Dataflow Gen1 to Gen2: Using the New Upgrade Wizard

Navy Solv Systems title card reading 'Dataflow Gen1 to Gen2 Upgrade' with a loop motif.

In Short: Upgrading Dataflow Gen1 to Gen2 With the New Wizard

Microsoft now offers an in-place upgrade from Power BI Dataflow Gen1 to Dataflow Gen2 (CI/CD). The Dataflows Upgrade Wizard replaces each Gen1 dataflow with a Gen2 (CI/CD) item that keeps the same ID, name, schedule and connections. For most dataflows, that means downstream semantic models and reports keep working without being repointed.

Three caveats come first. The wizard is in preview, it only works in workspaces assigned to Fabric capacity, and the upgrade cannot be undone. It removes most of the rebuild work that made Gen1 migrations slow, but it does not test your estate for you. This post covers why to move, what the wizard does and does not do, the capacity and licensing questions, and the upgrade plan we use.

Why Move Off Gen1 at All

Microsoft describes Power BI Dataflow Gen1 as being "in a legacy state" with no new feature investment. It remains supported, and existing dataflows keep running. Microsoft has not announced a retirement date, and we would not plan around rumoured ones.

For Premium customers with Fabric access, Microsoft recommends Gen2. For Pro and Premium Per User (PPU) customers, Microsoft says Gen1 remains a supported option and that Gen2 guidance for those scenarios is still evolving.

What Gen2 adds:

Output destinations. Gen1 keeps its results in its own internal storage. Gen2 can also write to Lakehouse tables and files, Warehouse, Fabric SQL database, KQL databases, Azure SQL, Azure Data Lake Gen2, Snowflake, PostgreSQL and SharePoint files, so one dataflow can feed many consumers.
Staging and fast copy. Gen2 can stage data in a lakehouse and use warehouse compute, and fast copy uses the pipeline Copy activity for large volumes.
Fabric integration. Gen2 runs as a pipeline activity, shows up in the Monitoring hub and has a much more detailed refresh history.
CI/CD. Since April 2026 every new Dataflow Gen2 is created with Git integration and CI/CD support, and the wizard upgrades to that same item type. Dataflows can sit in source control and move through deployment pipelines with everything else. Our CI/CD and Git guide covers the set-up.

The Power Query logic itself does not change. If your team knows Power Query, it already knows how to author Gen2. For where dataflows sit next to pipelines, copy job and mirroring, see our guide to Data Factory in Microsoft Fabric.

What the Upgrade Wizard Does

You start it from the ellipsis menu next to a Gen1 dataflow in the workspace item list, by choosing Upgrade to Dataflow Gen2 (Preview). It runs in four steps: an overview, a selection step where you can add other Gen1 dataflows from the same workspace, a review step, and a monitor step that shows progress and lets you download the results as a CSV upgrade report.

During selection, the wizard assesses each dataflow and gives it one of three statuses:

Ready to migrate - nothing detected that needs a manual step
Needs Attention - the upgrade can go ahead, but there is follow-up work
Upgrade unavailable - blocked, either because you are not the dataflow owner or because a data source connection cannot be resolved

For larger estates there is a REST API, also in preview. One endpoint upgrades a batch of between 1 and 50 Gen1 dataflows from a workspace per request, and a companion endpoint returns a readiness verdict for every Gen1 dataflow in a workspace. In its FabCon Europe 2026 round-up for Data Factory, Microsoft described API support, the downloadable report and new validation rules as the next steps for the wizard. The wider conference picture is in our FabCon Europe 2026 recap.

