Data warehouse, data lake or lakehouse: what do the terms mean on the Microsoft stack?

Three terms, three Microsoft services

The three terms describe how data is stored and who can work with it. A data warehouse holds curated, structured tables designed for SQL queries and reporting. A data lake holds raw data of any shape in files, so it can be stored first and modeled later. A lakehouse puts warehouse-style tables, with transactions and schemas, directly on lake storage so both data engineers and analysts can use the same copy. On the Microsoft stack, the warehouse is Fabric Data Warehouse, the lake is Azure Data Lake Storage or OneLake, and the lakehouse is the Fabric lakehouse item, and in Fabric all of them store Delta tables in OneLake and share one SQL engine (Microsoft Learn, Microsoft Fabric decision guide: Choose between Warehouse and Lakehouse, accessed 2026-10-20). EPC Group's Microsoft Fabric consulting practice uses these definitions to settle naming debates early, because most organizations need more than one of the three, and the real decision is which data goes where.

The data warehouse: Fabric Data Warehouse

Microsoft describes Fabric Data Warehouse as an enterprise-scale relational warehouse on a data lake foundation. It is developed primarily with T-SQL, shares a large surface area with the SQL Database Engine, supports full multi-table ACID transactions, materialized views, functions and stored procedures, and loads data through COPY INTO, pipelines, dataflows and cross-database queries. Storage and compute are separated, workload management is autonomous, and the data itself sits in Delta tables in OneLake (Microsoft Learn, What is Fabric Data Warehouse?, accessed 2026-10-20). Microsoft recommends the warehouse for enterprise, departmental or self-service warehousing and for structured analysis in T-SQL, while the lakehouse SQL analytics endpoint is read-only, with no DML (Microsoft Learn, Microsoft Fabric decision guide: Choose between Warehouse and Lakehouse, accessed 2026-10-20). EPC Group still builds dimensional warehouses for finance and regulatory reporting, where conformed dimensions and transactional loads matter more than raw flexibility.

The data lake: Azure Data Lake Storage and OneLake

A data lake stores everything first. Microsoft defines it as a single, centralized repository for structured and unstructured data kept in its raw or native format, usually as files or blobs, without conforming it to an existing structure. Azure Data Lake Storage is not a separate account type: enabling the hierarchical namespace on a Blob Storage account turns on Hadoop-compatible access through the ABFS driver, directories with atomic rename and delete, Azure RBAC and POSIX access control lists at directory and file level, and encryption at rest (Microsoft Learn, Introduction to Azure Data Lake Storage, accessed 2026-10-20).

OneLake is the Fabric version of the same idea. Every Fabric tenant includes exactly one, built on Azure Data Lake Storage, organized by workspace and item, readable through ADLS Gen2 APIs, and able to reference external ADLS, Blob Storage or Amazon S3 data through shortcuts instead of copies. Microsoft states plainly what OneLake was built to fix: separate lakes per business group created management overhead, slowed projects and increased duplication (Microsoft Learn, What is OneLake?, accessed 2026-10-20). EPC Group keeps existing ADLS accounts where they work and connects them with shortcuts rather than moving data twice.

The lakehouse: Delta tables on the lake

The Fabric lakehouse is where the two models meet. It stores structured and unstructured data in one location, manages tables with Delta Lake for ACID transactions, schema enforcement and time travel, and lets data engineers work in Spark notebooks while analysts query the same tables in T-SQL. Files placed in the Tables folder are validated and registered automatically, and the automatically generated SQL analytics endpoint exposes only Delta tables, including those reached through shortcuts (Microsoft Learn, What is a lakehouse in Microsoft Fabric?, accessed 2026-10-20). Each Fabric engine reads the same Delta data, so a warehouse, a Spark notebook and a Power BI report in Direct Lake mode can all use one copy (Microsoft Learn, What is OneLake?, accessed 2026-10-20). EPC Group's Azure analytics architecture designs commonly land and refine data in lakehouses and serve curated star schemas from a warehouse in the same workspace domain.

Which one an organization needs

Microsoft's decision guide reduces the warehouse and lakehouse choice to three questions. If the team develops in Apache Spark, with Python, Scala, Spark SQL or R, start with a lakehouse; if it develops in T-SQL, start with a warehouse. If loads need multi-table transactions, choose the warehouse. If the data includes unstructured files, or the team is not sure, choose the lakehouse (Microsoft Learn, Microsoft Fabric decision guide: Choose between Warehouse and Lakehouse, accessed 2026-10-20). Microsoft adds that the choice is not final: either item can be added later as needs change, and both use the same SQL engine for T-SQL queries (Microsoft Learn, What is Fabric Data Warehouse?, accessed 2026-10-20).

In practice, EPC Group sees three common answers. Organizations with a mature SQL team and mostly structured line-of-business data start with a warehouse and add a lakehouse only for files, logs or data science. Organizations with large volumes of semi-structured or external data start with lakehouses arranged in bronze, silver and gold layers and serve the gold layer to Power BI. Organizations that already run Azure Data Lake Storage keep it as the raw landing zone and connect it to Fabric with shortcuts, so the lake, the lakehouse and the warehouse all read one copy. In each case the decision is recorded per data domain, not once for the whole company, so a finance warehouse and an operations lakehouse can sit side by side.

Where Synapse and Purview fit

Azure Synapse Analytics brought SQL pools, Spark pools and pipelines together in Azure, and Microsoft now documents a Fabric destination for each of those workloads, from the Migration Assistant for dedicated SQL pools to guides for Spark pools, notebooks, Spark job definitions, Hive Metastore metadata and pipelines (Microsoft Learn, Microsoft Fabric migration overview, accessed 2026-10-20). Governance spans all three models. Fabric provides discovery, domains, endorsement and lineage in the OneLake catalog, and Microsoft Purview adds sensitivity labels that persist through supported exports, data loss prevention for lakehouses, warehouses and semantic models, audit of every Fabric user activity and insider risk indicators (Microsoft Learn, Use Microsoft Purview to protect Microsoft Fabric, accessed 2026-10-20). EPC Group's Microsoft Purview practice sets one label taxonomy for the whole estate before data moves.

Documented work

ANON-CASES: EPC Group engagement data for data platform design covers mid-size and large healthcare, financial-services and manufacturing organizations choosing between warehouse, lake and lakehouse patterns on Azure and Microsoft Fabric, delivered by the firm's Microsoft Fabric consulting practice. Client names are not published here. Engagements are described by industry and organization size only, and no outcome figures are quoted that cannot be traced to a signed deliverable. EPC Group is a Microsoft Solutions Partner with all six Solutions Partner Designations as of October 2026.

Frequently Asked Questions

Talk to a senior architect

If your teams use the words warehouse, lake and lakehouse for the same storage account, or Synapse workloads need a plan into Fabric, a 30-minute discovery call with a senior EPC Group architect maps which data belongs in which model and which Microsoft service implements it. See the Microsoft Fabric consulting service hub for engagement structures.

Email contact@epcgroup.net or call 888-381-9725.

a Microsoft Gold Partner (2000 until Microsoft retired the program in 2022) — today holding all six Microsoft Solutions Partner Designations.