In Fabric you pay by capacity, not per user. Managed well the model is economical; managed badly it produces surprise invoices.
01. How Capacity Works
The capacity you buy sets the compute power available per second. Pipeline runs, queries, model refreshes and report views all draw from the same pool.
02. Smoothing
Momentary overages aren't blocked immediately; usage is evaluated spread over time. That can let heavy overnight jobs eat into daytime capacity. Scheduling has to be deliberate.
03. Throttling Risk
If capacity is constantly full, jobs queue and reports slow down. Users experience this as Power BI being slow; the cause is often a data pipeline running on the same capacity.
04. Pausing
Capacity can be paused outside working hours and isn't billed during that time. In environments used only during the day, this noticeably lowers monthly cost.
05. Separating Workloads
Running development and production on the same capacity means a heavy test query can affect live reports. Separating them raises cost slightly but raises predictability a great deal.
06. Monitoring
The capacity metrics app shows which item consumes what. Reviewing that report monthly catches a single runaway pipeline early.