HPA calculator: how many replicas will the Horizontal Pod Autoscaler pick?

Enter where your Deployment is now and the target from your HPA spec — see the next scaling step and where it settles. Formula: desired = ceil(current × currentMetric / target), skipped within the 10% tolerance, clamped to min/max and the scaling policy.

1Where you are now — pick a scenario or type your numbers
2Your HPA spec — averageUtilization, minReplicas, maxReplicas
More options: tolerance, scale-up policy

Step by step

Assumes the total load stays constant, so per-pod utilization falls as replicas grow (utilization × replicas = const). Real HPA also waits for metrics from new pods (readiness, the 5-minute scale-down stabilization window), so treat the steps as the shape, not the timing. CPU utilization is measured against the pod's request, not its limit — no request, no HPA.

Example

4 replicas at 90% CPU with a 60% target: 4 × 90/60 = 6 → the HPA scales to 6 replicas, and the load spreads to 60% per pod — the steady state.

Next: how many nodes those replicas need · node reservation formulas.