Dataflow Gen1 to Gen2: when to migrate and how

Migrate a Power BI Dataflow Gen1 to Gen2 only if its workspace already runs on a Fabric capacity: Gen2 doesn't work in Pro or Premium Per User workspaces. Gen1 still works but gets no new features. To move, use Save As for a safe side-by-side copy, or the preview Upgrade Wizard to replace it in place.

Dataflows are the Power Query part of the Power BI service: the same steps you build in Power BI Desktop, run in the cloud and shared between reports. Since Microsoft Fabric arrived there have been two generations, and Microsoft is now steering everyone with a capacity towards the second. This guide covers whether you need to move, what Gen2 gives you, what it costs in capacity, and how to migrate without breaking the reports that depend on your dataflows. Every behaviour below was checked against Microsoft Learn as of November 2026.

Is Dataflow Gen1 being retired?

No date has been set. Microsoft's migration guidance for Dataflow Gen1 (updated 25 October 2026) says Gen1 is "now in a legacy state and doesn't receive new feature investment". For Premium and Fabric customers, Gen2 is the recommended path. For Pro and Premium Per User customers, Microsoft says Gen1 "remains a supported option" and that it will share guidance as Gen2 options for those licences become available.

So there are two different answers depending on where your dataflows live:

• On a Fabric or Premium capacity: plan the move. Nothing will stop working tomorrow, but every new feature lands in Gen2, and the tooling to migrate is now in place.
• On Pro or Premium Per User only: stay where you are for now. Gen1 is still supported, and Gen2 isn't available in those workspaces at all.

Can you use Dataflow Gen2 on a Power BI Pro licence?

Not in a Pro workspace. Microsoft's Save As guide lists "a Fabric capacity (F2 or higher) or Fabric trial" as a prerequisite and states that "Power BI Pro and Power BI Premium Per User (PPU) workspaces don't support Dataflow Gen2". The Dataflow Gen2 overview adds Power BI Premium capacity to the list.

Licences work differently once you're on a capacity. Gen1 needs a Pro or PPU licence to publish or manage a dataflow. Gen2 needs only a free Fabric licence to author one in a capacity workspace. The licences of the people reading your reports don't change: on a capacity smaller than F64 they still need Pro, as our Power BI licensing guide explains.

The smallest capacity, an F2, costs about £231 a month pay-as-you-go or about £137 a month on a one-year reservation in UK South, as of November 2026 (Azure Retail Prices API, as used by our Power BI and Fabric pricing estimator). Microsoft Fabric pricing explained walks through the larger sizes. In our view, buying a capacity just to run dataflows rarely pays on its own; the case is stronger when you also want a lakehouse, a warehouse or Direct Lake reports.

What do you gain, and lose, by moving to Gen2?

Microsoft's comparison tables in the migration guidance and the overview list the differences. These are the rows that matter most for a small team:

Capability	Dataflow Gen1	Dataflow Gen2
Build transformations in Power Query	Yes	Yes
Write output to a lakehouse, warehouse, Azure SQL, SharePoint or others	No (stores its own data)	Yes, multiple destinations
Run as a step in a Fabric pipeline	No	Yes
Git integration and CI/CD	No	Yes, on by default for new items since April 2026
Monitoring hub and enhanced refresh history	No	Yes
Fast copy for large loads	No	Yes
Copilot for Data Factory	No	Yes
Incremental refresh	Yes	Yes (but it must be set up again)
DirectQuery to the dataflow	Yes	No
CDM folders, bring-your-own-lake, enhanced compute engine	Yes	No (OneLake and data destinations replace them)
AutoML	Deprecated	Not listed

The biggest practical change is the output. A Gen1 dataflow holds its results in its own storage, and reports read them through the dataflows connector. A Gen2 dataflow can write the results into a lakehouse or warehouse table, which notebooks, SQL and Direct Lake semantic models can then use. If you're moving towards a lakehouse, that's the point of the move.

The biggest loss is DirectQuery. If any report reads a dataflow in DirectQuery mode, it has to switch to Import or Direct Lake as part of the migration.

What does a Gen2 refresh cost in capacity?

Gen2 refreshes draw capacity units (CUs) from your Fabric capacity rather than adding a separate bill. Microsoft's Dataflow Gen2 pricing page (updated September 2026) sets the rates:

What runs	Rate
Standard compute, each query's first 10 minutes	12 CU per second
Standard compute, beyond 10 minutes	1.5 CU per second
Standard compute, older Gen2 items without CI/CD	16 CU per second
High-scale compute (only when staging is on)	6 CU per second
Fast copy	1.5 CU per second of core time

The pricing page says the rates are "subject to change at any time", and they don't apply to trial capacities.

To put that in scale: an F2 provides 2 CUs every second, which is 172,800 CU-seconds a day. Using the rates above, a query that runs for 5 minutes uses 300 × 12 = 3,600 CU-seconds, about 2% of an F2's day. A 20-minute query uses 600 × 12 plus 600 × 1.5, or 8,100 CU-seconds, about 5%. That's our arithmetic on Microsoft's published rates, and real refreshes vary. The upgrade guidance warns that CU consumption can differ from Gen1 and tells you to validate refresh duration and consumption afterwards. The Capacity Metrics app is the tool Microsoft points to for the real figures.

Save As or the Upgrade Wizard: which should you use?

There are four routes, and two of them do most of the work.

