Microsoft Fabric

Microsoft’s software-as-a-service analytics platform: Spark, a SQL warehouse, pipelines, a KQL eventhouse and Power BI all reading and writing one tenant-wide lake, OneLake, paid for from shared capacity.

What it is: Lakehouses,Data warehouses
Used for: ETL & ELT pipelines,Dashboards and BI
Type: Managed service on Azure
Made by: Microsoft
License: Proprietary
Facts checked: 24 October 2026

What it is

Fabric bundles what used to be separate Azure services into one SaaS product, run from a web portal with no Azure resources to set up. Every tenant gets exactly one OneLake, a logical data lake on Azure Data Lake Storage Gen2 that stores tables as Delta Lake, and through metadata virtualization also exposes them as Iceberg. On top of it sit workloads: Data Engineering (Spark notebooks and jobs over lakehouses), Data Warehouse (a T-SQL engine that separates compute from storage and writes Delta tables), Data Factory (pipelines and Power Query dataflows, with more than 200 connectors), Real-Time Intelligence (eventstreams, and eventhouses that hold KQL databases, as in Azure Data Explorer), Data Science, Databases (a SQL database and mirroring of external databases into OneLake) and Power BI. Each engine reads the same files, so a table loaded by T-SQL can be read by Spark, and Power BI’s Direct Lake mode reports on it without importing a copy.

Shortcuts mount data from other workspaces, ADLS, Amazon S3, S3-compatible stores, Google Cloud Storage and Iceberg sources without copying it; mirroring replicates databases such as Azure SQL Database, Cosmos DB and Snowflake into Delta tables, or syncs an external catalog such as Azure Databricks’. Everything runs on capacities: an F SKU from F2 to F8192 is a pool of capacity units (CUs) that all workloads share, billed by the second in Azure with no commitment, or about 41% less with a reservation. A 60-day trial gives the equivalent of an F64. Viewing Power BI content on capacities smaller than F64 still needs a Power BI Pro or Premium Per User licence for each viewer.

Why teams pick it

One copy of the data in open Delta (and Iceberg-readable) tables, shared by Spark, T-SQL, KQL and Power BI without exports
No infrastructure to provision: workspaces, lakehouses and warehouses are created in a browser and governed centrally in the OneLake catalog
Close ties to Power BI, Excel, Teams and Entra ID make it the shortest path for organisations already on Microsoft 365
Shortcuts and mirroring bring in S3, GCS, Snowflake, Databricks and operational databases without hand-built pipelines

Watch out for

Capacity-based billing pools every workload, so one heavy Spark job or report can throttle others on the same capacity; sizing takes monitoring
Licensing is layered: F capacities, Power BI Pro or PPU per user below F64, and Power BI Premium P SKUs that Microsoft is retiring
A proprietary SaaS tied to Azure and Entra ID; the open formats keep the data portable, but pipelines, notebooks and semantic models are Fabric items
A broad, fast-moving product: features such as Fabric IQ are still in preview, and item types and licences keep changing

Specs

Compare with alternatives →

Data model: Delta Lake tables in one tenant-wide lake, OneLake
Query language: T-SQL, Spark and KQL over the same tables
How it scales: Capacity units (F2 to F8192) shared by every workload
Storage format: Delta Lake on ADLS Gen2, also exposed as Iceberg
Ways to run it: SaaS on Azure
Pricing model: F capacity billed per second, about 41% less reserved; Power BI Pro per viewer below F64

Alternatives

What all lakehouses have in common →Compare side by side →

DatabricksCommercialA managed lakehouse platform built on Apache Spark and Delta Lake, with Unity Catalog for governance, on AWS, Azure or Google Cloud.

SnowflakeCommercialA fully managed cloud data warehouse that separates storage from compute, with virtual warehouses billed per second.

Google BigQueryGoogle CloudGoogle Cloud’s serverless data warehouse: no clusters, and you pay per TiB scanned or per slot-hour.

Azure Data ExplorerAzureMicrosoft’s managed analytics database for logs, telemetry and time series, known internally as Kusto and queried with KQL; it began as the back end of Azure Monitor’s Log Analytics and Application Insights.

Amazon S3 TablesAWSA bucket type for Apache Iceberg tables: S3 stores them, acts as their catalog, and compacts and cleans them up for you.

