after seeing how hard it was for business teams and AI to work with complex internal data. He was previously co-founder and CEO of Lingumi and CPO at Novakid.

    Talk to Rig

    Check the migration before you commit to it

    An engineer sends a short written view of the scope, systems and risks, then offers a call to go through it.