Type in what your workload requests. See what it actually costs — itemized like a receipt, not buried in a dashboard.
Workload estimate
millicores (1000m = 1 vCPU)
MiB
running pods for this workload
% of the request, from `kubectl top` or your dashboard
Try:
Item
Monthly
Total requested capacity, billed$0.00
Estimate only, based on public list pricing for requested (not used) vCPU and memory,
excluding node overhead, storage, and networking. Real bills vary by region, commitment
discounts, and provider. Built to make the size of the gap visible, not to replace your
cost explorer.
Want to find these across your whole cluster?
This calculator checks one workload at a time. The
Kubernetes Cost Audit Toolkit runs three read-only
scripts against your real cluster and finds every over-provisioned pod, orphaned volume,
and empty load balancer automatically — no dashboard, no agent, just kubectl.
Cost questions, answered
How much does a Kubernetes pod actually cost?
A pod's cost is driven by its CPU and memory requests (not limits) multiplied by replica count — the scheduler reserves that capacity on a node whether the pod uses it or not. At typical cloud list pricing, 1 vCPU requested runs roughly $25-35/month and 1GB of memory roughly $4-6/month, before node overhead.
Why do requests cost money even when the pod is idle?
The scheduler reserves the requested CPU and memory exclusively for that pod, whether or not it's actually using it. That reserved capacity is what determines node size — and node size is what the cloud provider bills for, not real-time usage.
What's the difference between requests and limits for cost?
Requests drive scheduling and cost — a node needs enough allocatable capacity to satisfy every pod's requests. Limits only cap usage once a pod is already running; they don't change what capacity gets reserved or billed.
Is trimming requests always safe?
Not from a single snapshot. Check usage trends over 1-2 weeks, not one quiet moment, before reducing requests — a low reading during off-peak hours isn't the same as low peak load. Tools like VPA in recommendation mode can do this trend analysis for you.