What It Does Not Do

Microsoft's documentation is clear about the limits. These are the ones that matter most:

No way back. The Gen1 item is replaced and cannot be reverted. To keep the original, Microsoft points you to Save As, which creates a separate Gen2 (CI/CD) copy with a new ID.
No data carried over. The upgraded dataflow holds no data until its first refresh. The Gen1 cache does not survive, so if you cannot reload history from the source, back it up first, for example to a lakehouse.
No DirectQuery. Gen2 does not support DirectQuery through the dataflow connector. Semantic models that use it must move to Import or Direct Lake.
Incremental refresh is not carried over. You set it up again on the upgraded dataflow, and it needs a data destination that supports it.
No cascading refresh. Linked dataflows downstream need their own schedule or a pipeline.
The legacy connector breaks. The old Power BI Dataflows connector cannot read a Gen2 (CI/CD) item, so consumers must switch to the Power Platform Dataflows connector. The wizard does not detect this, and it cannot see Power BI Desktop or Excel files stored outside Fabric.
A query limit. Gen2 (CI/CD) supports up to 50 enabled queries per dataflow. Anything above that must be cut down before the upgrade.

Microsoft also says the assessment does not detect every limitation. Treat a green status as a good sign, not a test result.

Capacity and Licensing

The wizard only works on Gen1 dataflows in a workspace assigned to Fabric capacity. You also need the Users can create Fabric items admin setting enabled for your scope, Admin, Member or Contributor access to the workspace, and ownership of the dataflow.

Licensing shifts in a way that can help. Publishing or managing a Gen1 dataflow needs a Pro or PPU licence. Microsoft says a Gen2 dataflow can be authored with a Fabric (Free) licence in a Premium capacity workspace. The cost moves to the capacity instead, so the real question is whether the capacity can carry the load.

Gen2 uses a different compute and billing model. Standard compute for Gen2 (CI/CD) is rated at 12 CU for each second of a query's first 10 minutes and 1.5 CU per second after that. Staging uses high scale compute at 6 CU per second, and fast copy is rated at 1.5 CU per second of aggregate core time. Microsoft says plainly that CU consumption can differ from Gen1, so check refresh duration and CU in the Capacity Metrics app before you upgrade the whole estate. Our capacity sizing guide explains how to read that evidence.

If your dataflows live in Pro or PPU workspaces today, upgrading means moving them onto capacity first. That is a sizing and cost decision, not a wizard setting.

A Step-by-Step Upgrade Plan

This is the order we follow. It favours small waves over a big bang, which matches Microsoft's own advice to start with a representative set and expand once it is validated.

1. Inventory and map lineage

List every Gen1 dataflow, its owner, its schedule and everything that reads it. Use Lineage view in the workspace, and search Power BI Desktop and Excel files for the PowerBI.Dataflows connector, because the wizard cannot see them.

2. Run the assessment without upgrading

Open the wizard, or call the readiness endpoint, and record each status and reason. Sort the estate into ready, needs attention and blocked.

3. Fix the blockers and the pre-work

Take over ownership where needed, then confirm a clean Gen1 refresh. Rebind unresolved connections. Cut any dataflow above 50 enabled queries. Rename dataflows yourself if a name has unsupported characters or clashes with another, rather than letting the wizard strip characters or add a "Migrated" prefix. Back up any history you cannot reload.

4. Pilot a representative wave

Pick a few dataflows that cover your common patterns, ideally one with linked entities and one with an Import consumer. Upgrade them in a quiet window, because manual and scheduled refreshes do not run during the upgrade.

5. Complete the post-upgrade steps, in order

Refresh each upgraded dataflow. Then open and save every downstream dataflow so it rebinds; if you skip this, it keeps returning stale data. Then finish the Needs Attention actions: destinations, incremental refresh, schedules for linked dataflows, connector changes and storage mode changes.

6. Validate, then scale

Compare row counts and key totals against the last Gen1 output. Check refresh times and CU in the Capacity Metrics app. Update any automation that calls the Power BI REST API, because Gen2 uses the Fabric REST API, which Microsoft says does not have full parity. Then move to the next wave, using the API for bulk upgrades where it saves time.

Testing and Rollback

Because the upgrade is irreversible, rollback has to be planned before the upgrade, not after.

