📊 Series: Modern Data Platforms Part 5 of 10

"Enterprise analytics does not have one winning platform. It has many platforms serving different needs, different teams, and different historical decisions. The question is never 'which one should we use' — it is 'how do we make them work together and govern the result.'"

The Four Major Platforms — What Each Actually Is

PLATFORM POSITIONING MAP

                    ┌──────────────────────────────────────────┐
                    │         UNIFIED PLATFORMS                │
                    │  (One platform for all analytics tasks)  │
                    │                                          │
                    │   Microsoft Fabric    Snowflake          │
                    │   (Microsoft stack)   (Cloud-native DWH) │
                    └──────────────────────────────────────────┘

                    ┌──────────────────────────────────────────┐
                    │         SPECIALISED PLATFORMS            │
                    │  (Excellent at specific workloads)       │
                    │                                          │
                    │   Azure Synapse       Databricks         │
                    │   (SQL+Spark hybrid)  (Spark+ML first)   │
                    └──────────────────────────────────────────┘

None of these is "the best." Each represents a different set of design decisions optimised for different scenarios. Understanding those decisions is what helps you choose correctly — or recognise why your organisation already has multiple.

Microsoft Fabric — Unified SaaS Platform for Microsoft Ecosystem

What it is: A unified analytics platform that brings together data engineering, data warehousing, data science, real-time analytics, and Power BI into a single SaaS offering on OneLake.

Dimension
Detail
Core strength
Everything in one platform — no data movement between tools. Power BI, Spark, SQL, ML all read the same OneLake data.
Best fit
Organisations already on Microsoft 365/Azure, Power BI-heavy teams, organisations wanting to consolidate platforms.
Licensing
Fabric capacity (F-SKUs) — one licence covers all workloads. Works alongside M365.
Limitation
Newer platform (GA 2023) — some enterprise features still maturing. Spark experience less mature than Databricks.

Azure Synapse Analytics — SQL Warehouse + Spark, Tightly Azure-Integrated

What it is: A hybrid platform combining a dedicated SQL data warehouse (MPP architecture) with Apache Spark, tightly integrated with Azure services.

Dimension
Detail
Core strength
Dedicated SQL pool for very large structured data warehouses (petabyte scale). Deep Azure integration (Azure ML, Purview, Defender).
Best fit
Enterprises with large existing SQL Server workloads, regulated industries needing deep Azure security controls, organisations requiring dedicated (not serverless) compute SLAs.
Licensing
Pay-per-use for serverless; reserved DWU capacity for dedicated pools.
Limitation
More complex to set up and govern than Fabric. Microsoft is shifting investment toward Fabric — Synapse remains supported but new features land in Fabric first.
Direction of travel Microsoft's announced strategy is that Fabric is the future unified platform. Synapse Analytics will continue to receive support and security updates, but new capabilities — especially around unified governance and AI integration — are being built in Fabric. For new projects, Fabric is the default Microsoft recommendation.

Databricks — Spark-First Platform, ML and Data Science Leader

What it is: A cloud-native unified data and AI platform built on Apache Spark. Created by the original Apache Spark and Delta Lake developers. Available on Azure (Azure Databricks), AWS, and GCP.

Dimension
Detail
Core strength
Best-in-class Spark execution engine (Photon). Most mature Delta Lake implementation. Strong ML lifecycle management (MLflow). Multi-cloud.
Best fit
Data science teams with Python/Scala expertise, ML-heavy workloads, organisations needing multi-cloud flexibility, large-scale ETL with complex transformations.
Licensing
DBU (Databricks Unit) based pricing on top of cloud compute. Can be expensive at scale without careful management.
Limitation
Not a BI platform — needs Power BI or another tool for reporting. Not natively integrated with Microsoft 365. Higher skill requirement (Spark expertise).

Snowflake — Cloud-Native SQL Data Warehouse, Any Cloud

What it is: A cloud-native data warehouse built from scratch for the cloud. Runs on Azure, AWS, and GCP. Pure SQL focus with exceptional multi-cloud data sharing capabilities.

Dimension
Detail
Core strength
Automatic query optimisation. Unique data sharing (share live data with external partners without copying). True multi-cloud. Near-zero administration.
Best fit
Organisations that need to share data with external partners, multi-cloud organisations, teams that want SQL simplicity without infrastructure management, data marketplaces.
Licensing
Credit-based consumption pricing. Separate from Microsoft ecosystem — no M365 licence synergy.
Limitation
Additional cost on top of Azure spend. No native Power BI integration (connector available but not seamless). Spark/ML requires additional tooling.

