CSA Loom vs Microsoft Fabric — feature-by-feature¶
Comparative positioning note
This document is written from the perspective of Microsoft Azure, Cloud Scale Analytics, and CSA Loom. Any description of third-party or competing products, services, pricing, or capabilities is derived from publicly available documentation and sources believed accurate at the time of writing, and is provided for general comparison only. We do not claim expertise in, or authority over, any non-Microsoft product or service; the respective vendor's official documentation is the authoritative source for their offerings, which may change over time. Nothing here is intended to disparage any vendor — where a competing product has genuine advantages, we aim to note them honestly. Verify all third-party details against the vendor's current official documentation before making decisions.
Last updated: 2026-05-22
CSA Loom is a productized Azure-native parity layer for Microsoft Fabric, deployable into any Azure tenant where Fabric isn't yet generally available. This page is the side-by-side capability comparison.
For the older / broader csa-inabox vs Fabric comparison, see CSA-in-a-Box vs Microsoft Fabric. This page is CSA Loom-specific — Loom is the productized SaaS-feel form of csa-inabox.
Positioning at a glance¶
| CSA Loom | Microsoft Fabric | |
|---|---|---|
| Shape | Productized + customer-deployable parity layer on Azure-native + open source | Unified SaaS analytics platform |
| Where it ships | Commercial + GCC + GCC-High + IL5 (v1.1) | Azure Commercial GA; Gov Forecasted (no public date) |
| Audit boundaries | FedRAMP High + DoD IL4 + IL5 + ITAR-eligible | FedRAMP High + DoD IL2 (Commercial baseline) |
| Compute model | Hybrid Databricks + Synapse Serverless + ADX + Power BI Premium | F-SKU capacity (single CU pool) |
| Pricing | Free in v1 (Azure consumption only) | F-SKU + overage |
| Vendor coupling | Microsoft 1P + open-source under the covers | Microsoft 1P SaaS |
| Best fit | Federal / DoD / regulated commercial blocked from Fabric | Commercial; wants managed SaaS simplicity |
Per-capability comparison¶
See the full CSA Loom parity matrix for the workload-by-workload table. Headline differences:
Storage + namespace¶
| Capability | CSA Loom | Fabric |
|---|---|---|
| Storage primitive | ADLS Gen2 + unified path tree | OneLake (single tenant-wide namespace) |
| Shortcuts | Loom Shortcuts service (cross-cloud, ADLS) | OneLake shortcuts (ADLS, S3, GCS) |
| RBAC enforcement | Engine-layer (UC / Purview / Synapse / Power BI / ADX) | Storage-protocol-layer (OneLake Security) |
| Cross-engine | UC Iceberg REST + Delta UniForm (Commercial); manual (Gov) | OneLake native |
Direct Lake (the headline difference)¶
| Capability | CSA Loom | Fabric |
|---|---|---|
| Engine | Power BI Premium Import + warm-cache materializer | VertiPaq direct over Delta-Parquet |
| Freshness | 5-30 seconds (partition refresh) | Sub-second |
| Storage mode | Import (with Direct-Lake-Shim refresh) | Direct Lake (native) |
| F-SKU required | Yes (in GCC-H/IL5); no parity in GCC | Yes (F-SKU only) |
Honest gap: sub-second freshness is not achievable in Loom. See Direct Lake parity workload page for the engineering rationale.
