Cloud Cost Reduction Checklist: 20 Quick Wins (2026)
A practical checklist of 20 actionable steps to reduce your cloud bill. From quick wins to architectural changes, ranked by effort and impact.
TL;DR
- • Quick wins (1-2 hours): delete zombies, rightsize instances, schedule dev/test shutdown — saves 10-30% immediately.
- • Medium effort (1-2 days): buy Savings Plans for baseline, set up spot for batch jobs, review egress costs — saves 30-60%.
- • Strategic (1-2 weeks): evaluate multi-provider for commodity compute, containerize for bin-packing, adopt serverless for bursty workloads.
This checklist provides 20 actionable steps to reduce your cloud bill across all 8 major providers, organized from quick wins to architectural changes. Each item includes estimated savings potential and implementation effort. Use this as a quarterly review checklist to keep costs optimized. The checklist now covers strategies specific to both Big 3 (AWS, Azure, GCP) and smaller providers (DigitalOcean, Hetzner, Linode, Oracle, Vultr).
Side-by-Side Comparison
| Feature | AWS | Azure | GCP | DO | Hetzner | Linode | OCI | Vultr |
|---|---|---|---|---|---|---|---|---|
| Rightsizing tools | AWS Compute Optimizer | Azure Advisor | GCP Recommender | Monitoring Dashboard | Cloud Console | Cloud Manager | Operations Insights | Monitoring API |
| Cost dashboard | AWS Cost Explorer | Azure Cost Management | Cloud Billing Reports | Billing page | Invoice overview | Billing & Usage | Cost Analysis | Billing page |
| Budget alerts | AWS Budgets | Azure Budget Alerts | GCP Budget Alerts | Billing alerts (email) | — (manual) | Billing alerts | Budget Alerts | Billing alerts (email) |
| Anomaly detection | Cost Anomaly Detection | Anomaly Alerts | Budget threshold alerts | — | — | — | Cost threshold alerts | — |
| Savings recommendations | Savings Plans / RI recs | Azure Advisor reservations | CUD recommendations | N/A (flat pricing) | N/A (flat pricing) | N/A (flat pricing) | Universal Credits recs | N/A (flat pricing) |
| Unused resource detection | Trusted Advisor | Azure Advisor | Recommender (idle VMs) | Manual (API) | Manual | Manual (API) | Recommender | Manual (API) |
Quick Wins (1-2 hours, 10-30% savings)
1. Delete unused resources: Check for unattached volumes, unused floating IPs/reserved IPs, idle load balancers, and orphaned snapshots. This applies to ALL providers. On DigitalOcean, check for forgotten Droplets running staging apps. On Hetzner, check for unused floating IPs (€4/mo each) and volumes.
2. Right-size instances: Use provider tools to identify over-provisioned VMs. On Big 3, use Compute Optimizer / Advisor / Recommender. On smaller providers, check CPU/memory graphs — if consistently below 40%, downsize one tier. Example: Downsizing from DO Premium 4 vCPU ($48/mo) to 2 vCPU ($24/mo) saves $288/yr per Droplet.
3. Enable auto-shutdown for dev/test: Schedule non-production instances to stop outside business hours. Running only 10 hours/day (weekdays) cuts costs by 70%. On Hetzner/DO/Vultr, use cron + API to automate stop/start.
4. Move to latest generation instances: AWS m7i vs m5, Azure Dv5 vs Dv4, GCP C3 vs N2 — 10-20% better price-performance. On Hetzner, ensure you're on CCX or CAX (newer) rather than CX (older) series.
5. Review data transfer: Check for unnecessary cross-region or cross-AZ transfers. On Big 3, this is a major hidden cost. On Hetzner (20 TB included) and DO (generous transfer), this is less of a concern.
Commitment Discounts (1-2 days, 30-60% savings)
6. Buy Reserved Instances or Savings Plans: Commit to 60-70% of your steady-state Big 3 usage. Start with 1-year terms. Example: 10 m5.xlarge instances: on-demand $1,400/mo → 1yr Savings Plan $883/mo (save $517/mo). Not applicable to DO, Hetzner, Linode, Vultr (already cheap).
7. Reserve databases: RDS/Cloud SQL/Azure SQL Reserved Instances provide 30-60% discounts. Example: RDS db.r5.xlarge 1yr reservation saves ~$140/mo vs on-demand.
8. Use GCP Sustained Use Discounts: These apply automatically for VMs running 25%+ of the month. Verify they're being applied in your billing console. No action needed.
