Glossary

Microsoft Fabric

By Emil Björk · Microsoft business apps consultant, Gothenburg

Microsoft's unified analytics platform — lakehouse, data warehouse, real-time analytics, and Power BI on a shared storage layer called OneLake.

Microsoft Fabric is Microsoft's unified analytics platform, combining lakehouse, data warehouse, real-time analytics, data engineering, data science, and Power BI on a shared SaaS storage layer called OneLake. For Dynamics 365 customers, Fabric is the increasingly default destination for cross-system analytics. Synapse Link streams Dataverse and F&O data into OneLake; Power BI in DirectLake mode queries the parquet directly with near-import performance and no refresh schedules. Fabric is sold in F-SKU capacity tiers; capacity sizing — peak Power BI users, lakehouse compute, model size — is the main cost lever. Microsoft Purview integrates for governance.

In practice, Fabric replaces a hand-built data pipeline for cross-system reporting rather than replacing Power BI's day-to-day reporting itself. A company running both Business Central and Dynamics 365 Sales historically needed a separate data warehouse and ETL jobs to build a single report combining both; Synapse Link streams both sources into OneLake as they change, and DirectLake mode lets Power BI query that combined data without a scheduled refresh at all. The reporting experience for an end user looks identical to a normal Power BI report — the difference is entirely in how fresh and how unified the underlying data is.

Where confusion shows up: Fabric capacity (the F-SKU) is a shared resource pool, not a per-report or per-user cost, which makes it easy to under- or over-provision if nobody models peak concurrent usage across every workload sharing that capacity — a lakehouse refresh, a data science notebook, and a Power BI report can all compete for the same capacity at once. Sizing conservatively and monitoring the capacity metrics app from day one avoids the common failure mode of a capacity that runs fine in testing and throttles the moment real concurrent usage arrives.

Related terms

Mentioned in these guides

And 6 more — this term shows up across the library.