Microsoft Fabric licensing — capacity is the unit, and most enterprises buy too much of it.

F2 to F2048 — the capacity ladder that determines almost everything.

Fabric capacity is sold as F-SKUs ranging from F2 (the entry-level capacity, roughly equivalent to a starter dev environment) through F2048 (suited to multi-petabyte enterprise warehousing). Each step up the ladder roughly doubles capacity and cost. Capacity Units (CUs) are the underlying compute metric. Workloads consume CUs at rates that depend on workload type — Power BI reporting consumes differently from Synapse warehouse queries, which consume differently from Spark notebook runs.

The architectural decision is whether to consolidate workloads on a single large F-SKU or distribute across several smaller capacities. Consolidation is operationally simpler but creates noisy-neighbour risk: a single runaway Spark notebook can throttle the executive Power BI dashboard. Distribution gives isolation but raises floor cost. For most enterprises, the right answer is two or three F-SKUs aligned to workload class (Power BI delivery, data engineering, data science) rather than the single mega-capacity that Microsoft sales motion frequently proposes.

The reservation discount is meaningful, but only if the capacity number is right.

Fabric F-SKUs can be purchased pay-as-you-go (hourly, Azure billing) or via annual reservation at approximately 41% discount. The reservation economics are compelling at face value — the median customer recovers the reservation premium within seven months. The risk is the same as any reservation: locking in a capacity tier that the workload outgrows or undershoots within the reservation term.

The correct approach is to run on PAYG for the first 60–120 days post-deployment, observe actual CU consumption, and reserve only the established baseline — typically the F-SKU one tier below peak observed usage, with pause-during-off-hours scripting to handle the peaks. Reserving the peak F-SKU annually almost always leaves capacity idle.

Power BI Premium migration

Customers on Power BI Premium per-capacity (the P-SKU model) are being migrated to F-SKUs. The mechanical conversion is one P-SKU to the equivalent F-SKU, but the pricing is not equivalent. F-SKUs are billed through Azure and can be paused; P-SKUs were billed through M365 and ran continuously. For customers who used Power BI Premium primarily during business hours, the migration to F-SKU with scheduled pause delivers 30–55% cost reduction at the same capacity. Microsoft sales motion will not surface this option proactively. Capturing the pause-and-reserve saving estate-wide is a textbook license cost reduction exercise — observed-consumption sizing, not the peak number on the quote.

Power BI Pro and PPU are not going away.

For enterprises with fewer than 500 Power BI consumers, per-user Power BI Pro ($10/user/month) is usually still more economical than Fabric capacity. The capacity model becomes the right answer once the consumer count multiplied by Pro pricing exceeds the cost of the smallest F-SKU that can serve the workload — typically around the 500–800 user range for read-only consumption, lower for organisations using paginated reports or AI features. Power BI Premium Per User (PPU) sits as a niche middle path for organisations with a small number of advanced authors.