Microsoft Fabric Overview
Microsoft Fabric is an enterprise-grade SaaS data platform that unifies data engineering, data integration, data science, real-time analytics, and business intelligence (BI). It became generally available (GA) in 2023.
Its core value proposition is enabling data workloads that previously required multiple separate products and licenses under a single platform and single license.
Core Concepts
OneLake (Unified Data Lake)
At the foundation of Fabric is OneLake — a single unified data lake built on Azure Data Lake Storage Gen2 (ADLS Gen2), shared across the entire tenant.
- One per tenant — all workspaces store their data here
- Data is stored in open Delta Parquet format, accessible by a wide variety of tools
- The Shortcuts feature allows accessing data from other storage sources (ADLS Gen2, S3, etc.) without physical copies
Fabric Workloads (Experiences)
Fabric consists of the following workloads:
| Workload | Description |
|---|---|
| Data Engineering | Spark-based Lakehouse processing and data pipelines |
| Data Factory | Data integration / ETL/ELT (copy activities, dataflows) |
| Data Science | ML model development and experiment tracking (Notebooks + MLflow) |
| Data Warehouse | Enterprise DWH powered by T-SQL |
| Real-Time Intelligence | Streaming analytics with Eventstream and KQL databases (Eventhouse) |
| Power BI | BI reports and dashboards |
| Database (Preview) | Operational SQL Database (including Mirroring) |
Power BI and Fabric
Integration Background
Power BI was originally a standalone Microsoft BI service. Since Fabric was announced in 2023, Power BI has been fully integrated as a core part of Fabric.
Accessing app.powerbi.com now redirects to the same UI and feature set as Microsoft Fabric (fabric.microsoft.com).
Workspaces
Workspaces are the common organizational unit shared by both Fabric and Power BI.
- A workspace is a container for Fabric items, where Power BI reports and semantic models coexist with other Fabric items
- Power BI workspaces created before Fabric seamlessly migrated to Fabric workspaces
Workspace License Modes
Workspaces operate under the following license modes:
| License Mode | Description |
|---|---|
| Fabric (Trial/Capacity) | Assigned to a Fabric Capacity. All Fabric workloads available |
| Premium Capacity (P SKU) | Assigned to Power BI Premium. Power BI-specific features (large models, etc.) available |
| Embedded (EM/A SKU) | Capacity for embedded scenarios in external applications |
| Pro | Traditional Power BI Pro license. Fabric workloads not available |
| Shared | Free shared capacity. Functionality is limited |
Direct Lake Mode
Direct Lake mode lets Power BI read Delta tables on OneLake directly — without importing data, with DirectQuery-level flexibility, and near Import-mode performance.
Billing Model
Two Primary Billing Axes
Fabric billing is structured around two main axes: Capacity (compute) and Storage (OneLake).
Fabric Capacity (F SKUs)
| SKU | Compute Units (CU) | Typical Use Case | Monthly (Pay-As-You-Go) |
|---|---|---|---|
| F2 | 2 CU | PoC / Small scale | ~$260/month |
| F4 | 4 CU | Small production | ~$520/month |
| F8 | 8 CU | Medium scale | ~$1,040/month |
| F16 | 16 CU | Medium to large | ~$2,080/month |
| F32 | 32 CU | Large scale | ~$4,160/month |
| F64 | 64 CU | Large scale | ~$8,320/month |
| F128–F2048 | 128–2048 CU | Enterprise | — |
Prices are approximate values for the US East region. Azure Reservations (1-year or 3-year) can provide up to 41% discount.
How Capacity Works (CU / Smoothing)
Fabric compute is managed in units called Capacity Units (CU).
- Each workload job (Spark, SQL, KQL, etc.) consumes CUs
- Smoothing: A buffer mechanism that spreads short-term CU spikes across a 24-hour window, mitigating billing overages from momentary high loads
- Burst Limit: The burst allowance scales with capacity size; jobs are throttled when the limit is exceeded
OneLake Storage Billing
| Storage Type | Price (Reference) |
|---|---|
| OneLake LRS (Locally Redundant) | ~$0.023/GB/month |
| OneLake GRS (Geo-Redundant) | ~$0.046/GB/month |
- Storage charges are billed independently from Capacity
- Delta Parquet compression typically results in significantly smaller sizes than source data
Power BI User Licenses and Fabric
Users who consume Fabric content also need appropriate licenses:
| Scenario | Required License |
|---|---|
| View/create reports in a Fabric workspace | Power BI Pro or Premium Per User (PPU) |
| View reports published as Power BI Apps | Free for viewer with F64+ Capacity |
| Operate Fabric items (Lakehouse, Notebook, etc.) | Fabric (free) license + capacity assignment |
With F64 or higher Capacity, content published as Power BI Apps can be viewed without requiring Pro licenses for end users (equivalent to Power BI Premium behavior).
Fabric Trial
- Anyone with a Microsoft account can try Fabric free for 60 days
- The trial allocates the equivalent of F64 Capacity
- After the trial, you must either move to paid Fabric Capacity or change the workspace mode
Fabric vs. Traditional Azure Services
| Capability | Microsoft Fabric | Traditional Azure Equivalent |
|---|---|---|
| Data Lake | OneLake | ADLS Gen2 |
| Data Integration / ETL | Data Factory in Fabric | Azure Data Factory |
| Spark Processing | Lakehouse (Spark) | Azure HDInsight / Azure Databricks |
| Data Warehouse | Fabric Data Warehouse | Azure Synapse Analytics |
| Streaming | Real-Time Intelligence | Azure Event Hubs + Stream Analytics |
| BI | Power BI (integrated) | Power BI Service (standalone) |
Fabric consolidates all of the above under a single SaaS platform and Capacity-based billing.
Summary
- Microsoft Fabric = a unified SaaS data platform, with Power BI as one of its core components
- Workspaces are the shared organizational unit for both Fabric and Power BI — all items are managed in the same space
- All data is centralized in OneLake (Delta Parquet), enabling fast analytics via Power BI's Direct Lake mode
- Billing consists of two axes: Capacity (F SKU CU consumption) + OneLake storage
- With F64 or higher, Power BI content viewers no longer need Pro licenses