Skip to content
Migration centers — AWS, GCP, Snowflake, Databricks, Teradata, Cloudera, Informatica, Palantir, SAS, Oracle to Azure

Migration — Databricks to Microsoft Fabric (Package Index)

Comparative positioning note

This document is written from the perspective of Microsoft Azure, Cloud Scale Analytics, and CSA Loom. Any description of third-party or competing products, services, pricing, or capabilities is derived from publicly available documentation and sources believed accurate at the time of writing, and is provided for general comparison only. We do not claim expertise in, or authority over, any non-Microsoft product or service; the respective vendor's official documentation is the authoritative source for their offerings, which may change over time. Nothing here is intended to disparage any vendor — where a competing product has genuine advantages, we aim to note them honestly. Verify all third-party details against the vendor's current official documentation before making decisions.

Status: Authored 2026-04-30 Audience: Teams running Databricks (on Azure, AWS, GCP) evaluating Microsoft Fabric as a strategic consolidation target for BI, data engineering, real-time analytics, and AI workloads. Scope: Complete migration package — assessment through decommission — with feature-level mapping, hands-on tutorials, benchmarks, and best practices for hybrid and full-migration scenarios.

Quick decision: should you migrate?

flowchart TD
    Start[Current Databricks workload] --> Q1{Primary use?}
    Q1 -->|BI / dashboards / Power BI| A1[Strong Fabric fit<br/>Direct Lake + unified capacity]
    Q1 -->|Heavy ML / DL training| A2[Stay on Databricks<br/>Photon + MLflow + GPU clusters]
    Q1 -->|Data engineering / ELT| Q2{Notebook-heavy<br/>or dbt-native?}
    Q2 -->|dbt-native / SQL-first| A3[Strong Fabric fit<br/>dbt-fabric + Lakehouse SQL]
    Q2 -->|PySpark notebook-heavy| A4[Evaluate carefully<br/>see notebook-migration.md]
    Q1 -->|Streaming / real-time| Q3{Latency requirement?}
    Q3 -->|Sub-second| A5[Fabric RTI / Eventhouse<br/>better for sub-second BI]
    Q3 -->|Seconds-to-minutes| A6[Either platform works<br/>evaluate cost + ops burden]
    Q1 -->|Multi-cloud requirement| A7[Stay on Databricks<br/>Fabric is Azure-only]

    style A1 fill:#0078d4,color:#fff
    style A3 fill:#0078d4,color:#fff
    style A5 fill:#0078d4,color:#fff
    style A2 fill:#ff6b35,color:#fff
    style A7 fill:#ff6b35,color:#fff
    style A4 fill:#e8a317,color:#fff
    style A6 fill:#e8a317,color:#fff

Most enterprises land on hybrid. Databricks handles heavy ML/training; Fabric handles BI, ad-hoc analytics, and real-time. OneLake shortcuts let both engines read the same Delta tables. See best-practices.md for the hybrid playbook.

Decision matrix

Workload category Databricks strength Fabric strength Recommendation
BI semantic models DBR SQL endpoint + Power BI Import Direct Lake (zero-copy) + native PBI Fabric
Ad-hoc SQL analytics DBSQL warehouse, Photon Lakehouse SQL endpoint, auto-optimized Fabric (cost)
PySpark notebooks Photon runtime, GPU attach Fabric Spark (forked OSS) Databricks (perf)
dbt transformations dbt-databricks adapter, mature dbt-fabric adapter, growing Either
Delta Live Tables DLT (declarative, expectations) Data Pipelines + dbt tests Databricks (maturity)
MLflow experiments Native MLflow, Unity Catalog lineage Fabric ML experiments (limited) Databricks
Model serving Databricks Model Serving, GPU Azure ML managed endpoints Databricks
Feature store Databricks Feature Store + UC Fabric feature engineering (preview) Databricks
Structured streaming Structured Streaming, Auto Loader Real-Time Intelligence / Eventhouse Fabric (sub-second)
Governance / catalog Unity Catalog (3-level namespace) OneLake + Purview Databricks (maturity)
Cost model DBU tiers (Jobs, SQL, All-Purpose) Fabric CU (single capacity) Fabric (simplicity)
Multi-cloud AWS, Azure, GCP Azure only Databricks

Package contents

Strategic & cost

Document Description Lines
why-fabric-over-databricks.md Strategic white paper: when Fabric is the right move and when it is not ~400
tco-analysis.md DBU pricing vs Fabric CU, reserved capacity, storage, worked examples ~350
benchmarks.md Photon vs Fabric Spark, DLT vs RTI, SQL warehouse comparisons ~300

Feature mapping & migration guides

Document Description Lines
feature-mapping-complete.md 40+ feature-by-feature mapping: Databricks to Fabric equivalents ~400
notebook-migration.md PySpark notebooks, dbutils, library management, Databricks Connect ~350
unity-catalog-migration.md Unity Catalog to OneLake + Purview: catalogs, schemas, RBAC, lineage ~400
dlt-migration.md Delta Live Tables to Fabric Data Pipelines + dbt, expectations, monitoring ~350
ml-migration.md MLflow, Model Serving, Feature Store, AutoML, Vector Search ~350
streaming-migration.md Structured Streaming to Fabric Real-Time Intelligence / Eventhouse ~300

Hands-on tutorials

Document Description Lines
tutorial-notebook-to-fabric.md Step-by-step: convert a Databricks PySpark notebook to Fabric ~350
tutorial-dlt-to-fabric-pipeline.md Migrate a DLT pipeline to Fabric Data Pipeline + dbt with quality tests ~350

Operations

Document Description Lines
best-practices.md Hybrid strategy, workspace mapping, capacity planning, pitfall avoidance ~300

Reading order

If you have 30 minutes: Read this index + why-fabric-over-databricks.md.

If you are building a business case: Add tco-analysis.md + benchmarks.md.

If you are planning the migration: Read feature-mapping-complete.md first, then the specific migration guide for your primary workload (notebooks, DLT, ML, streaming).

If you are hands-on-keyboard: Jump to the tutorials and best-practices.md.

Maintainers: csa-inabox core team Source finding: CSA-0083 (HIGH, XL) -- approved via AQ-0010 ballot B6 Last updated: 2026-04-30