Apache AmoroIncubatingOpen sourceAn incubating Apache “lakehouse management system”, started at NetEase as Arctic, that watches Iceberg tables and keeps them fast on its own: it compacts small files, rewrites deletes and expires old snapshots and files in the background.

DremioCommercialA SQL lakehouse engine built on Apache Arrow, with its own Iceberg catalog and “reflections” that speed up queries without anyone changing them.

LakeOpsCommercialA commercial “control plane” for Apache Iceberg lakehouses that connects to your catalog, compacts and maintains tables on its own based on their health, and routes SQL from one endpoint to Trino, Snowflake, Athena, DuckDB, StarRocks or ClickHouse.

OnehouseCommercialA managed lakehouse from the creator of Apache Hudi that runs in your own cloud account: it ingests data into Hudi, Iceberg or Delta Lake tables in your buckets, keeps them optimized, syncs them to other catalogs and runs Spark, SQL, Trino, Flink and Ray on them.

SpiralCommercialA closed-source storage and query engine for multimodal AI data (video, images, sensor traces and tables) that keeps everything as Vortex files in your own object storage, versions every change and streams results to GPUs; access is by request.

Tiger LakeBetaCommercialTiger Cloud’s built-in Iceberg connector, in public beta: it copies a Postgres table or TimescaleDB hypertable into Apache Iceberg, then keeps it current from the write-ahead log.

How it fits with other technologies

What Microsoft Fabric uses

Workloads on one lake#

From Microsoft’s overview of Fabric, checked 2026-10-24. All of them store data in OneLake and draw on the same capacity.

Data Engineering

Spark

Lakehouses, Spark notebooks and job definitions for processing large datasets, scheduled through Data Factory.

Data Warehouse

T-SQL

A transactional SQL warehouse with compute separated from storage, writing its tables in Delta Lake format.

Data Factory

ingestion

Pipelines and Power Query dataflows, with more than 200 connectors to sources on premises and in the cloud.

Real-Time Intelligence

KQL

Eventstreams, the Real-Time hub, and eventhouses: KQL databases for logs, telemetry and time series.

Databases

operational

SQL database in Fabric, and mirroring that replicates Azure SQL Database, Cosmos DB, Snowflake and others into OneLake.

Power BI

Reports, dashboards and semantic models; Direct Lake mode reads OneLake tables without an import.

Data Science

Model training with experiment tracking and a model registry, with predictions written back for reports.

Fabric IQ

preview

Ontology, Fabric Graph, data agents and semantic models for shared business definitions.

See it in a real stack

No architecture uses Microsoft Fabric yet. Drop it into the playground and wire it up yourself.Open playground →

Where it came from2023

A Microsoft analytics project used to combine Azure Data Factory for ingestion, Azure Synapse Analytics for Spark and SQL, a data lake in Azure Storage and Power BI for reports, each a separate service. On 23 May 2023, at Build, Arun Ulag announced Microsoft Fabric as one product that combined Data Factory, Synapse and Power BI, with seven workloads on a built-in lake, OneLake, and Delta on Parquet as the default format for all of them.

Microsoft set Fabric to be enabled for Power BI tenants from July 2023, and it became generally available on 15 November 2023, by which time Microsoft said about 25,000 organisations had adopted it. Since then Microsoft has added a SQL database, mirroring, Iceberg interoperability for OneLake and Fabric IQ, renamed the real-time workload Real-Time Intelligence, and steered Power BI Premium capacity customers toward Fabric F SKUs.

Why it’s built this way

Software as a service, not a set of Azure resources: Wiring separate services together took expertise most analysts lacked. A SaaS that works like Office, with one lake per tenant and no clusters to provision, puts every workload behind one sign-in.
Delta Parquet as the format for every engine: If Spark, T-SQL and Power BI all read and write the same open tables, data needs no copies between workloads and can leave Fabric without conversion.
One pool of capacity for all workloads: Buying compute per service left some idle while others ran short; a shared capacity lets unused units in one workload serve another.
Shortcuts instead of copies: Data already lives in other lakes and clouds; mounting it in OneLake avoids ETL and egress for data that doesn’t need to move.

Timeline

2023
Announced in preview at Build on 23 May; enabled for Power BI tenants from July
2023
Generally available on 15 November

Sources5 · facts checked 24 October 2026

Improve this page on GitHub