	Save As	Upgrade Wizard (preview)
What it does	Creates a new Dataflow Gen2 alongside the Gen1	Replaces the Gen1 in place
Original Gen1	Kept	Replaced, and can't be recovered
Dataflow ID	New, so consumers must be repointed	Kept
Schedule	Not documented for Gen1 copies; check after saving	Kept
Connections	Check each one after saving	Kept
Incremental refresh	Not copied	Not carried over
Reversible	Yes, delete the copy	No
Who can run it	Contributor or above on the target workspace	The dataflow's owner, with Contributor or above

Save As is the safe first step. In the workspace, select the ellipsis next to the dataflow, choose Save as Dataflow Gen2, name it and select Create. You get a Gen2 copy to test against while the original keeps serving your reports. Microsoft also exposes it through the REST API for bulk moves.

The Upgrade Wizard is the cut-over tool. Microsoft's Upgrade Wizard page says it preserves "each dataflow's ID, name, schedule, and connections", which is why most reports keep working afterwards. It can upgrade several dataflows from the same workspace in one pass and gives you a CSV report at the end. Because it can't be undone, Microsoft's own advice is to use Save As first if you want to keep the original.

The other two routes are manual: export a Power Query template (.pqt) from the Gen1 editor and import it into a new Gen2, or copy the queries and paste them in. Microsoft documents both on its move queries page. They're useful when you want to rebuild rather than lift and shift.

How do you migrate a dataflow step by step?

This is the order we'd follow for a small team, built on Microsoft's documented prerequisites and limitations:

1. Inventory the consumers. List every semantic model, report, Excel workbook and downstream dataflow that reads each dataflow. The wizard checks items in Fabric, but Microsoft notes it can't see .pbix or .xlsx files outside Fabric that use the legacy connector, so you have to find those yourself.
2. Check the blockers. The wizard flags dataflows that need attention: DirectQuery consumers, bring-your-own-lake storage, incremental refresh, linked entities, more than 50 enabled queries, and name clashes. It won't run at all if you aren't the owner or a data source connection is unresolved.
3. Back up anything you can't reload. A Gen1 dataflow's stored data is a cache, and Microsoft says it isn't available after the upgrade. If a source only keeps recent history, copy the data to a lakehouse first.
4. Test with Save As. Make a Gen2 copy in a test workspace, refresh it, compare the output with the original and check its CU use in the Capacity Metrics app.
5. Upgrade or repoint. Either run the Upgrade Wizard on the original, or keep the Save As copy and repoint each consumer to its new ID. Don't refresh a dataflow while it's being upgraded, and if an upgrade fails, the wizard restores the Gen1 but you can't retry for 24 hours.
6. Finish the job. The upgrade doesn't refresh the new dataflow, so run each one. Open and save any downstream dataflows so they rebind, otherwise they keep returning stale data. Set up incremental refresh again on a destination that supports it.

What breaks after an upgrade?

Most things carry on, but Microsoft lists four catches worth knowing before you start:

• The legacy connector. Anything that reads the dataflow through the old Power BI Dataflows connector fails on its next refresh. It needs switching to the Power Platform Dataflows connector, which adds a "Workspaces" navigation step; the workspace ID, dataflow ID and table names stay the same, but you have to sign in again.
• Workspace Viewers. Viewers can't read tables from an upgraded dataflow through the Power Platform Dataflows connector; they need Contributor, Member or Admin.
• REST API scripts. Gen2 uses the Fabric REST API, which doesn't fully match the Power BI REST API, so any automation that manages dataflows needs updating.
• Workspace security settings. The upgraded Gen1 is kept soft-deleted for up to 30 days, and until then you can't turn on outbound access protection for that workspace.

What should a Pro-only team do now?

Keep using Gen1 and keep your Power Query logic tidy, because that's the part that moves across unchanged whichever route you take later. Watch for Microsoft's promised Pro guidance: our FabCon Europe 2026 round-up covers what was announced for Pro users this autumn. If you're weighing a capacity for other reasons, cost it with the pricing estimator first.

If you'd rather someone senior planned the migration, mapped what reads each dataflow and moved it without a broken Monday report, that's the pipeline work our data engineering team does, alongside our Power BI and Fabric consultancy. See all our data services.

Frequently asked questions

Is Power BI Dataflow Gen1 being retired?

Not as of November 2026. Microsoft Learn says Gen1 is in a legacy state: still supported, but it receives no new feature investment. No end-of-support date has been published, and Microsoft says Pro and Premium Per User customers can keep using Gen1 while guidance for them evolves.

Can I use Dataflow Gen2 with a Power BI Pro licence?

Not in a Pro or Premium Per User workspace. Dataflow Gen2 needs a workspace on a Fabric capacity (F2 or higher), a Fabric trial or a Power BI Premium capacity. Inside such a workspace, a free Fabric licence is enough to author a Gen2 dataflow.

What is the difference between Save As and the Upgrade Wizard?

Save As makes a new Dataflow Gen2 next to the original, so the Gen1 stays untouched but the copy gets a new ID that consumers must be repointed to. The Upgrade Wizard, in preview, replaces the Gen1 in place, keeping its ID, name, schedule and connections, and cannot be reversed.

Will my Power BI reports break when I upgrade a dataflow?

Usually not. Microsoft says most downstream semantic models and reports keep working. The exceptions are items using the legacy Power BI Dataflows connector, which fail on their next refresh, DirectQuery consumers, and downstream dataflows, which must be opened and saved to rebind.

Does Dataflow Gen2 cost more to run than Gen1?

It can do either. Gen2 refreshes consume capacity units from your Fabric capacity, at 12 CU per second for each query's first ten minutes and 1.5 CU per second after that. Microsoft says consumption differs between the two, so test a copy and check the Capacity Metrics app before switching.

---

Easy Insight is a UK consultancy for AI, web, apps and data — senior specialists only, no juniors.