What Is Microsoft Fabric?
Microsoft Fabric is Microsoft's unified, end-to-end SaaS analytics platform — bringing data engineering, data warehousing, real-time analytics, data science, and Power BI into a single tenant-wide environment built on a shared lake called OneLake. Generally available since November 2023, Fabric consolidates capabilities that previously lived as separate Azure services into one product, billed against one capacity, governed by one permissions model.
This guide covers what Fabric is, what it includes, how it differs from Azure Synapse and other Microsoft analytics products, and where it fits in modern enterprise data architectures.
Microsoft Fabric is a SaaS analytics platform that unifies six core workloads — Data Engineering, Data Warehouse, Real-Time Intelligence, Data Science, Data Factory, and Power BI — on top of OneLake, a single tenant-wide Delta Lake. All workloads share storage, governance, capacity, and identity. Customers buy capacity (F-SKUs) priced by Capacity Units; everything Fabric does runs against that pool. The breakthrough capability is Direct Lake, which lets Power BI read OneLake Delta tables at import-mode speed with no copy or refresh.
What Is Microsoft Fabric?
Before Fabric, a typical Microsoft analytics stack looked like this: Azure Data Factory for ingestion, Azure Synapse for warehousing and big-data engineering, Azure Stream Analytics for streaming, Azure Data Explorer for log/telemetry, Azure ML for machine learning, and Power BI for visualization. Each was a separate service, with separate billing, separate identity, separate connectors, separate governance, and separate storage.
Fabric collapses that into one SaaS environment. The same data sits in OneLake. The same capacity powers every workload. The same Microsoft Entra identity governs access. The same Power BI workspace surfaces the results. For organizations already standardized on the Microsoft estate, this collapse is the value proposition.
Core Components
Fabric ships six core workloads, all wrapped in shared services:
- Data Factory — pipelines and Dataflows Gen2 for ingestion and orchestration. Over 200 connectors. Low-code Power Query M for transformations.
- Data Engineering — Lakehouses (Delta + files) and Spark notebooks. Code-first authoring in PySpark, Scala, SQL, R.
- Data Warehouse — fully managed warehouse on the Polaris engine, with T-SQL, multi-table ACID transactions, stored procedures, views.
- Real-Time Intelligence — Eventhouses (KQL/Kusto), Eventstreams, and Activator for high-volume streaming and event-driven workflows.
- Data Science — notebooks, MLflow, SynapseML, AutoML, and integration with Azure AI Foundry for advanced LLM scenarios.
- Power BI — reports, dashboards, semantic models, paginated reports. Direct Lake reads OneLake at import-mode performance.
Cross-cutting: Copilot for AI assistance throughout, Microsoft Purview for enterprise governance, Microsoft Entra ID for identity, and a unified Catalog for asset discovery.
OneLake — The Foundation
OneLake is the most important architectural decision in Fabric. It is a single, tenant-wide Delta Lake automatically provisioned for every Fabric tenant. Every Fabric workload reads and writes here. There is no concept of a separate storage account per service.
Key OneLake capabilities:
- One copy of data — all workloads see the same Delta tables. A Lakehouse table created in a notebook is queryable from the Warehouse, accessible to a Power BI semantic model, and visible to Purview without any copy step.
- Shortcuts — virtual references to data in Amazon S3, Azure Data Lake Gen2, Dataverse, Google Cloud Storage, or other OneLake locations. Data appears native to the workspace without being moved.
- Open format — OneLake uses Apache Parquet on disk in Delta Lake format. Support for Apache Iceberg interop allows Snowflake, Databricks (UniForm), and other Iceberg-aware engines to read and write OneLake tables.
- Domains — logical groupings of workspaces aligned to business domains, enforcing data mesh-style federation.
Workloads in Fabric
Each workload uses its own engine but writes to the same OneLake Delta format.
- Lakehouse vs Warehouse — Lakehouses are Spark-backed and code-first; Warehouses are SQL-backed and fully managed. Both store Delta tables in OneLake. The choice is mostly about who builds the workload — a data engineer (Lakehouse) or a SQL developer / analyst engineer (Warehouse).
- Spark engine — Microsoft-tuned with V-Order optimization for Direct Lake performance. Notebooks, jobs, and Spark Job Definitions.
- Eventhouse / KQL — Kusto-based real-time engine for telemetry, logs, and time-series at billions of rows per minute.
