Microsoft Fabric: What It Is, What It Does Well, and Who Should Consider It

Microsoft Fabric is the most significant data platform launch from Microsoft since Azure Synapse. Announced in 2023 and reaching general availability in late 2023, Fabric is Microsoft's attempt to unify the fragmented Azure analytics ecosystem — Azure Synapse Analytics, Power BI, Azure Data Factory, Azure Data Lake Storage, and Azure Machine Learning — into a single integrated platform under a shared governance model and a unified billing unit (Fabric Capacity).

The pitch is appealing: one platform, one bill, one identity and governance model. The reality, as with most platform integration plays, is more nuanced.

What Fabric Actually Is

Fabric is not a single product — it is a collection of experiences hosted under the Fabric umbrella:

**OneLake:** A single logical data lake per organisation, built on Azure Data Lake Storage Gen2. All Fabric experiences (Synapse, Power BI, data engineering) write to and read from OneLake via the Delta Lake table format. The OneLake design is the most architecturally significant aspect of Fabric — it establishes a shared storage layer that all Fabric components access without data copying.

**Data Engineering:** A Spark-based environment for data transformation and pipeline building. Notebooks (Python, Scala, SQL, R) and Spark jobs run on Fabric-managed Spark clusters against OneLake storage. Equivalent to Azure Synapse Spark or Databricks.

**Data Factory in Fabric:** Visual ETL/ELT pipeline builder (dataflows and pipelines) for ingestion from external sources. Similar to Azure Data Factory, with a simplified UI.

**Synapse Data Warehouse:** A dedicated SQL warehouse (separate from Spark) for T-SQL analytics workloads. Supports standard SQL queries against Delta Lake tables in OneLake. Equivalent to Azure Synapse Dedicated SQL Pool or Snowflake for SQL analytics.

**Real-Time Intelligence (formerly Real-Time Analytics):** KQL (Kusto Query Language) database for streaming and time-series analytics. Data is ingested via Eventstream (Kafka-compatible) and queried with KQL. This is a genuine Fabric differentiator — it brings Azure Data Explorer (ADX) capabilities into the Fabric platform.

**Power BI in Fabric:** Power BI is natively integrated. Semantic models (formerly datasets) live in OneLake. DirectLake mode allows Power BI to query Delta Lake files directly without import — combining the speed of imported data with the freshness of direct query.

**Data Science:** ML model development and deployment using notebooks and MLflow integration. Connects to OneLake for training data and model artefacts.

**Data Activator:** No-code alerting and automation triggered by data changes. Alert when a metric crosses a threshold; trigger a Power Automate flow. Not analytically sophisticated but useful for operational alerting.

What Fabric Does Well

**One storage layer, zero data copying.** OneLake with Delta Lake tables means a data engineering team's Spark output is immediately available to Power BI via DirectLake, without an import refresh cycle. For organisations struggling with data freshness in Power BI, this is a meaningful improvement over the traditional import → refresh → report cycle.

**Unified governance with Microsoft Purview.** Microsoft Purview integrates with Fabric to provide data cataloguing, lineage, classification, and sensitivity labels across all Fabric artefacts. For organisations already using Microsoft 365 and Azure AD, the identity model (same users and groups, same security policies) is a genuine simplification versus managing separate identity in Snowflake or Databricks.

**Power BI native integration.** If Power BI is your BI layer, Fabric is purpose-designed for it. DirectLake mode, semantic model integration, and the shared OneLake storage layer create a tighter BI integration than any third-party platform can match.

**Microsoft 365 and Teams integration.** Data alerts, report sharing, and Power BI content embedded in Teams work natively with Fabric. For organisations deeply embedded in the Microsoft 365 ecosystem, Fabric integration reduces friction.

**Real-Time Intelligence (KQL database).** Fabric's KQL database (Azure Data Explorer under the hood) is a genuine competitive advantage for streaming and time-series analytics. ADX/KQL is exceptionally fast for log analytics, IoT telemetry, and time-series queries. Databricks and Snowflake do not have an equivalent native capability.

Where Fabric Has Limitations

**Spark environment maturity.** Fabric's Spark environment is functional but not as mature as Databricks. Delta Lake support is present, but Databricks-specific features (Z-ordering, Auto Optimize, Photon engine, Delta Live Tables for streaming) are not available. Teams migrating from Databricks will find the Fabric Spark experience less capable.

**SQL warehouse performance.** The Synapse Data Warehouse in Fabric is not as fast or cost-predictable as Snowflake for complex analytical SQL workloads. For organisations with demanding SQL analytics workloads, Snowflake or BigQuery still outperform the Fabric SQL endpoint.

**Connector ecosystem.** Azure Data Factory has a mature connector library (200+ connectors). Fabric's pipeline and dataflow experiences have a smaller subset. Some enterprise source system connectors are available only in Data Factory proper, requiring hybrid use.

**Delta Lake lock-in.** All Fabric storage is Delta Lake. If your organisation uses Iceberg (for Snowflake or Trino compatibility), you need conversion or separate storage. The open table format wars (Delta vs Iceberg) have not resolved, and betting entirely on Delta creates dependency on Databricks-controlled open source.

**Pricing complexity.** Fabric capacity units (F SKUs) are the billing unit. F2 through F2048 capacities are available. The capacity reservation model (you reserve capacity, not individual services) is more complex than Snowflake's per-second billing or BigQuery's per-TB model. Right-sizing Fabric capacity requires understanding workload concurrency patterns.

Who Should Consider Fabric

**Microsoft-first organisations** with existing investment in Azure, Microsoft 365, and Power BI. The Fabric integration with these services creates genuine value for organisations already inside the Microsoft ecosystem. Introducing Fabric costs less in integration work than introducing Snowflake or Databricks when the existing stack is Microsoft.

**Power BI-centric organisations** who want DirectLake (eliminating import refresh latency) and tighter semantic model governance. If Power BI is your primary BI tool and refresh latency is a problem, Fabric's DirectLake is the best current solution.

**Organisations with Azure Data Explorer workloads** who want to consolidate Real-Time Intelligence into the broader analytics platform rather than operating ADX separately.

**Organisations evaluating Azure Synapse replacements.** Microsoft is steering customers from Azure Synapse Analytics to Fabric. If you are on Synapse, Fabric is the migration path.

**Who should not default to Fabric:** Organisations with existing Databricks or Snowflake investments (migration cost is high), multi-cloud organisations that need cloud neutrality, and analytics teams without existing Microsoft ecosystem dependency who value the more mature tooling in Databricks (for ML/Spark) or Snowflake (for SQL analytics).

Fabric is a serious platform that will mature significantly over the next 2–3 years. It is not yet a Databricks replacement for ML-heavy organisations or a Snowflake replacement for SQL-performance-critical environments. For the Microsoft-native organisation, it is now the most coherent data platform option in the Azure ecosystem.

Our data architecture consulting practice evaluates Microsoft Fabric alongside Databricks, Snowflake, and BigQuery for platform selection engagements — contact us to discuss your cloud data platform options.

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