Microsoft Fabric
Microsoft Fabric is Microsoft's SaaS analytics platform. Data integration, Spark data engineering, warehousing, real-time analytics, data science, operational databases and Power BI share one lake (OneLake) and one capacity-based license, so teams do not have to wire separate Azure services together.1
- Intended for
- Data engineers, Data scientists, Business analysts and report authors, Database administrators, Data platform and IT administrators
- Runs on
- Microsoft-hosted SaaS in the Fabric portal, Capacity purchased through Azure or a Cloud Solution Provider, Cross-tenant sharing through OneLake external data sharing
At a glance
- What is it?
- Microsoft Fabric is an analytics platform delivered as software as a service. It groups role-based workloads (Data Factory, Data Engineering, Data Warehouse, Data Science, Real-Time Intelligence, Databases, Industry Solutions, Power BI and the preview Fabric IQ) on a shared platform layer: OneLake for storage, Copilot for AI assistance, and central governance through the OneLake Catalog and Microsoft Purview. Each tenant gets one OneLake, built on Azure Data Lake Storage Gen2, and Microsoft states that you do not need an Azure account to use Fabric. Compute is bought as capacity (F SKUs) that all workloads in assigned workspaces share.1
- What does it do?
- Fabric ingests data with pipelines and Dataflow Gen2 (more than 200 connectors) or replicates operational databases into OneLake through mirroring. Spark notebooks and jobs transform the data, the warehouse and SQL analytics endpoints query it with T-SQL, Real-Time Intelligence handles streams and events, and Data Science trains and registers models with MLflow. Power BI reads the same Delta tables in Direct Lake mode. Shortcuts let OneLake reference data in Azure Data Lake Storage, Amazon S3 or Google Cloud Storage without copying it, and SQL database and Cosmos DB in Fabric add operational databases that mirror themselves into OneLake.123
- Who needs it?
- Organizations that want one governed analytics estate instead of separate ingestion, lake, warehouse and BI services, especially those already on Power BI or Microsoft 365. Microsoft's architecture guidance positions Fabric for teams whose machine learning is part of a wider analytics workflow and who want a managed SaaS service with little infrastructure to run. Power BI Premium P SKU customers are also directed to Fabric F SKUs as the P SKUs retire.45
- What does it need?6
- A Microsoft Entra tenant (one is created automatically when you acquire any free, trial or paid Microsoft online service license)
- Fabric enabled for the tenant by an administrator
- A Fabric capacity (F SKU) or a 60-day Fabric trial capacity
- Power BI Pro or Premium Per User licenses for people who create and share Power BI items outside My workspace
- On capacities smaller than F64, Pro or Premium Per User licenses for everyone who views Power BI content
- What it is not
- Fabric is not a renamed Power BI: Power BI is one Fabric workload, and Power BI Pro or Premium Per User licenses do not provision Fabric capacity for lakehouses, warehouses or notebooks. It is not an Azure PaaS service deployed into your subscription; it runs as SaaS in your Microsoft Entra tenant, with capacity bought through Azure. Microsoft says it is not a replacement for a dedicated machine learning platform when you need deep learning at scale, full MLOps or fine-grained compute control. Fabric Spark also lacks GPU pools, an external Hive Metastore and .NET for Spark.467
Availability and licensing. Generally available. Capacity is sold as F SKUs (F2 to F8192) through Azure or a Cloud Solution Provider, billed per second with a one-minute minimum and optional yearly reservations; prices vary by region. A 60-day trial gives F64-equivalent capacity. Fabric IQ is in preview, and at Build 2026 (June 2026) Microsoft announced GPU-accelerated query acceleration for Fabric Data Warehouse as an early access preview and the Rayfin SDK as a preview. Fabric is not yet available in sovereign clouds (checked 26 Oct 2026).5689
The problem it solves
Many analytics stacks keep separate tools for ingestion, the data lake, the warehouse, data science and reporting, each with its own copy of the data, its own security model and its own bill. Moving data between them takes ETL pipelines that are slow to build and hard to keep in sync.