- Activator — no-code event triggers. Watches Real-Time data and fires actions (notifications, Power Automate, custom hooks) when conditions are met.
- Mirroring — replicates external operational databases (Cosmos DB, Snowflake, Azure SQL, Databricks) into OneLake near real-time, without ETL.
Direct Lake Mode
The most consequential Power BI capability in Fabric is Direct Lake mode. Direct Lake reads Delta tables in OneLake directly into Power BI's VertiPaq engine — column by column, on demand — bypassing the traditional Import / DirectQuery dichotomy. Performance is comparable to Import mode; freshness is comparable to DirectQuery; storage cost is essentially zero because there is no copy.
For a deeper look, see our Power BI Fabric Integration guide.
One copy of data is the design principle. Pre-Fabric, getting the same data into a warehouse, a Power BI dataset, and a Spark job typically meant three copies kept in approximate sync by orchestration and luck. OneLake collapses this into one Delta table that every workload reads. Whether the gain materializes for a given organization depends on whether the underlying workloads can actually share storage — which usually depends on adopting Delta as the lingua franca.
Pricing Model
Fabric pricing is capacity-based. Customers purchase F-SKU capacities — pools of Capacity Units (CUs) — sized from F2 (small team) to F2048 and beyond. Within a capacity, all Fabric workloads consume CUs at workload-specific rates. Pricing options:
- Pay-as-you-go (PAYG) — billed by the second of capacity uptime. Capacity can be paused.
- 1-year reservation — discounts of approximately 41% versus PAYG.
Fabric also includes smoothing: short-term spikes are deferred and amortized across the capacity, so a 10-second burst at 4x capacity does not throttle. Sustained over-utilization eventually throttles the capacity until consumption recovers.
Power BI Premium Per User and Power BI Pro licenses remain valid for users without dedicated capacity. Fabric capacities replace the older Power BI Premium per Capacity SKUs (P1 ≈ F64, etc.).
Fabric vs Azure Synapse
Microsoft positions Fabric as the successor to Azure Synapse Analytics. The two products overlap heavily, with key differences:
- Deployment model — Fabric is SaaS (no networking to manage). Synapse is PaaS (more control, more setup).
- Storage — Fabric uses OneLake by default; Synapse uses customer-managed ADLS Gen2.
- Integration with Power BI — Fabric is unified; Synapse is connected.
- Roadmap — Microsoft directs new investment to Fabric. Synapse continues to receive maintenance and security updates but is not the strategic product going forward.
For most new analytics workloads inside the Microsoft estate, Fabric is the recommended starting point. Existing Synapse deployments do not need to migrate immediately, but new capabilities (Direct Lake, OneLake, Real-Time Intelligence) are Fabric-only.
Use Cases
Fabric is well-suited to:
- End-to-end analytics in Microsoft-centric organizations — ingestion, transformation, warehouse, BI, and ML in one product.
- Self-service and citizen analytics — Dataflows Gen2, Power BI, and Copilot lower the technical bar for non-engineers.
- Real-time operational analytics — Eventhouses + Activator handle telemetry, fraud detection, and operational monitoring at scale.
- Mirroring and operational reporting — replicate Cosmos DB, Azure SQL, Snowflake, or Databricks into OneLake for unified analytics without ETL pipelines.
- Data mesh adoption — Domains and workspaces support federated data ownership while preserving cross-domain discoverability.
Getting Started
The fastest paths to evaluate Fabric:
- 60-day free trial capacity — Microsoft offers free Fabric trial capacities for evaluation; tenants enable trial through Fabric admin.
- Power BI customers — every Power BI Premium customer effectively already has Fabric capacity. The new workloads are unlocked at the tenant level.
- Microsoft Learn paths — official, free training for each workload (Data Factory, Lakehouse, Warehouse, Power BI, Data Science).
- Reference architectures — Microsoft publishes opinionated patterns for medallion architecture, streaming, and master data management on Fabric.
Fabric is a substantial commitment to the Microsoft analytics roadmap. Within that ecosystem, it is the most coherent and integrated platform Microsoft has shipped — and for organizations whose data team and business users already live in Microsoft 365, the time-to-value is genuinely fast. Outside that ecosystem, the integration story is less compelling, and Fabric competes head-on with Databricks and Snowflake on more even footing.
