Data Plane · Lakehouse Platforms

Microsoft Fabric

SaaS analytics platform that unifies data engineering, warehousing, real-time analytics and Power BI on OneLake.

Overview

Microsoft Fabric is a bundle rather than an engine. It packages Power BI, Data Factory, Spark-based data engineering, a warehouse, real-time intelligence and data science into one software-as-a-service platform, all sitting on a single storage layer called OneLake.

OneLake is the part that matters architecturally. Every tenant gets exactly one, it is built on Azure Data Lake Storage Gen2, and it stores tables as Delta Parquet or Iceberg, so any tool that reads those formats can read the data. You cannot create a second one or delete it, which is a deliberate constraint.

Shortcuts give zero-copy access to data that lives elsewhere, including Amazon S3, Google Cloud Storage and on-premises sources, so data can be used without being moved. Mirroring does the same for databases such as Azure SQL, Cosmos DB, Snowflake and Databricks.

Iceberg support works through metadata virtualisation: Iceberg tables can be read as Delta, and Delta tables can be read by Iceberg readers, so Snowflake and Fabric can share the same files.

Licensing is the thing to model carefully. You buy capacity, not products, in F SKUs from F2 upward, and all workloads draw on that shared pool of capacity units. Below F64 every Power BI viewer needs their own paid licence; at F64 and above, free-licence users can view content, which changes the economics sharply.

Governance leans on Microsoft Purview rather than a native catalogue, and Copilot is enabled by default on paid capacities, drawing on the same capacity units.

Key features and capabilities

The same headings are used for every lakehouse platforms entry, so two tools can be read side by side.

Architecture and storage
  • Software as a service: no infrastructure to provision, with workloads on a shared platform layer
  • One OneLake per tenant, built on Azure Data Lake Storage Gen2
  • Shortcuts give zero-copy access to S3, Google Cloud Storage, Azure and on-premises data
  • Compute is a shared pool of capacity units rather than per-engine clusters
Engines and workloads
  • T-SQL warehouse, Spark for data engineering, and Analysis Services for BI
  • Notebooks in Python, Scala, SQL and R
  • Real-time intelligence with event streams and KQL databases
  • Power BI Direct Lake reads from OneLake without copying data
Table formats and catalogues
  • Delta Parquet is the default; only Delta tables appear in the SQL endpoint
  • Iceberg through metadata virtualisation, readable both ways
  • Hudi support is not published
  • OneLake catalog for discovery, with mirroring from Azure SQL, Cosmos DB, Snowflake and Databricks
Security and governance
  • OneLake security roles down to folder, table, row and column, enforced across every engine
  • Microsoft Purview for sensitivity labels, data loss prevention and audit, licensed separately
  • Lineage and impact analysis across items in a workspace
  • HIPAA, ISO 27001, 27017, 27018 and 27701 certifications
AI and machine learning
  • Copilot enabled by default on F2 and above, drawing on the same capacity units
  • Copilot variants per workload, including natural language to SQL and KQL
  • Fabric data agents, with conversation history kept for 28 days
  • Data science integrates with Azure Machine Learning for tracking and model registry
Ecosystem and integrations
  • Data Factory ships more than 200 connectors
  • OneLake supports ADLS Gen2 APIs, so Synapse, Databricks and Storage Explorer connect directly
  • Three published routes for software vendors to integrate or build workloads
  • Scanner APIs expose item metadata to third-party catalogues
Where it runs and what it costs
  • Azure only, across the Americas, Europe, Middle East, Africa and Asia Pacific
  • Some regions are Power BI only; US Government regions are available
  • No self-hosted or bring-your-own-cloud option
  • Geo-replication to a paired region is opt-in per capacity

Pricing

SubscriptionF2 from $262.80 a month

You buy capacity in F SKUs, billed per second. Pay-as-you-go runs from $262.80 a month for F2, $8,409.60 for F64 and upward to F8192; a one or three-year reservation saves around 41%. OneLake storage is charged separately, from $0.023 per GB a month hot down to $0.0050 cold. The licensing cliff matters: below F64 every Power BI viewer needs a paid per-user licence, at F64 and above free users can view. A 60-day trial at F64 scale is offered.

Vendor pricing page →

Demos and videos

About Microsoft

Fabric is a Microsoft first-party service sold through Azure, so it has no founding story or funding of its own. It is proprietary, billed per second with a one-minute minimum, and bought directly in Azure or through a partner or cloud solution provider. Microsoft publishes HIPAA, ISO 27001, 27017, 27018 and 27701 compliance for it. Unlike Databricks, Fabric contributes no engine of its own to open source; it adopts open formats instead.

microsoft.com

Other lakehouse platforms tools

Cloudera

Data Plane · Lakehouse Platforms

Hybrid data and AI platform running the same containerised data services on-premises and in cloud, with its own S3-compatible object store built on Apache Ozone.

  • Commercial

Databricks

Data Plane · Lakehouse Platforms

Lakehouse platform built on Apache Spark and open table formats, governed through Unity Catalog.

  • Commercial

Dremio

Data Plane · Lakehouse Platforms

Lakehouse query engine and semantic layer built on Apache Iceberg and Apache Arrow.

  • Open core

Starburst

Data Plane · Lakehouse Platforms

Commercial data platform based on Trino, for querying data across many sources.

  • Commercial

Drafted with AI assistance and checked against the vendor’s own documentation.