Microsoft Fabric makes it easy to start new workloads.

Semantic models, Dataflows, pipelines, notebooks, warehouses, Copilot and AI workloads can all compete for the same capacity — and sooner or later someone asks the uncomfortable question:

Where did all our CUs go?

This session presents a practical Fabric FinOps approach based on real production experience managing multiple Fabric capacities.

Instead of simply walking through the Capacity Metrics App, we will focus on the decisions that come after monitoring.

How do you identify workloads that actually matter?
When should you optimize, reschedule, move, redesign or scale?
Why can background operations affect capacity long after a refresh has finished?
Which optimizations produce meaningful savings — and which only make dashboards look greener?

We will examine real examples involving semantic model refreshes, Dataflows, workload distribution, workspace cleanup and capacity planning.

I will also show how Copilot and AI workloads change the capacity conversation.

The goal is simple: move from capacity monitoring to capacity management.