Cloud Cost Savings Estimator

Where the money actually goes

Compute usually dominates a cloud bill, but the surprises are elsewhere. Data egress is charged per gigabyte leaving the provider and is invisible until it is large. Unattached storage volumes and old snapshots accrue indefinitely after the instances they belonged to are gone. Idle load balancers, unused elastic IPs and forgotten development environments running at weekends all bill continuously. An audit of what is running but not used typically finds more savings than any pricing optimisation.

Rightsizing before commitment

The common mistake is buying reserved capacity for oversized instances, which locks in the waste for one to three years. Look at actual utilisation first: instances averaging under 20 percent CPU across a month are candidates for a smaller size or a different family. Only once workloads are correctly sized does a commitment discount make sense.

Commitment discounts

Reserved instances and savings plans typically cut 30 to 60 percent against on-demand in exchange for a one or three year commitment. The risk is straightforward: you pay whether or not you use it. Commit to your steady baseline, the capacity you are confident you will run regardless, and leave variable load on demand. Committing to peak capacity converts a variable cost into a fixed one at exactly the wrong level.

Spot capacity

Spot or preemptible instances offer 60 to 90 percent discounts and can be reclaimed with a short warning. That makes them excellent for batch processing, CI runners, rendering and anything checkpointable, and unsuitable for a stateful database. The saving is large enough that making a workload interruption-tolerant is often worth the engineering effort.

Storage tiering

Most stored data is never read after the first month. Lifecycle policies that move objects to infrequent-access and then archive tiers cut storage cost substantially. Check retrieval charges and minimum storage durations before applying them, because archive tiers bill for early deletion and retrieval can be expensive enough to erase the saving if the data is actually accessed.