There is no Master Data Services in Microsoft Fabric, and no migration path from it. But “Fabric has nothing” is no longer accurate, and if you read that somewhere in the last year it is worth updating. PowerTable, a component of the Plan item, is described by Microsoft as a no-code reference and master data management platform: an Excel-like grid over a table, with write-back, approval workflows, change history, and row- and column-level security. That is a real stewardship surface, and it did not exist before 2026.
What it is not is MDS. There is no model-and-entity structure, no domain-based attributes, no derived hierarchies, and no governed versions you can lock and commit. It is also in preview, and it arrives inside an enterprise planning and budgeting product rather than as a master data tool in its own right. So the honest answer to “is MDS in Fabric” is still no — but the gap is now shaped differently than it was.
The confusion is understandable. MDS shipped inside SQL Server. Fabric is where Microsoft's data story now lives. It is a reasonable assumption that the thing in the old box turns up in the new box. It did not, and what did turn up came in through a side door marked “planning”.
What Fabric gives you, and what it leaves out
Fabric is genuinely good at the analytical half of the problem. If your master data already exists and is already correct, Fabric will happily store it, join it, model it, and put it in front of Power BI. The gap is everything that happens before that: deciding what a valid cost centre looks like, letting a human create one, stopping them from creating a duplicate, and recording who approved it.
Purview is not the answer either
When MDS teams ask Microsoft what to use, they are usually pointed at Purview. It is worth understanding why that answer disappoints almost everyone who follows it.
Purview is a catalog. It scans systems, classifies what it finds, tracks lineage, and tells you where your data lives and who owns it. MDS was the opposite kind of tool: an authoring system. Someone opened it and typed a value, and that value became the truth other systems consumed. A catalog documents the cost centre list. It does not give anyone a place to add a cost centre.
Microsoft's own documentation reflects this — it positions Purview alongside partner MDM products rather than as a replacement for one. That is an honest position, and it is worth reading it literally rather than hopefully. We went deeper on the distinction in MDM is not a data catalog.
How far PowerTable actually gets you
Worth being concrete, because this is the part most write-ups have not caught up with. PowerTable covers the day-to-day stewardship loop properly: a business user opens a grid, edits records the way they would in Excel, the change is written back to the source, an approver gets a notification in email or Teams, and the whole thing lands in a change log. Add row- and column-level security and Type II slowly changing dimensions and you have a genuine governed editing surface. If your master data problem is “a few reference tables need a controlled way to be updated”, that may be the whole answer.
Where it stops is structure. MDS gave you a model containing entities, attributes typed as domain-based references to other entities, explicit and derived hierarchies built from those references, and versions you could lock and commit. PowerTable gives you tables with an excellent front end. Rebuilding a cost centre hierarchy or a set of cascading domain lookups is work you do yourself, in the semantic layer, and maintain by hand.
Two practical caveats. It is in preview, so it is not a thing to bet a governance programme on this quarter. And it is a component of Plan, an enterprise planning and budgeting product — which shapes the roadmap. Reference data management is a listed use case, not the reason the product exists. Whether that matters depends on how central master data is to what you are doing.
The three routes teams actually take
Compose it inside the Microsoft stack. Either PowerTable as above, or the older pattern: Dataverse tables as entities, a Power App as the editing screen, Power Automate for validation and approval routing, output landing in a lakehouse. PowerTable is the shorter path if your data already lives in Fabric; Dataverse still wins if you need custom app logic around the editing. For a handful of flat code lists either genuinely works. What you are buying is ownership: hierarchies, versioning, per-attribute permissions, and audit are now yours to build and maintain forever. Teams tend to be happy with this until the second hierarchy arrives.
Buy enterprise MDM. Informatica, Profisee, Semarchy, Reltio. You get fuzzy matching, survivorship rules, and golden-record management, which are real capabilities MDS never had. You also get enterprise pricing and an implementation measured in quarters. If your problem is genuinely customer-360 deduplication across a dozen source systems, this is the right shape of tool. If your problem is that four hundred cost centres need an owner, it is considerable overkill.
Keep authoring on SQL Server and feed Fabric from it. The part of MDS that worked was that master data lived in a SQL Server database your team already ran, backed up, and queried. Nothing about Fabric requires you to give that up. You run a dedicated master data tool against SQL Server, your stewards work there, and Fabric reads the result through a shortcut or a pipeline like any other source. Authoring stays where the governance is; analytics stays where the analytics are.
Where Primentra fits
Primentra is the third route. It runs on your SQL Server, gives data stewards a browser grid that behaves like a spreadsheet, and keeps the MDS concepts your models were built on — models, entities, attributes, domain-based attributes, hierarchies — so the migration is a mapping exercise rather than a redesign.
For Fabric specifically, the integration is unremarkable in the best way: your master data sits in SQL Server tables and views, and Fabric consumes them the way it consumes any other SQL source. No connector to license, no sync to monitor. If you are mapping out the move, the migration guide covers the sequence, and what SQL Server teams are doing instead covers the wider option set.
Frequently asked questions
Is there Master Data Services in Microsoft Fabric?
Not as such, and there is no migration path from MDS. But Fabric is no longer empty here: PowerTable, part of the Plan preview, is a no-code reference and master data management grid with write-back, approval workflows, audit logs, and row- and column-level security. What it lacks is MDS-style entity modelling, domain-based attributes, derived hierarchies, and governed versions — and it arrives inside a planning and budgeting product rather than as a dedicated MDM tool.
Can Microsoft Purview replace Master Data Services?
No. Purview catalogs and governs data that lives elsewhere; MDS was an authoring system where people created the official record. A catalog tells you the cost centre list exists and who owns it. It gives nobody a screen to add one. Microsoft positions Purview alongside partner MDM products, not as a replacement.
Can I rebuild MDS with Dataverse and Power Platform?
Partly. Dataverse holds entities, Power Apps gives an editing screen, Power Automate handles validation and approvals. It works for simple flat code lists. Derived hierarchies, versioning, per-attribute permissions and audit are yours to build and maintain — which is where most of these projects get expensive.
How do I get master data into Fabric?
Keep authoring it in a governed tool and let Fabric read the result. If your master data lives in SQL Server, Fabric consumes it through a shortcut, mirroring, or a pipeline like any other SQL source. The authoring surface and the analytics surface do not need to be the same product.
Keep the authoring. Let Fabric do the analytics.
Primentra runs on the SQL Server you already have, keeps your MDS model structure, and hands Fabric clean tables to read. Import your MDS models into the 60-day trial and see how much of your model maps straight across.