Power BI Fabric capacity: what it is and how to size it
Power BI Fabric capacity is a pool of compute that your organization reserves for its own workspaces. Without one, content runs on shared resources, and every user needs a paid license. With one, you gain larger models, more features and the wider Fabric toolset. But a capacity costs money every hour it runs. This guide explains how capacities work and how to choose a size without overpaying.
- F SKUs explained
- Smoothing and throttling
- Sizing from real usage
Last updated
What is Power BI Fabric capacity?
A capacity is dedicated compute. You buy it in Azure, pick a size and assign workspaces to it. After that, refreshes, queries and other jobs in those workspaces draw on that pool.
Sizes come as F SKUs. Each step up roughly doubles the capacity units available. So a small SKU suits a pilot, while a larger one carries a department or a whole company.
One capacity serves every Fabric workload. Power BI models, dataflows, lakehouses, warehouses and notebooks all share it. Therefore a heavy data engineering job can affect report speed, and the reverse. That shared design makes planning more important than it was with Premium alone. If you are still choosing between the two models, read Power BI Premium vs Fabric capacity, and for moving an existing tenant, see Power BI Premium to Fabric capacity migration.
Ideas you need before sizing
Six ideas explain nearly everything about how a capacity behaves.
Capacity units
A capacity unit measures compute, and each SKU provides a set number of them.
Interactive and background
Report queries are interactive, while refreshes and pipelines count as background work.
Smoothing
Fabric spreads the cost of an operation over time, so a short spike does not overload the pool.
Throttling
When use stays above the limit, Fabric first delays and then rejects new requests.
Pausing a Power BI capacity
A pay-as-you-go capacity can pause when idle, which then stops the compute charge.
Reservation
Committing for a term lowers the rate, because Microsoft rewards steady use.
When does a capacity make sense for a small team?
Not always. A team of report builders with Pro licenses on shared capacity can go a long way. So the first question is what you need that shared capacity cannot give.
There are three common triggers. First, models outgrow the size limit of a Pro workspace. Second, the team wants Fabric items such as a lakehouse or a warehouse. Third, many people only read reports, and licensing each one costs more than a capacity would.
That third trigger has a condition. As of October 2026, free-license viewing starts at F64. Below F64, every reader still needs Pro or Premium Per User. That threshold shapes the whole business case, so confirm it on Microsoft's licensing page before you commit.
Signs your Power BI Fabric capacity is the wrong size
These six signs point to a capacity that is too small, too large or badly scheduled.
- Reports slow down at the same hour daily
- Refreshes fail with capacity errors
- Utilization sits low all week
- Background jobs pile up overnight
- One workspace uses most of the units
- Nobody looks at the metrics app
How to size Power BI Fabric capacity in four steps
Sizing works best as a measured trial, not a guess.
List the workloads
First write down the models, refreshes and Fabric items that will run on the capacity.
Trial a small Power BI capacity
Then run real work on a trial or a small pay-as-you-go SKU for a few weeks.
Read the metrics
Next open the Capacity Metrics app and see how close usage comes to the limit.
Adjust and commit
Finally scale up or down, spread heavy jobs across the day and reserve once usage is steady.
Questions about Fabric capacity for Power BI
Is Fabric capacity the same as Power BI Premium capacity?
They are closely related. Power BI Premium came in P SKUs bought through Microsoft 365. F SKUs are bought through Azure, and they run every Fabric workload, reports included. Microsoft announced the retirement of P SKUs in March 2024. As of October 2026, new customers buy F SKUs, and existing P customers move to F SKUs as their agreements renew.
Do report users still need Pro licenses on a Fabric capacity?
Authors do. Anyone who publishes or shares content needs a Pro or Premium Per User license, whatever the capacity. Readers are different. On F64 or larger, people with a free license can view content. On F2 to F32, readers still need a paid license.
What happens when a Power BI capacity gets throttled?
Fabric applies limits in stages. First, interactive requests meet a short delay. If overload continues, the capacity rejects interactive requests. After prolonged overload, it rejects background jobs as well. Smoothing means brief peaks rarely trigger this. The Capacity Metrics app shows when throttling happened and which items caused it.
Can we pause a capacity at night?
Yes, on pay-as-you-go. Pausing stops the compute charge, although storage still bills. Content in the capacity is unavailable while it is paused, and scheduled refreshes will not run. So pausing suits development capacities more than production ones. A reserved capacity bills for the term whether it runs or not.
Can you help us choose a size?
Yes. One of our consultants reviews your workloads and usage, then recommends a size with the reasoning written down. Power BI Agency does not resell licenses, so the advice is neutral. We are a division of Sataware Technologies, and each engagement starts with a free assessment and a fixed scope.
Let's build your first dashboard together.
Tell us what you report on today. A consultant replies within 2 hours with next steps.
Trusted partners
Microsoft Solutions Partner · AWS Partner
Cloud & AI Platforms
Azure SQL, Microsoft Fabric and the data pipelines behind your reports.
AI Business Solutions
Power BI reporting on your Dynamics 365, Business Central and Power Platform data.
Security
Row-level security and Entra ID sign-in on every report and workspace.
Amazon Web Services
Power BI on your AWS data, with connectors for Redshift, Athena and S3.