Moving from Power BI Premium to Fabric: a practical guide
If your organization runs Power BI on a Premium capacity, it has an end date. Microsoft no longer sells Premium per capacity (the P SKUs), and customers can’t renew them when their Enterprise Agreement or Microsoft Cloud Agreement comes up. Each P SKU retires at the end of its current term, and the replacement is a Microsoft Fabric capacity (an F SKU).
The move is not automatic. Someone has to buy the F capacity in Azure and move every workspace onto it. If the Premium term ends first, Microsoft’s migration FAQ spells out what happens: 30 days of grace, then interactive use is throttled from day 31, and from day 91 everything is rejected until the workspaces are on a Fabric capacity.
The good news is that for most organizations this is a reassignment, not a rebuild. Reports, semantic models, refresh schedules, apps and deployment pipelines move with the workspace. The move itself takes minutes. The work is in sizing the new capacity, choosing a quiet window, proving afterwards that everything still works, and being able to undo it. This guide covers each of those, and there’s a printable checklist at the end.
What changes when you move to Fabric
Your content doesn’t change. How the capacity is bought and run does.
| Premium (P SKU) | Fabric (F SKU) | |
|---|---|---|
| Bought through | Microsoft 365 agreement | Azure, pay-as-you-go or reserved |
| Pause and resize | Not available | Pause when idle, resize in minutes |
| Autoscale | Optional add-on | Gone. Resize instead, or use capacity overage |
| Free viewers | Yes | Only on F64 and larger. Below F64, every viewer needs Pro |
| Power BI Embedded | EM or A SKUs | Included in every F SKU |
| Power BI Report Server | Included | Needs a Fabric reservation, or SQL Server Enterprise with Software Assurance |
Two of those rows catch people out. If you size down below F64 to save money, your report viewers suddenly need Pro licenses. And if you run Power BI Report Server on-premises, check how you’ll license it before you cancel Premium.
1. Size the new capacity from your own data
Microsoft maps by capacity units: P1 to F64, P2 to F128, P3 to F256. That’s the safe starting point, not necessarily the right answer.
Open the Capacity Metrics app against your Premium capacity and look at the last 30 days. Sustained peaks near 100% mean you need the equivalent F SKU or larger. Long quiet stretches mean you may be able to run smaller, or split production from development and pause the development capacity when nobody is using it. If you plan to add Lakehouses, notebooks or pipelines later, leave room for them: in Fabric they draw from the same capacity as your reports.
2. Get the new capacity ready
- Create the F capacity in the same Azure region as your Premium capacity. Same-region moves are simple. Cross-region moves are a separate project, because large-format semantic models and Fabric items don’t survive the trip and have to be recreated.
- Add the capacity admins before moving anything. The person doing the move needs to be a Fabric administrator and an admin or contributor on the new capacity.
- Write down your Premium capacity settings, such as workload settings, memory limits and notification thresholds, and recreate them on the F capacity. Tenant settings carry over. Capacity settings don’t.
- Decide who may create Fabric items. The “Users can create Fabric items” setting doesn’t affect existing reports, but it does decide whether someone’s first notebook ends up competing with your finance refreshes.
3. Record what “working” looks like before you move
This is the step most migrations skip, and the one that makes the rest provable. Before anything moves, write down for every semantic model:
- its refresh schedule: days, times, time zone, and whether it’s on
- its data connections, and which gateway each one uses
- whether its last refresh succeeded
Anything already failing goes on a list now, so nobody blames the migration for it later.
4. Pilot two or three workspaces
Pick low-risk workspaces, including at least one with a scheduled refresh, and move them in a quiet window. Moving a workspace cancels any refresh or query running at that moment, so avoid your refresh peaks.
5. Check that everything still works
Compare against the list from step 3:
- every workspace shows the new capacity
- refresh schedules are still on, with the same days and times
- data connections and gateway bindings are unchanged, and no credentials need re-entering
- an on-demand refresh succeeds, and so does the next scheduled one
- reports open, including DirectQuery, live connection and paginated reports
- subscriptions send, and XMLA tools such as Tabular Editor still connect
- the Capacity Metrics app shows the new capacity comfortably below its limits
Microsoft’s own guidance is the same: trigger a refresh after the move and confirm the next scheduled refresh succeeds. Run these checks again the morning after, once the overnight refreshes have run.
6. Move the rest, then keep watching
Move the remaining workspaces in one planned window, or in waves by department. Re-run anything the move cancelled. Repeat the checks after the first scheduled refreshes and keep an eye on refresh failures and utilization for a couple of weeks.
Only then cancel Premium. Until that point it’s your way back: every workspace can be moved back to it.
7. Then start saving
The reason to like F SKUs is that you can pause and resize them. Once the dust settles, look at the utilization again. Development and test capacities can often be paused overnight and at weekends, and steady production usage is cheaper on a reservation than pay-as-you-go.
When it isn’t simple
Plan extra time if you have:
- a Premium capacity in a different region from where you want Fabric
- more than one Premium capacity, or Embedded (EM or A) capacities as well
- Private Link, customer-managed keys or strict network rules
- very large semantic models, or Fabric items already on the Premium capacity
- workspaces owned by people who have since left
Download the checklist
The steps above as a one-page printable checklist, with space to fill in your capacities, dates and owners: Premium to Fabric migration checklist (PDF).
How we do it
We run this migration as a flat-fee service. We size the capacity from your usage. Your team gets a written plan to approve before anything moves. The pilot comes first, then the cutover in one window. Afterwards we prove that every refresh schedule, connection and refresh still works, with a report you can keep. Your Premium capacity stays on until you’re satisfied. It’s $6,000 for tenants with fewer than 100 workspaces and $12,000 for larger ones, and the first conversation is free.