The short version
- Microsoft Fabric is a unified SaaS analytics platform built on OneLake as shared storage.
- It bundles data engineering, warehousing, real-time intelligence, data science, and Power BI.
- It is the natural choice for organizations already on the Microsoft stack who want a single integrated platform.
The longer explanation
What's in the box
Fabric packages together what used to be separate Microsoft data products:
- Data Factory. Data integration and pipelines.
- Synapse Data Engineering. Spark-based engineering workloads.
- Synapse Data Warehouse. T-SQL analytical warehouse.
- Synapse Real-Time Intelligence. KQL-based analytics for streaming and high-cardinality data.
- Synapse Data Science. Notebooks, ML tracking, model serving.
- Power BI. Semantic modeling, reports, dashboards.
- Activator. Event-driven alerting and action on analytics outputs.
The unifying substrate is OneLake plus a shared workspace experience, shared security and governance via Purview, and shared capacity-based billing.
Why the platform argument matters
In the old stack, a client might buy Azure Data Factory for ingestion, Azure Databricks for engineering, Azure Synapse dedicated pools for warehousing, a streaming service for real-time, and Power BI for BI. That is five products with five commercial relationships, five governance surfaces, five access control models, and a lot of copy-paste data movement. Fabric collapses those into one.
The trade-off is vendor concentration. Organizations that value platform diversity — or that have workloads where best-of-breed outperforms the unified option — take the more distributed approach. For Microsoft-centric enterprises, the unification benefits usually dominate.
OneLake as the architectural bet
OneLake is the part of Fabric that is genuinely new. It is a tenant-wide lake where every workload stores Delta Parquet tables in a canonical layout. A Data Engineering pipeline writes; a Data Warehouse query reads; a Power BI semantic model ingests — all against the same physical storage, with no copies. "Shortcuts" extend the model to external sources (ADLS, Amazon S3, Databricks Unity Catalog) without data movement.
The consequences matter. Storage cost is paid once, not per tool. Governance applies once, across tools. New workloads compose against the existing lake instead of requiring a data-movement project.
The realistic migration pattern
Most Fabric adoptions we run follow a pattern:
- Enable Fabric capacity on an existing Power BI environment.
- Land the first data domain in OneLake (ingest, engineer, model).
- Migrate a single BI workload to read from the new lake-backed model.
- Expand domain by domain, reusing the CI/CD, governance, and observability assets established in the first domain.
- Retire legacy warehouse and movement pipelines as their downstream consumers migrate.
The first domain takes 12-20 weeks. Subsequent domains ship in 4-8 weeks each.
How Thoughtwave approaches this
Our enterprise data modernization on Microsoft Fabric case study documents our canonical approach. For engagements in discovery, see our Data Analytics & Engineering service and the broader accelerators portfolio.
For the comparison conversation with alternative platforms, see our Microsoft Fabric vs Databricks comparison and the broader data platform decision insight. Our position is vendor-neutral: we deliver Fabric engagements where the fit is strongest, and Databricks or Snowflake where those are the better choice for the client's workload and existing stack.