Rehearse with Save As. Save As creates a Gen2 (CI/CD) copy and leaves the Gen1 untouched. Use it to test destinations, refresh times and CU on real logic before running the in-place upgrade on the original. The copy has a new ID, so it is a test bed, not a drop-in replacement.
Keep a copy of the queries. Export each Gen1 dataflow as a Power Query template so the logic exists outside the service.
Know the failure path. If an upgrade fails, the wizard restores the original Gen1 dataflow, and you cannot retry that dataflow for 24 hours.
Note the 30-day window. Upgraded Gen1 items are kept in a soft-deleted state for up to 30 days, and during that time you cannot turn on outbound access protection for the workspace. Microsoft does not document a way to restore from that state, so do not count it as a rollback.

Common Gotchas

Most problems sit at the edges of the estate, not in the dataflow itself:

Reports that look fine but are not. A semantic model on the legacy connector keeps showing its imported data until its next refresh fails. Check refreshes, not visuals.
Viewers lose access. Workspace Viewers can see an upgraded dataflow but cannot consume its tables through the connector. Give consumers Contributor or above, or better, point them at a data destination.
Credentials do not carry over. The Power Platform Dataflows connector is a different source, so each consumer has to sign in again.
Staging items appear. Gen2 creates system staging lakehouse and warehouse items in the workspace. They are expected; leave them alone.
Preview timing. Third-party trackers of Microsoft's public release plan show general availability for the wizard targeted at Q4 2026, after slipping from September. Microsoft has not confirmed a date, so plan production waves with that in mind.

How We Help

Our Microsoft Fabric consultants run Gen1 to Gen2 upgrades as a short, staged project: inventory and lineage, a readiness assessment, a pilot wave with CU measured against your capacity, then bulk upgrades with the post-upgrade steps scripted rather than done by hand. Where dataflows sit in Pro or PPU workspaces today, we size the capacity first, so the move does not turn into a throttling problem.

Sources and Further Reading

Frequently asked

It performs an in-place upgrade of a Power BI Dataflow Gen1 to Dataflow Gen2 (CI/CD). The upgraded item keeps the same ID, name, schedule and connections, so most downstream semantic models and reports keep working without changes. You can upgrade one dataflow or several from the same workspace together. The wizard is currently in preview.

No. The original Gen1 dataflow is replaced and cannot be reverted. If you want to keep it, use Save As instead, which creates a separate Dataflow Gen2 (CI/CD) copy with a new ID and leaves the Gen1 untouched. If an upgrade fails, the wizard restores the original Gen1, and you must wait 24 hours before retrying.

Yes. The wizard only works on Gen1 dataflows in a workspace assigned to Fabric capacity. You also need the Users can create Fabric items setting enabled, Admin, Member or Contributor access to the workspace, and ownership of the dataflow. Dataflows in Pro or PPU workspaces need to move onto capacity first.

Most keep working because the dataflow ID is preserved. The exceptions are consumers using the legacy Power BI Dataflows connector, which cannot read a Gen2 (CI/CD) item, and semantic models using DirectQuery, which Gen2 does not support. Those need the Power Platform Dataflows connector and an Import or Direct Lake storage mode.

Microsoft says Dataflow Gen1 is in a legacy state and will not receive new feature investment, but it remains supported and existing dataflows keep running. Microsoft has not announced a retirement date. It recommends Gen2 for Premium customers with Fabric access, while Pro and PPU customers can keep using Gen1.

It can differ. Gen2 uses a different compute and billing model, consuming capacity units for standard compute, staging and fast copy. Microsoft advises validating refresh duration and CU consumption after upgrading, so pilot a few dataflows and check the Capacity Metrics app before upgrading the whole estate.

Yes. The wizard can upgrade several Gen1 dataflows from one workspace together, and a REST API, also in preview, upgrades between 1 and 50 dataflows per request with a companion endpoint that reports readiness for every Gen1 dataflow in a workspace.