Fabric runs these workloads on one SaaS platform that stores tables once in OneLake in the open Delta Lake format, so each engine reads the same data. Mirroring and shortcuts reduce the need to copy operational or multicloud data before analysis, and permissions, sensitivity labels and auditing are handled centrally.2
How it works
Fabric runs inside a Microsoft Entra tenant. Its building blocks are:
- Capacity: an F SKU pool of capacity units (F2 to F8192) bought in Azure and billed per second; every workload in its workspaces draws on it.
- Workspaces: containers for items such as lakehouses, warehouses, notebooks, pipelines, eventhouses and semantic models.
- OneLake: one logical lake per tenant on ADLS Gen2. Tables are stored in Delta Lake format and shared by all engines without copies. Shortcuts point to external storage; mirroring replicates databases in near real time.
- Engines: Spark, the warehouse SQL engine, KQL databases for Real-Time Intelligence and Power BI with Direct Lake.
- Governance: the OneLake Catalog for discovery and administration, and Microsoft Purview for information protection, data loss prevention and audit on supported items.16
Diagram as a list
Copilot & business apps
- Power BI (Direct Lake)Semantic models and reports that read OneLake tables directly
Agents & automation
- Fabric IQ (preview)Ontology, data agents and operations agents over shared semantics
Data & analytics
- Operational databases and external storageAzure SQL, Cosmos DB, PostgreSQL, Snowflake, S3 and other systems that hold source dataConnects to Data Factory, mirroring and shortcuts
- Data Factory, mirroring and shortcutsCopies, replicates or references source data into OneLakeConnects to OneLake (Delta tables on ADLS Gen2)
- OneLake (Delta tables on ADLS Gen2)Single tenant-wide store that every engine reads and writes
- Spark (Data Engineering and Data Science)Transforms data and trains models in notebooks and jobsConnects to OneLake (Delta tables on ADLS Gen2)
- Data Warehouse and SQL analytics endpointsT-SQL querying and modeling over Delta tablesConnects to OneLake (Delta tables on ADLS Gen2)
- Real-Time IntelligenceEventstreams and KQL databases for data in motionConnects to OneLake (Delta tables on ADLS Gen2)
Security, identity & governance
- OneLake Catalog and Microsoft PurviewDiscovery, permissions, sensitivity labels, DLP and auditConnects to OneLake (Delta tables on ADLS Gen2)
Cloud & hybrid infrastructure
- Fabric capacity (F SKU)Shared compute meter for all workloads in assigned workspaces
Capabilities
OneLake and shortcuts1
One tenant-wide logical data lake on Azure Data Lake Storage Gen2 that every workload uses. Shortcuts give zero-copy access to Azure Data Lake Storage, Amazon S3, Google Cloud Storage and other stores.
Why it matters: Engines work on one copy of the data, which cuts duplication and pipeline work across teams.
Limits: Shortcut reads depend on the external source; cross-cloud access relies on caching to limit egress costs.
Mirroring2
Continuously replicates databases such as Azure SQL Database, Azure SQL Managed Instance, Azure Cosmos DB, Azure Database for PostgreSQL, SQL Server, Oracle, Snowflake, Google BigQuery and SAP into OneLake as Delta tables. Azure Databricks Unity Catalog is mirrored as metadata only, through shortcuts.
Why it matters: Operational data becomes queryable in Fabric without building ETL pipelines.
Limits: Some sources (Azure Database for MySQL, SharePoint List, Dremio) are in preview. A running capacity is required, and free mirroring storage is capped at one TB per purchased capacity unit.
Data Warehouse and migration110
A T-SQL warehouse that separates compute from storage and stores data in Delta Lake format. The Migration Assistant converts schemas from Azure Synapse dedicated SQL pools and SQL Server.
Why it matters: Gives Synapse and SQL Server warehouse customers a documented path into Fabric.
Limits: T-SQL compatibility with source warehouses is not complete; SQL authentication users, column-level encryption and some objects need rework.
Real-Time Intelligence1
Ingests, stores, analyzes and acts on data in motion such as IoT readings, logs and clickstreams. The Real-Time hub lists streams, KQL database tables and change data capture sources such as Azure SQL Database, Azure Cosmos DB and PostgreSQL.
Why it matters: Operational monitoring and alerting can run on the same platform as historical analytics.
