What is OneLake? The Tenant-Wide Storage Layer Inside Microsoft Fabric

OneLake is Microsoft Fabric's tenant-wide unified storage layer, built on ADLS Gen2, with a single namespace and shortcut features that let Fabric reference data in S3, ADLS, or Dataverse without copying.

OneLake is Microsoft Fabric's tenant-wide unified storage layer, built on Azure Data Lake Storage Gen2 under the hood and addressable via a single tenant-wide namespace. It launched with Fabric in May 2023 and is the storage primitive every Fabric workload reads from and writes to.

What OneLake actually is

OneLake is three things at once:

• A logical namespace, one per Microsoft tenant, where every Fabric workspace lives under a single root path.
• A physical storage layer built on ADLS Gen2, using Delta-Parquet as the default file format and inheriting ADLS access controls underneath.
• A shortcut system that lets a OneLake path point at data in S3, ADLS, Dataverse, or another OneLake region without copying, so a single OneLake path can mirror external data on read.

The shortcut feature is the part most teams underestimate. It is the difference between "we need a daily Data Factory pipeline to land S3 data" and "we point a OneLake shortcut at the S3 bucket and query it directly." Egress costs and read latency still apply, but the operational complexity drops by an order of magnitude for read-mostly workloads.

What people confuse it with

• "OneLake is just rebranded ADLS." Half-true. OneLake sits on top of ADLS Gen2, but the tenant-wide namespace, shortcut model, and Fabric-aware access controls are net-new.
• "OneLake replaces our data lake." Eventually, for Fabric-native workloads, yes. For non-Fabric Azure workloads (Azure ML, Azure Databricks running outside Fabric, custom Azure services), ADLS containers are still the right primitive. Most organizations run both during the transition.
• "OneLake is Iceberg-native." No. OneLake is Delta-Parquet native. Microsoft has shipped read support for external Iceberg tables via the Mirroring feature in 2026, but writes still target Delta.

When OneLake matters

OneLake matters when:

• The organization is building or migrating to Fabric and needs a coherent storage layer across Lakehouse, Warehouse, and Direct Lake semantic models.
• The data architecture includes inbound data from S3 or Dataverse that the team would otherwise have to copy or schedule pipelines for.
• Workspace partitioning is the operational question, and a single tenant-wide namespace simplifies the security model.

When OneLake does not help

OneLake does not help when:

• The organization has no Fabric footprint and is not planning one.
• The data layer is Iceberg-first (Snowflake polaris, Databricks Unity Catalog with Iceberg) and Delta interop is a constraint.
• Data residency requirements exclude the Microsoft cloud regions OneLake supports.

How Thinklytics works on OneLake architecture

OneLake namespace strategy, shortcut design, and workspace partitioning come up in every Fabric assessment we run. The decisions on day one shape what the platform looks like at year three. See Microsoft Fabric data engineering for the architectural pattern we recommend and Microsoft Fabric consulting for the procurement-side scope.

Frequently asked questions

What is OneLake in one sentence?

OneLake is Microsoft Fabric's tenant-wide unified storage layer, built on ADLS Gen2 under the hood and addressable via a single tenant-wide namespace, with a shortcut feature that mirrors data from S3, ADLS, or Dataverse without copying.

Is OneLake different from ADLS Gen2?

OneLake uses ADLS Gen2 storage primitives but presents them as one logical namespace per Microsoft tenant. ADLS containers are still there under the hood, but the access pattern, security model, and shortcut feature are Fabric-specific. Think of OneLake as ADLS with a tenant-wide front door.

What file format does OneLake use?

Delta-Parquet by default. OneLake stores tabular data as Delta Lake tables (Parquet files with a transaction log). It supports reading external Apache Iceberg tables via Fabric Mirroring in 2026, but writes still target Delta. Teams that need true Iceberg portability across Snowflake and Databricks typically standardize on external Iceberg tables.

What are OneLake shortcuts?

Shortcuts are pointers from one OneLake path to another storage location (S3, ADLS, Dataverse, another OneLake region). The shortcut presents the external data as if it lived in OneLake, with no copy and no scheduled refresh. Egress costs and latency apply when querying through the shortcut.

Do I still need ADLS containers?

For Fabric-native workloads, no. OneLake replaces the per-workspace ADLS containers Synapse used. For non-Fabric Azure workloads (Azure ML, Azure Databricks, custom Azure services), ADLS containers are still the right primitive. Most organizations end up running both during the transition period.

How is OneLake billed?

OneLake storage is billed separately from Fabric capacity. Storage is roughly $0.026 per GB per month for hot data in 2026 (similar to ADLS Gen2 pricing). Caching for Direct Lake mode is included in the Fabric capacity cost. Shortcut traffic does not add storage cost but can add egress cost when crossing clouds or regions.

What workspaces structure should I use in OneLake?

One workspace per logical domain (Sales, Finance, Operations, Engineering, etc.), with workspace-level permissions held by the domain owner. The default of one workspace per team creates an access-control mess at 8+ workspaces. The medallion layout (bronze, silver, gold) typically lives as folder prefixes within a workspace, not as separate workspaces.

How does Thinklytics work on OneLake architecture?

OneLake namespace strategy, shortcut design, and workspace partitioning are part of our 4-week Fabric current-state assessment. See Microsoft Fabric data engineering for the full architectural pattern we recommend.