What is the fastest AWS cost saving with no risk?

Delete clearly orphaned resources (unattached EBS volumes, idle Elastic IPs, idle load balancers and NAT gateways), and migrate gp2 volumes to gp3 — a live, ~20% cheaper change for the same performance. These are zero-risk and need no commitment.

How often should I run an AWS cost review?

Monthly, after the bill closes, for the formal review and your month-over-month trend; plus a lightweight weekly waste-and-anomaly sweep. Cloud usage changes constantly, so a recurring cadence beats a one-off clean-up that quietly undoes itself.

Which AWS services help with cost optimization?

AWS Cost Explorer and the Cost and Usage Report (CUR) for visibility, Cost Optimization Hub and Compute Optimizer for rightsizing recommendations, AWS Budgets and Cost Anomaly Detection for guardrails, and Organizations Tag Policies plus SCPs for governance. The CloudFinOpsKit AWS Tool stitches these authoritative sources into one report.

AWS · The pillar guide · Updated June 2026

The AWS Cost Optimization Checklist for 2026

By the CloudFinOpsKit team — the people who build the AWS tool's 70+ Well-Architected cost checks. 12 min read.

Industry surveys consistently put wasted cloud spend near 30%, and a disciplined practice recovers 20–35% in the first year. This is the working checklist we run, organized by lever and mapped to the AWS Well-Architected Cost Optimization Pillar. Start at the top — the early sections are zero-risk and need no commitment.

1. Kill waste first (COST 04 / 09)

Delete unattached (available) EBS volumes — snapshot-first, grace period. Full guide →
Release idle Elastic IPs (and remember: all public IPv4 is billed since Feb 2024).
Delete idle load balancers (ALB/NLB with no targets, Classic ELBs with no instances) and idle NAT gateways (confirm zero bytes in CloudWatch first).
Remove idle Global Accelerator accelerators and unused ENIs.
Clean up old/disconnected EBS snapshots and unused AMIs (deregister the AMI before deleting its snapshot).
Find stopped EC2 instances still paying for attached EBS — terminate or snapshot.

2. Right-size & modernize (COST 06 / 07)

Action AWS Compute Optimizer and Cost Optimization Hub recommendations for EC2, EBS, Lambda and RDS.
Migrate gp2 → gp3 (≈20% cheaper, live change) and downsize over-provisioned io1/io2.
Move eligible workloads to Graviton (Arm) for better price/performance.
Right-size RDS/Aurora (smaller class, Graviton, or Aurora Serverless v2 where idle).
Use Karpenter / Cluster Autoscaler on EKS; consolidate under-used node groups.
Schedule non-prod off out-of-hours (Instance Scheduler).

3. Buy the right rate (COST 07)

Cover your stable baseline with Compute Savings Plans (flexible) or RIs; target 60–80% coverage at ≥95% utilization. SP vs RI guide →
Review existing commitments for under-utilization; don't auto-renew stranded ones.
Run fault-tolerant / batch / EKS work on Spot.

4. Optimize storage & data (COST 09)

Add S3 lifecycle rules (Standard → Standard-IA → Glacier → expire) or enable Intelligent-Tiering.
Set CloudWatch Logs retention — log groups default to "never expire"; trim noisy groups.
Review cross-Region / inter-AZ data transfer and NAT data-processing; use VPC endpoints for AWS-service traffic.
Right-size managed services (DynamoDB on-demand vs provisioned, ElastiCache, OpenSearch).

5. Govern & allocate (COST 01 / 02 / 03)

Create AWS Budgets with 80/100/120% actual + forecast alerts on every account.
Turn on Cost Anomaly Detection monitors.
Enforce allocation tags (cost-center, owner, environment, application) with Organizations Tag Policies + SCPs, and activate cost allocation tags in Billing.
Use the account as your primary cost boundary; let consolidated billing roll it up.
Enable Cost Explorer and the CUR for history and granular analysis.

6. Optimize AI / Bedrock spend (COST 06)

Get per-model token visibility from CloudWatch (AWS/Bedrock).
Enable prompt caching, cap output tokens, and right-size / route to cheaper models.
Only keep Provisioned Throughput at sustained high utilization. Bedrock guide →

7. Make it a habit (operate)

Run a monthly cost review with named owners and real decisions.
Track realized vs projected savings month over month to prove ROI to finance.
Move from showback to chargeback as the org matures.
Treat cost as an engineering quality metric, not a once-a-year finance task.

Run all 40+ checks in one click. The CloudFinOpsKit AWS Tool automates this entire checklist — read-only, across your account or whole Organization — pricing every finding from your actual Cost Explorer cost (with a region-accurate Price List fallback), scoring your FinOps maturity, and tracking month-over-month trends. Recommendations come from AWS's own engines (Cost Optimization Hub, Compute Optimizer, Cost Explorer), so the report is free of false positives.

Want the manual version? Grab the free AWS Cost Review Checklist (PDF + Word).

FAQ

Where do I start if I only have an hour?

Section 1. Unattached EBS volumes, idle Elastic IPs, idle load balancers and NAT gateways, plus gp2→gp3, are zero-risk and usually the biggest immediate win.

Do I need Cost Explorer enabled?

It's strongly recommended (free) for exact billed figures. Without it, tools fall back to region-accurate list prices — directionally right, but enable Cost Explorer for invoice-grade numbers.

How is this different from Trusted Advisor?

Trusted Advisor surfaces some checks; this checklist is the full FinOps lifecycle (inform → optimize → operate) and pulls together Cost Optimization Hub, Compute Optimizer, Cost Explorer and governance into one repeatable review.