Side-by-Side Comparison

Criterion
Fabric
Synapse
Databricks
Snowflake
Primary workload
Unified (all)
SQL DW + Spark
Spark + ML
SQL DW
Power BI integration
Native (Direct Lake)
Good (connector)
Connector only
Connector only
ML / Data Science
Good (improving)
Moderate
Best-in-class
Limited
Multi-cloud
Azure only
Azure only
Yes
Yes
Governance with M365
Native Purview
Native Purview
Requires setup
Requires setup
Skill requirement
Low-medium
Medium-high
High (Spark)
Low (SQL)
Admin overhead
Low (SaaS)
Medium-high
Medium
Very low

The Reality of Multi-Platform Enterprises

Large enterprises rarely run on a single platform. A common real-world combination:

TYPICAL ENTERPRISE MULTI-PLATFORM SETUP

Data Engineering team → Databricks (Spark expertise, ML pipelines)
                              ↓ writes Delta tables to ADLS Gen2
Business Intelligence team → Microsoft Fabric (Power BI + Direct Lake)
                              reads from same ADLS via OneLake shortcuts
SQL/Finance team → Azure Synapse serverless (ad-hoc SQL on lake)
External data sharing → Snowflake (partner data exchange)

Each tool doing what it does best.
Governance challenge: Microsoft Purview spans all four.
📌 Governance is the hard part Having four platforms is not the problem. Having four platforms with four separate access control systems, four data catalogues, and no unified lineage — that is the problem. Microsoft Purview is designed to govern across all of these platforms. Part 8 covers multi-platform governance in depth.

Common Misconceptions

  • "We should pick one platform and standardise everything on it" — Rarely practical. Different teams have different skills, different workloads have different needs, and historical investments cannot be abandoned overnight. The realistic goal is minimising unnecessary platform sprawl and governing the platforms you have well.
  • "Fabric replaces Databricks" — Not for data science/ML-heavy organisations. Fabric's Spark is capable and improving, but Databricks has a decade of Spark optimisation and MLflow maturity. Many enterprises run Databricks for data engineering and ML, and Fabric for BI — the platforms complement each other via OneLake shortcuts.
  • "Snowflake has no place in a Microsoft Azure environment" — Snowflake's data sharing capability (sharing live data with external partners without data movement) has no equivalent in Azure currently. Organisations with heavy partner data exchange requirements legitimately need Snowflake alongside Azure services.
  • "Azure Synapse is being retired" — Not retired — it continues to be supported and funded. Microsoft is directing new feature development to Fabric, but existing Synapse workloads are not being forced to migrate.

🎯 Quick Check

Q1: A pharmaceutical company has a Python data science team running complex ML pipelines on patient data, and a separate Power BI team building compliance reports. Which platform combination makes sense and why?

Show Answer

Azure Databricks for data science + Microsoft Fabric for BI. Databricks serves the Python/ML team with best-in-class Spark execution and MLflow model management. Fabric serves the BI team with Power BI Direct Lake for compliance reports. The bridge: Databricks writes Delta tables to ADLS Gen2, Fabric reads the same data via OneLake shortcuts — no data duplication. Both teams use the same underlying data in their preferred tool.

Q2: A global manufacturing company needs to share real-time inventory data with 50 supplier partners worldwide — each on different cloud providers. Which platform feature addresses this requirement?

Show Answer

Snowflake's data sharing capability. Snowflake allows sharing live data with external organisations (even on different cloud providers) without physically copying the data. The supplier sees live inventory as it updates. Azure alternatives exist (Azure Data Share, Fabric shortcuts) but Snowflake's cross-cloud, zero-copy sharing model is uniquely suited for multi-partner, multi-cloud data exchange scenarios.

Q3: An organisation is choosing between Fabric and Synapse Analytics for a new project. They are a Microsoft 365 E5 shop with an existing Power BI Premium capacity, and the primary workload is BI reporting with some data engineering. Which should they choose and why?

Show Answer

Microsoft Fabric. Three reasons: (1) Direct Lake integration with Power BI is native in Fabric — no connector overhead. (2) Microsoft's stated direction is Fabric for new projects — new governance and AI features land here first. (3) Their existing Power BI Premium capacity can be converted to Fabric capacity, avoiding additional licensing. For a primarily BI-focused workload in a Microsoft ecosystem, Fabric is the natural fit.

Key Takeaways — Part 5

  • Microsoft Fabric, Synapse, Databricks, and Snowflake each optimise for different workloads — there is no universally "best" platform
  • Microsoft Fabric is the recommended default for Microsoft-ecosystem organisations: unified SaaS, native Power BI Direct Lake, Purview governance
  • Databricks leads for Spark/ML-heavy workloads — many enterprises use both Databricks and Fabric together via OneLake shortcuts
  • Snowflake's unique strength is multi-cloud data sharing with external partners — a use case other platforms handle less elegantly
  • Multi-platform reality requires unified governance — Microsoft Purview is designed to govern across all four platforms from one control plane
📚 Continue the Series