Limits: Streaming items share the capacity with other workloads, so sustained event loads need capacity planning.
Data Science and Foundry integration4
Spark notebooks in Python, R, SQL and Scala, low-code AutoML, MLflow experiment tracking and a model registry. Fabric data agents can pass Fabric data to agents built in Microsoft Foundry.
Why it matters: Predictions and AI answers can be built where the governed data already lives and surfaced in Power BI.
Limits: Model endpoints are in preview and support a limited set of model flavors; there is no responsible AI dashboard, so Microsoft suggests pairing with Azure Machine Learning for broader model governance.
Fabric IQ (preview)18
A workload for shared business semantics that includes ontology, planning, Fabric Graph, data agent, operations agent and semantic models.
Why it matters: Aims to give agents and reports one consistent definition of business entities and metrics.
Limits: Preview. At Build 2026 Microsoft still listed the Fabric IQ ontology as preview.
Capacity licensing56
F2 to F8192 capacities bought through Azure, billed per second with a one-minute minimum, pausable and resizable, with optional yearly reservations. A trial gives F64-equivalent capacity for 60 days.
Why it matters: One meter covers all workloads, and capacity can be paused when idle.
Limits: Below F64 every Power BI viewer needs Pro or Premium Per User; prices vary by Azure region.
Practical use cases
Reports pull from several operational databases through nightly ETL jobs that break and leave data a day old.
- Approach
- Mirror the databases into OneLake and build Power BI semantic models on the mirrored Delta tables in Direct Lake mode.
- Role of Microsoft Fabric
- Fabric replicates the data, stores it in OneLake and hosts the models and reports.
- Data, infrastructure and skills
- A running F capacity, supported source databases and Pro licenses for report authors.
- Type of benefit
- Fresher data and less pipeline maintenance
- Caveats
- Some mirroring sources are in preview; queries on mirrored data consume capacity at normal rates.
- First step
- Mirror one Azure SQL Database into a trial capacity and build one report on it.
An Azure Synapse dedicated SQL pool runs core reporting and needs a long-term home.
- Approach
- Use the Fabric Migration Assistant to convert the schema into a Fabric warehouse, fix flagged objects, then copy data with a copy job.
- Role of Microsoft Fabric
- Fabric Data Warehouse becomes the target warehouse.
- Data, infrastructure and skills
- A DACPAC or connection to the source pool and a Fabric capacity.
- Type of benefit
- Platform consolidation
- Caveats
- T-SQL parity is incomplete; SQL authentication users and column-level encryption need replacements.
- First step
- Upload a DACPAC of the pool to the Migration Assistant and review the objects it cannot convert.
Staff want to ask questions in natural language over governed business data.
- Approach
- Keep the data in OneLake, expose it through a Fabric data agent and connect that agent to an agent built in Microsoft Foundry.
- Role of Microsoft Fabric
- Fabric supplies the governed data and semantics; Foundry hosts the agent.
- Data, infrastructure and skills
- Data modeled in Fabric, a Foundry project and permission design for both.
- Type of benefit
- Faster access to answers
- Caveats
- Fabric IQ items are in preview; check what data each agent can reach.
- First step
- Create a data agent over one well-modeled lakehouse and test it with a small group.
Who uses it
KPMG
ScalingKPMG: rebuilding the Digital Gateway data layer on Microsoft Fabric
about 2 hours instead of about 16 (87% less) (measured)
Professional services, Global (138 countries and territories)
Make-A-Wish America (national office and chapters)
In productionMake-A-Wish America: unified data, Copilot and an IT help bot
roughly two-thirds of staff (measured)
Nonprofit, United States
Chow Tai Fook Jewellery Group
In productionChow Tai Fook: 400+ AI agents for jewelry retail staff
over 400 agents (measured)
Retail and consumer goods, Hong Kong (story dateline); store network across several markets
EY
ScalingEY: Dynamics 365 Sales with agents, and Copilot across the firm
85,000 in 18 months; goal 100,000 (measured)
Professional services, Global
Works with
Optional integration
- Azure DatabricksFabric mirrors Azure Databricks Unity Catalog metadata into OneLake through shortcuts.
