Microsoft Fabric: What It Is, What It Replaces, and When to Use It

The quick answer

Microsoft Fabric is a unified analytics platform launched by Microsoft in 2023 that brings together data engineering, data warehousing, data science, real-time analytics, and BI in a single SaaS product. It replaces and consolidates Azure Synapse Analytics, Azure Data Factory, Power BI Premium, and Azure Data Lake Storage into one platform with shared storage (OneLake), a unified compute layer, and a single governance model. Fabric is Microsoft's primary strategic investment for enterprise data and analytics — Synapse is not the go-forward platform. For Microsoft-ecosystem organisations, Fabric is worth serious evaluation; for organisations not invested in Azure, it adds limited value compared to best-of-breed alternatives.

What Microsoft Fabric includes

Fabric is composed of multiple workloads, all operating on a shared storage layer (OneLake) and governed through a unified catalog:

**OneLake**: the shared storage foundation. OneLake is a single, multi-cloud data lake for every organisation — similar in concept to OneDrive for data. All Fabric workloads read and write to OneLake. Data stored in OneLake is in Delta Parquet format, accessible by any Fabric workload or external engine. Shortcuts allow OneLake to virtually reference data in Azure Data Lake Storage, Amazon S3, or Google Cloud Storage without copying it.

**Data Engineering (Notebooks and Spark)**: Spark-based data engineering for large-scale transformation and data processing. Fabric uses autoscaling Spark clusters with no cluster management overhead — similar to Databricks' managed Spark but within the Fabric ecosystem.

**Data Factory**: a re-imagined version of Azure Data Factory for data ingestion and integration. Dataflows Gen2 (based on Power Query) for no-code transformation and data movement. Data Pipelines for orchestrating complex multi-step workflows. Hundreds of pre-built connectors.

**Data Warehouse**: a cloud-native SQL data warehouse (distinct from Synapse's Dedicated SQL Pool) built on OneLake. Supports T-SQL with full ANSI SQL compliance, concurrent user access, and direct query of Delta Parquet files in OneLake. The Fabric Data Warehouse is the go-forward warehouse for Microsoft-ecosystem organisations.

**Lakehouse**: the lakehouse architecture within Fabric — OneLake storage with Delta Lake table format, accessible via Spark for engineering and SQL for analytics. Each Lakehouse generates an auto-generated SQL endpoint for BI tool connectivity. The Lakehouse is the recommended primary storage pattern for most Fabric deployments.

**Real-Time Intelligence (formerly Real-Time Analytics)**: streaming ingestion and analysis via Eventstream (Kafka-compatible event ingestion), KQL databases (Kusto Query Language for time-series and log analytics), and Activator (trigger automated actions based on real-time data patterns).

**Power BI**: Power BI is fully integrated into Fabric — all existing Power BI workspaces are Fabric workspaces. DirectLake mode (a major new capability) allows Power BI to query Delta Parquet files in OneLake at near-import speeds without materialising a separate dataset. This makes large semantic models practical in Power BI without the memory constraints of import mode.

**Data Science**: MLflow-based ML experiment tracking, model training via Spark notebooks or VS Code integration, and model serving. Feature engineering pipelines integrated with Lakehouse storage.

What Fabric replaces

**Azure Synapse Analytics**: Synapse is not the go-forward data platform in the Microsoft ecosystem. Fabric's Data Warehouse and Lakehouse replace Synapse's Dedicated SQL Pool and Spark Pool respectively. Microsoft has committed to maintaining Synapse but is investing new features in Fabric. Organisations building new data platforms on Azure should start with Fabric, not Synapse.

**Azure Data Factory (partially)**: Fabric Data Factory includes Data Pipelines and Dataflows Gen2, covering the majority of ADF use cases. Complex ADF pipelines (especially those with heavy IR infrastructure requirements or custom Activity types) may not migrate cleanly to Fabric Data Factory. ADF continues to be developed; it is not deprecated.

**Power BI Premium**: Fabric capacities (F SKUs) replace Power BI Premium capacities (P SKUs). F64 is roughly equivalent to P1. Existing P SKU customers can continue renewing; new capacity purchases are on F SKUs.

**Azure Data Lake Storage (partially)**: OneLake virtualises and unifies storage, replacing the need for separate ADLS accounts for each analytics workload. OneLake Shortcuts allow existing ADLS storage to be referenced without migration.

Licensing and pricing

Fabric uses a capacity-based model — you purchase Fabric capacity units (CUs), and all workloads within Fabric draw from the capacity pool. This is different from Synapse's workload-specific pricing.

**Capacity SKUs**: F2 (2 CUs, ~$0.36/hour), F4, F8, F16, F32, F64 (~$12.80/hour equivalent to Power BI P1), F128, F256, F512, F1024, F2048. Larger SKUs benefit from economies of scale.

**Pay-as-you-go vs reserved**: Fabric capacity is available on-demand (hourly billing) or as Azure reservations (1-year or 3-year commitment, 17–41% discount). For production workloads with predictable usage, reserved capacity is significantly cheaper.

**Cost comparison to Synapse + ADF + Power BI Premium**: because Fabric unifies billing, the comparison requires modelling your current spend across all replaced services against Fabric capacity. For organisations paying separately for Synapse Dedicated SQL, ADF, and Power BI Premium, Fabric typically achieves cost parity or savings at F64+ capacity with consolidation benefits.

DirectLake: the major Power BI advancement

DirectLake mode is the most significant capability in Fabric for Power BI users. It allows Power BI semantic models to read Delta Parquet files in OneLake directly, with performance approaching import mode — without the need to ingest data into the Power BI in-memory VertiPaq engine on a schedule.

In traditional Power BI, the options were: import mode (fast queries, stale data between refreshes) or DirectQuery mode (live data, slower queries). DirectLake provides fast queries against live data — the Delta Parquet files in OneLake are the source of truth, and Power BI reads them directly without a separate materialisation step.

DirectLake is only available for data stored in Fabric's OneLake. Power BI semantic models connecting to external databases (Snowflake, BigQuery, Redshift) cannot use DirectLake — only data in Fabric Lakehouse or Fabric Data Warehouse on OneLake.

When to use Fabric

Fabric is the right choice when: you are a Microsoft-first organisation (Azure infrastructure, Office 365, Power BI users); you are evaluating Azure Synapse and want the more modern platform; you want consolidated billing and governance across analytics workloads; you are a heavy Power BI user and want DirectLake mode for semantic model performance.

Fabric is less compelling when: your data stack is primarily on Snowflake, Databricks, or GCP (Fabric's advantages are OneLake-centric); your team has deep Databricks expertise (Databricks remains stronger for ML/AI and complex Spark engineering); or you want multi-cloud flexibility (Fabric is Azure-native, though OneLake Shortcuts provide limited cross-cloud storage virtualisation).

For the Synapse vs Databricks comparison, see azure synapse vs databricks. For the Microsoft Fabric migration path, see microsoft fabric migration guide. For the warehouse architecture context, see data lakehouse vs data warehouse.

Our data architecture consulting practice evaluates and implements Microsoft Fabric — from platform assessment and licensing guidance through Lakehouse design, Data Factory pipeline migration, and Power BI DirectLake semantic model implementation. Book a free 30-minute audit to discuss whether Fabric is the right choice for your environment.

A former Microsoft data architect audits your data foundation, identifies your top priorities, and sends you a written plan. Free. No pitch.