Microsoft Fabric

Microsoft Fabric is Microsoft's answer to the fragmented data stack: a single unified analytics platform that consolidates what previously required separate tools — Azure Data Factory for ingestion, Azure Synapse for warehousing and Spark, and Power BI for reporting — into one integrated experience with shared governance, a single identity model, and one underlying storage layer.

OneLake is the foundation. Every Fabric workspace gets a unified data lake backed by Azure Data Lake Storage Gen2, with data stored in open Delta/Parquet format. All workloads — Lakehouse, Warehouse, Spark notebooks, Real-Time Intelligence, Power BI — read from and write to the same OneLake, eliminating data duplication and movement. Shortcuts extend OneLake to external storage (AWS S3, Google Cloud Storage, ADLS) without copying data, treating remote buckets as if they were local OneLake folders.

Data Engineering in Fabric uses Apache Spark under the hood. Engineers write PySpark, Spark SQL, Python, Scala, or R in notebooks with full Git integration and CI/CD pipeline support. The Lakehouse item type gives each data domain a managed table layer (Delta tables) plus a raw files area, with automatic schema-on-read and T-SQL endpoint for SQL consumers.

Data Factory (Fabric's pipeline engine, distinct from Azure Data Factory) provides 200+ native connectors for batch and incremental ingestion. Pipelines support low-code drag-and-drop authoring and code-first configurations, scheduling, and monitoring — handling everything from database CDC to API pulls to file drops.

Data Warehouse provides a T-SQL serverless compute layer over OneLake tables. It is not a copy of the data — it is a SQL engine that reads Delta files directly, enabling BI teams to query lakehouse data with standard SQL without Spark knowledge. Cross-database queries across warehouses and lakehouses are supported natively.

Real-Time Intelligence (formerly Real-Time Analytics) targets sub-second streaming analytics. Events are ingested via Eventstream (Kafka-compatible), stored in Eventhouse (KQL database), and queried with Kusto Query Language (KQL). Reflexes trigger automated actions when data crosses thresholds.

Power BI is built into Fabric as a first-class workload. Direct Lake mode reads Delta files from OneLake directly into the Power BI engine without import or DirectQuery overhead, combining import-speed query performance with real-time data freshness. All Power BI reports and semantic models live in the Fabric workspace alongside the engineering artifacts that produce the data.

Copilot is woven throughout: in notebooks it generates and explains PySpark/SQL code; in pipelines it suggests transformations; in Power BI it answers natural language questions and generates DAX measures. Available on F64+ capacity.

Microsoft Purview integration provides cross-workspace data lineage, sensitivity labels, classification scanning, and access policies that apply uniformly across all Fabric workloads — a significant governance advantage for regulated industries.

Billing is capacity-based: organizations purchase an F-SKU (F2 through F2048), which represents Compute Units (CUs). All workloads within the org consume CUs from this shared pool. A 60-day free trial at F64 capacity is available.