Microsoft OneLake and Fabric as a Data Platform

Data-platform angle only. The product overview, licensing and BI/Copilot features are in Microsoft Fabric; terminology in microsoft-fabric-terminology-guide; BI in Power BI.

What it is

OneLake is the single logical data lake every Fabric tenant gets automatically: one instance per tenant, cannot be deleted or duplicated, built on Azure Data Lake Storage, storing tables as Delta Parquet or Iceberg.

Maker, ownership and history

Microsoft. Fabric is the successor of the Synapse/Power BI stack; no acquisition history verified.

Editions/deployment

SaaS only, billed by Fabric capacity per region; workspaces belong to a capacity. Pricing: https://azure.microsoft.com/pricing/details/microsoft-fabric/ (not opened).

Core architecture

• Hierarchy: tenant, workspace, items (lakehouse, warehouse, eventhouse). Domains give federated ownership.
• All engines (T-SQL, Spark, Analysis Services/Direct Lake, KQL) read one copy in Delta Parquet. Access via ADLS Gen2 APIs, Windows file explorer, Azure Databricks, Foundry.
• Metadata virtualization: Iceberg tables readable as Delta in Fabric, Delta readable by Iceberg clients such as Snowflake.
• Shortcuts (GA): pointers to OneLake, ADLS, Blob, S3 and S3-compatible, Iceberg sources, Dataverse, on-prem; shortcut transformations can convert format or remove PII.
• Mirroring: database mirroring replicates operational DBs into Delta; metadata mirroring syncs an external catalog and reads data in place.
• OneLake security roles enforce folder/table/row/column rules across SQL, Spark and Power BI; sensitivity labels and DLP apply.

Role in an enterprise AI rollout

Gives agents and Copilot one governed estate across clouds without copies; ontology gives them business meaning (ontology); Purview audits AI interactions.

AI-related features as of September/October 2026 (Microsoft Learn “What’s new”)

• Fabric IQ workload (Preview): ontology item (entities, relationships, constraints), ontology as data-agent context (Preview), ontology as Real-Time Dashboard source (Preview); Native Planning Engine (Preview). Fabric IQ in Microsoft 365 Copilot: GA.
• OneLake catalog search API, MCP tool and fab find CLI (Preview); table discovery (Preview); OneLake table read API for Delta/Iceberg (Preview); centralized governance experience (Preview).
• Purview: DSPM for AI covering Fabric Copilots and data agents (Preview).
• Mirroring previews: AWS Glue Iceberg, Google Lakehouse Runtime Catalog, Dremio catalog, Azure Database for MySQL, Azure Monitor, Snowflake security roles; Unity Catalog in Azure Databricks can query OneLake via catalog federation.

Strengths and weaknesses vs peers (opinion)

• Strong: zero-copy shortcuts and mirroring across clouds, tight Power BI/Microsoft 365 reach.
• Weaker: many features still Preview; capacity-based billing makes cost attribution harder; SaaS-only.

Self-learning

• OneLake overview: https://learn.microsoft.com/en-us/fabric/onelake/onelake-overview (updated 2026-10-02)
• Microsoft Certified: Fabric Data Engineer Associate (exam DP-700, paid exam; English version updates 2026-10-19; free practice assessment): https://learn.microsoft.com/en-us/credentials/certifications/fabric-data-engineer-associate/
• What’s new: https://learn.microsoft.com/en-us/fabric/fundamentals/whats-new

Sources (fetched 2026-10-07)

The three Microsoft Learn URLs above.

Open items

• Purview product details beyond the DSPM preview not covered; no vault Purview note exists.
• Fabric pricing page not opened.