How to reduce your cloud bill: the checklist that works, in order

Quick answer

Reduce a cloud bill in this order: delete waste (orphaned disks, idle load balancers, dead environments), right-size over-provisioned compute and storage, apply commitment discounts to the steady baseline, move fault-tolerant work to Spot, tier cold storage, and cut egress with CDNs and endpoints. Then keep it down by estimating cost in PRs and gating on a budget so it doesn't creep back.

Cloud bills grow through a thousand small defaults — an oversized instance here, an un-deleted environment there, full-rate storage on data nobody reads. Bringing one down is a checklist, run in order of impact and ease. Here's the one that works.

1. Delete waste (immediate, no downside)

Orphaned EBS/persistent disks from deleted instances.
Idle load balancers and unattached Elastic IPs.
Forgotten dev/test environments and stale snapshots.

2. Right-size (biggest single lever)

Right-size compute to actual CPU/memory use.
Match instance generation/family — e.g. Graviton for ~20% off.
Right-size databases and over-provisioned throughput (Cosmos RU, DynamoDB).

3. Commit to the baseline

Reserved Instances vs Savings Plans (AWS), Azure reservations/savings plans, GCP CUDs.
Only after right-sizing — never commit to oversized resources.

4. Spot for fault-tolerant work

AWS, Azure, GCP Spot — up to ~90% off for batch, CI, and stateless work.

5. Tier storage and cut egress

Move cold data to cheaper classes (S3, Azure, GCS).
Put a CDN in front of content and add VPC endpoints to cut egress and NAT.

6. Keep it down with a cost gate

One-time cleanups decay. The durable fix is making cost visible before deploy: estimate in pull requests and gate on a budget so the bill stops creeping back up.

FAQ

What's the fastest way to reduce a cloud bill?

Delete waste first — orphaned disks, idle load balancers, unattached IPs, forgotten environments — because it's pure savings with no downside. Then right-size over-provisioned resources, then apply commitment discounts (Reserved Instances, Savings Plans, Committed Use) to the steady baseline. Waste cleanup is immediate; commitments compound.

What are the biggest cloud cost levers?

In rough order of impact: right-sizing compute and storage, commitment discounts on steady workloads, Spot/preemptible for fault-tolerant work, eliminating idle and orphaned resources, storage tiering, and cutting egress with CDNs and endpoints. The mix depends on your workload, but compute right-sizing plus commitments usually dominates.

How do I stop my cloud bill from growing again?

Make cost visible before deployment. Estimate infrastructure cost in pull requests and gate on a budget, so expensive changes are caught at review instead of on the invoice. One-time cleanups decay; a cost gate in CI keeps the bill in check continuously.

Should I buy Reserved Instances or Savings Plans first?

Only after right-sizing — committing to oversized resources locks in waste. Once your baseline is right-sized and stable, cover it with the commitment model that fits (Savings Plans for flexibility, Reserved Instances for a fixed fleet). Commit to baseline, never peak.

How much can I realistically save?

Most under-managed cloud accounts have 20-40% of spend in waste and missed discounts: idle resources, oversized instances, no commitments, hot storage that should be cold, and egress that a CDN would absorb. The exact figure depends on how disciplined the account already is.

How does C3X help reduce the bill?

C3X estimates infrastructure cost from your Terraform before deployment, so you can right-size and compare options on real numbers, and gate PRs on a budget so the bill stops creeping back up. It turns cost from a monthly surprise into a number you see at review time.

What to do next

Work the checklist top to bottom, then lock the gains in with a cost gate. C3X estimates your infrastructure cost from Terraform so you can right-size and compare options on real numbers, and gate PRs so the bill stays down. The quickstart runs it in minutes.