Data Activator / Reflex¶
| Capability | CSA Loom Activator Engine | Fabric Reflex |
|---|---|---|
| Engine | ADX + NRules + Redis + Function dispatcher | KQL/Eventhouse + per-object state engine |
| Rule primitives | All 8 Fabric primitives implemented | Native |
| End-to-end latency | 5-30 s | 5-30 s |
| Action surface | Teams / Email / Power Automate / Logic App / Databricks Job / ADF / UDF / webhook | Same surface + native Fabric items |
Mirroring (zero-ETL CDC)¶
| Capability | CSA Loom Mirroring Engine | Fabric Mirroring |
|---|---|---|
| Sources | Azure SQL / Postgres / MySQL / Cosmos / Snowflake / Oracle / SAP (Open Mirroring) | Same sources native |
| Mechanism | OSS Debezium + Spark Structured Streaming + Delta MERGE | First-party |
| Latency | Sub-minute steady-state | Sub-minute steady-state |
| First-touch UX | Templated configs + CLI (v1); Console UI (v1.1) | Click-to-mirror native |
Data Agents¶
| Capability | CSA Loom Data Agents | Fabric Data Agents |
|---|---|---|
| Foundation | Extends apps/copilot/ + azure-functions/copilot-chat/ | Azure OpenAI Assistant API |
| Tools | NL2SQL + NL2DAX + NL2KQL + Microsoft Graph + custom AI Search | Same |
| Identity passthrough | OBO via MSAL BFF | Native (Entra Agent ID) |
| Up to 5 data sources per agent | Yes | Yes |
| Foundry integration | Commercial only (when Foundry Agent Service available) | Native |
Compliance + Sovereignty¶
| Boundary | CSA Loom | Microsoft Fabric |
|---|---|---|
| Azure Commercial GA | ✅ | ✅ |
| Azure Government GCC | ✅ | Forecasted |
| Azure Government GCC-High / IL4 | ✅ | Forecasted |
| Azure Government DoD IL5 | ✅ v1.1 | Forecasted |
| ITAR (in GCC-High) | ✅ | ❌ |
| HIPAA BAA | ✅ all boundaries | ✅ Commercial |
| CMMC L2 / L3 | ✅ | Commercial only |
This is the headline differentiator. Fabric is Forecasted across all four Gov boundaries today. Federal customers cannot adopt Fabric — and CSA Loom is what they should deploy.
When to choose which¶
Choose Microsoft Fabric if¶
- You're on Azure Commercial
- Power BI is central to your delivery
- You want the simplest possible operational footprint (single CU pool, one workspace plane)
- You don't need fine-grained Bicep / per-resource control
- You're comfortable with Microsoft-SaaS coupling
Choose CSA Loom if¶
- You're on Azure Government (or migrating to it)
- ITAR / CMMC L2/L3 requirements
- IL5 audit scope (v1.1)
- You have existing Synapse / Databricks investment to evolve
- Spark / ML workloads are heavy enough to want Photon + UC
- You need per-DLZ subscription isolation (multi-domain federal)
- You want
publicNetworkAccess = disabledeverywhere via Bicep
Choose BOTH (hybrid)¶
- Federal customer with both Commercial + Gov estates
- Pattern: Fabric in Commercial for public data; Loom in Gov for CUI / classified
- See Hybrid topology use case
Forward migration¶
When Microsoft Fabric reaches your audit boundary:
| CSA Loom artifact | Fabric equivalent | Migration mechanism |
|---|---|---|
| Delta tables | OneLake Delta tables | Zero data movement (OneLake shortcut) |
| dbt models | dbt in Fabric Data Factory | Low (dbt-fabric adapter) |
| Notebooks | Fabric Spark notebooks | Medium (runtime swap) |
| TMDL semantic models | Direct Lake on OneLake | Medium (re-author) |
| ADX databases / KQL | Fabric Eventhouse | Low (same engine) |
| Activator rules | Fabric Reflex | Low-Medium (JSON port) |
| Data Agents | Fabric Data Agents | Low (config export/import) |
| Purview catalog | Fabric Purview | Zero (same engine) |
See Forward to Fabric runbook.
Related¶
- CSA Loom — the pillar
- CSA Loom — parity matrix (workload-by-workload)
- CSA Loom — reference architecture
- CSA-in-a-Box vs Microsoft Fabric (legacy comparison)
- Microsoft Fabric in Azure Government — availability
- ADR:
fiab-0012— Forward migration
See also¶
- ← Previous: Microsoft Fabric Platform Guide
- → Next: CSA Loom — overview
- ⌂ Index: Documentation home