- Azure SQLAzure SQL Database and Managed Instance can be mirrored into OneLake.
- Azure Cosmos DBAzure Cosmos DB can be mirrored into OneLake; Cosmos DB in Fabric uses the same engine.
- Azure Database for PostgreSQLAzure Database for PostgreSQL can be mirrored into OneLake.
- Microsoft FoundryFabric integrates with Microsoft Foundry for models and agents.
Complementary tools
- Microsoft PurviewPurview adds information protection, DLP and audit for supported Fabric data.
- Azure Machine LearningMicrosoft suggests pairing Fabric with Azure Machine Learning for broader model governance.
Same family
- Power BIMicrosoft describes Power BI as a core component of Microsoft Fabric.
Alternative approaches
- Azure Synapse AnalyticsMicrosoft says to prefer Fabric over Synapse for new investments.
- Azure DatabricksMicrosoft's guide presents Fabric and Azure Databricks as alternative starting points for analytics-integrated ML.
- Azure Machine LearningWhen ML is a small extension of analytics, Microsoft suggests Fabric or Databricks as a starting point.
Optional integration for
- Azure Cosmos DBCosmos DB can be mirrored into OneLake; Cosmos DB in Fabric uses the same engine.
- Azure Database for PostgreSQLAzure Database for PostgreSQL can be mirrored into OneLake.
- Azure SQLAzure SQL Database and Managed Instance can be mirrored into OneLake.
- Azure DatabricksFabric mirrors Unity Catalog metadata and reads data through shortcuts.
- Microsoft Dragon CopilotA documented integration brings DAX Copilot conversation data into Fabric OneLake.
- Azure IoT OperationsData flows can send data to Microsoft Fabric OneLake, and Microsoft suggests Fabric for dashboards.
- Microsoft Industry Solutions and Dynamics 365 SustainabilityIndustry Solutions in Microsoft Fabric include healthcare and nonprofit data solutions.
- Microsoft PurviewPurview covers Copilot in Fabric and can export catalog data to OneLake.
- Azure AI SearchOneLake is a supported data source and knowledge source.
Relationship labels follow Microsoft's documentation. "Alternative approaches" does not mean one is better: each profile says when it fits.
Compared with
Getting started
Get capacity
Start the 60-day Fabric trial, which provides F64-equivalent capacity, or create an F SKU capacity in the Azure portal.
Check: The capacity is listed in the Fabric admin portal and can be assigned to a workspace.
Create a workspace
Create a workspace and assign it to the capacity so it supports all Fabric experiences.
Check: Workspace settings show a Fabric or Fabric Trial workspace type.
Land data
Create a lakehouse and load data with a pipeline or Dataflow Gen2, or create a mirrored database from an Azure SQL, Cosmos DB or PostgreSQL source.
Check: Tables appear in Delta format and answer queries from the SQL analytics endpoint.
Report on it
Build a semantic model in Direct Lake mode and a Power BI report on top of it.
Check: The report refreshes without an import step.
Govern and size
Apply sensitivity labels, review items in the OneLake Catalog and track usage in the Microsoft Fabric Capacity Metrics app.
Check: Capacity utilization stays below throttling levels during peak hours.
Official resources
Deciding whether you need it?
Three free tools apply Microsoft's own published guidance to your situation, and say when Microsoft Fabric or any new technology is not needed.
Sources
Each statement above links to the source it comes from. Labels say who reported it.
- What is Microsoft Fabric (opens in a new tab)
- Mirroring in Microsoft Fabric (opens in a new tab)
- Cosmos DB in Microsoft Fabric (opens in a new tab)
- AI and machine learning products (Azure Architecture Center) (opens in a new tab)
- Power BI Premium to Microsoft Fabric migration FAQ (opens in a new tab)
- Understand Microsoft Fabric licenses and capacity (opens in a new tab)
- Compare Fabric and Azure Synapse Spark (opens in a new tab)
- Microsoft Build 2026 live blog (opens in a new tab)
- Microsoft Fabric product page (opens in a new tab)
- Migration Assistant for Fabric Data Warehouse (opens in a new tab)
Thank you. Your correction was sent.
The editors check it against the sources. If you left an email address, they may reply about it.