Cloud agility is amazing — until the monthly invoice lands. If your AWS bill feels like it’s growing faster than revenue, you’re not alone. The good news: most teams can trim 25‑50% of AWS spend in under a quarter without rewriting their applications. This guide distills the highest‑impact moves — updated for the newest 2024‑25 features like Cost Optimization Hub and Graviton3 — so you can start saving today.

0. Hunt the Zombies First

Before dashboards and commitments, wipe out truly idle spend:

• EC2 & RDS instances from last quarter’s experiment still running 
• Orphaned EBS volumes / snapshots 
• Unused Elastic IPs and NAT Gateways (swap to t4g NAT instances or VPC Endpoints for low‑throughput dev nets) 
• Stale Lambdas, log groups, and AMIs

A one‑time cleanup here routinely chops 5‑15% off the bill and gives you a cleaner baseline for the steps that follow.

1. Measure Before You Optimize

You can’t optimize what you can’t see.

• Enable AWS Cost Explorer and tag 95% of resources within 30 days.
• Deploy the Cloud Intelligence Dashboards CloudFormation stack for exec‑friendly, real‑time KPIs.
• Set budgets & alerts for both overall spend and key services (e.g., >10% MoM increase triggers Slack alert).
• Create tiered billing alarms in CloudWatch (e.g., $100, $500, $1,000) so bill shocks never wait for month‑end.

Here is a CLI one-liner to see top 5 services by cost last month (excluding tax):

aws ce get-cost-and-usage \
  --time-period Start=$(date -v-1m +%Y-%m-01),End=$(date +%Y-%m-01) \
  --granularity MONTHLY --metrics UnblendedCost \
  --group-by Type=DIMENSION,Key=SERVICE |
  jq '[.ResultsByTime[0].Groups[]
        | select(.Keys[0]!="Tax")
        | {service:.Keys[0], cost:(.Metrics.UnblendedCost.Amount|tonumber)}]
        | sort_by(.cost) | reverse | .[:5]'

2. Right‑Size Continuously

• Turn on AWS Compute Optimizer recommendations for EC2, EBS, and Lambda.
• For Kubernetes, pair Karpenter with Cluster Autoscaler to downscale idle nodes in minutes.
• Adopt Target Tracking Scaling Policies instead of manual instance sizing.
• Run Trusted Advisor idle checks (Business/Enterprise Support) to surface under‑utilized RIs, EIPs, and NAT Gateways.

Pro tip: Treat right‑sizing as a daily automated job, not a one‑time project.

3. Switch to Efficient Silicon

• Graviton3 instances deliver ~25% better price‑performance vs x86 for typical workloads.
• For ML inference, evaluate Inferentia2 or AWS Trainium.
• Benchmark first; many workloads need only a simple --platform linux/arm64 rebuild.

4. Commit Smartly (Savings Plans, RIs, Spot)

Option	Typical Savings	Best For	Caveats
Compute Savings Plan	40–66%	Flexible EC2/Fargate/Lambda usage	1‑ or 3‑year commitment
Standard RI	40–72%	Steady, long‑running instances	Tied to family/region
Spot Instances	70–90%	Fault‑tolerant & batch jobs	Can be interrupted

Blend Savings Plans for baseline load and Spot for bursty tasks to maximize ROI.

5. Automate Shutdowns

• Use the Instance Scheduler (Solution Library) to power off dev/test stacks outside business hours.
• Lambda cron jobs can tag & stop idle RDS clusters and Redshift dev clusters.
• For containers: employ horizontal-pod-autoscaler with zero‑replica floor during nights/weekends.
• Replace small NAT Gateways in non‑prod VPCs with cheaper t4g NAT instances or private endpoints.

6. Storage Hygiene

• Enable S3 Intelligent‑Tiering for unpredictable access patterns; no retrieval fees for >128KB objects.
• Expire old EBS snapshots to Snapshot Archive — costs 75% less than standard.
• Compress & partition logs before shipping to S3/Glacier; consider Parquet over raw JSON.

7. Database Savings

• Migrate from general‑purpose to Graviton‑based RDS/TiDB/Aurora instances.
• Adopt Aurora Serverless v2 for spiky workloads (provision to zero in <1 min).
• Turn on RDS storage autoscaling to pay only for used GB.
• Archive cold data from RDS to S3/Glacier, then query it on demand via Athena or Redshift Spectrum.

8. Leverage the Newest Services & Features

• Amazon S3 Express One Zone (Dec 2024) offers SSD‑grade latency for 60% less than Multi‑AZ.
• Run GPU‑intensive ML training or fine‑tune jobs on AWS Batch + Spot instead of long‑lived on‑demand GPU nodes — restart‑tolerant workloads often save 60–80%.

9. Observability on a Budget

• Use FireLens to route container logs directly to S3 -> Athena rather than CloudWatch Logs.
• Lower CloudWatch Logs retention to 14 days; ship older logs to cold storage.
• Sample traces (X‑Ray) at 10–20% for prod to cut ingestion costs.

10. Adopt AWS Cost Optimization Hub

A single console aggregates Compute Optimizer, Trusted Advisor, and Savings Plan hints. Prioritize actions with highest $ impact and automate via AWS Systems Manager OpsCenter tickets.

11. Review & Repeat

Schedule a quarterly cost‑optimization sprint:

1. Refresh Compute Optimizer & Hub reports.
2. Triage actions >$500/mo.
3. Automate the change via IaC pull requests.
4. Celebrate wins publicly.

Conclusion

AWS cost optimization isn’t a one‑off exercise — it’s a continuous discipline that blends tooling, automation, and culture. By executing the twelve tactics above, SaaS teams routinely cut 30‑50% of spend within a single quarter, freeing budget for innovation.

Need a jump‑start? Schedule a discovery call with an nScope architect. Walk away with a prioritized action list and an AWS Savings Plan Decision Matrix tailored to your workloads.

Book your session