9. Consider prepaid storage: AWS S3 reserved capacity, Azure Blob reserved capacity offer 20-40% discounts on storage.
10. Negotiate Enterprise Discounts: If spending $50K+/month, negotiate an EDP (AWS), Enterprise Agreement (Azure), or CUD commitment (GCP) for 5-15% additional discount. Oracle Universal Credits also provide volume discounts.
Architecture Optimization (1-2 weeks, 20-50% savings)
11. Use spot/preemptible for fault-tolerant workloads: CI/CD, batch processing, dev environments, and stateless workers. Save 60-90% on Big 3, 50% on Oracle. Alternative: Run these on Hetzner/DO at fixed rates often cheaper than spot.
12. Adopt serverless for variable workloads: Move bursty workloads to Lambda/Functions/Cloud Functions. Example: Lambda handling 1M requests/mo at 128 MB: ~$0.20/mo vs t3.micro 24/7: $7.50/mo.
13. Implement storage lifecycle policies: Automatically transition data from hot to cold tiers on Big 3 and Oracle. Example: 10 TB S3 Standard ($230/mo) → Glacier Instant ($40/mo). On smaller providers, object storage is already cheap ($0.02/GB/mo).
14. Use CDN for static content: Cloudflare (free tier), CloudFront, Azure CDN, Cloud CDN. Reduces origin server load and often costs less than direct serving. Works with any provider.
15. Containerize and use managed Kubernetes: Bin-packing multiple services onto shared node pools improves utilization from 10-30% to 60-80%. Available on AWS (EKS), Azure (AKS), GCP (GKE), DO (DOKS), Linode (LKE), Oracle (OKE), Vultr (VKE).
Multi-Provider Optimization (1-2 weeks, 30-50% savings)
16. Evaluate multi-provider for commodity compute: Move CI/CD, staging, and non-critical workloads to cheaper providers. Example: GitHub Actions runners on Hetzner ($0.021/hr) vs AWS ($0.170/hr) = 88% savings. Self-hosted GitLab runners on Vultr or DO at $48/mo vs equivalent AWS cost.
17. Use Hetzner/DO for development environments: A full development environment on Hetzner (8 vCPU, 32 GB) costs ~$34/mo vs $280/mo on AWS. Multiply by 10 developers = $2,460/mo saved.
18. Offload static hosting: Cloudflare Pages (free) or Vercel (free tier) instead of S3+CloudFront or Cloud Storage+CDN.
19. Use Oracle Free Tier for test databases: Oracle Always Free includes 2 AMD VMs + Autonomous Database. Run test/staging databases at $0/mo instead of paying for RDS test instances.
20. Monitor and compare quarterly: Prices change. New providers enter. Use our comparison pages to benchmark your current costs against alternatives each quarter.
Governance & Monitoring (Ongoing, prevents waste)
Set up budget alerts on every provider you use. AWS Budgets, Azure Cost Alerts, GCP Budget Alerts, and Oracle Budget Alerts are free. For DO/Linode/Vultr, set billing email alerts at your expected threshold.
Tag all resources with team, environment, and project on Big 3 providers. Untagged resources are the biggest source of shadow spend. On smaller providers, use naming conventions (e.g., staging-api-01) since tagging may be limited.
Review costs weekly — a 15-minute scan prevents month-end surprises. A misconfigured auto-scaling group, forgotten GPU instance, or runaway log storage can cost $1,000+ in a weekend.
Benchmark quarterly: compare your per-unit costs (cost per request, cost per user) against previous quarters. Track your blended cost per vCPU-hour across all providers.
Frequently Asked Questions
What is the fastest way to reduce my cloud bill?
The three fastest wins are: 1) Delete unused resources (immediate savings). 2) Right-size over-provisioned instances (30-50% per instance). 3) Move development and CI/CD workloads to a cheaper provider like Hetzner or DigitalOcean (50-88% savings on compute). These can typically be done in 1-2 days.
How much can I realistically save on cloud costs?
Most organizations can save 20-40% through optimization within a single provider. Organizations adopting a multi-provider strategy can save 40-60% total. The breakdown: 10-15% from waste elimination, 10-20% from commitment discounts, 5-10% from architecture optimization, and 20-40% from moving commodity compute to cheaper providers.
How often should I review cloud costs?
Weekly: quick scan of the cost dashboard for anomalies. Monthly: detailed review of top spending services and trends. Quarterly: strategic review of commitments, architecture, provider mix, and pricing changes. Annually: evaluate whether your current provider mix is optimal and renegotiate enterprise agreements.
Is it worth using multiple cloud providers?
Yes, for most organizations spending over $1,000/mo on cloud. The operational complexity of managing two providers is minimal for simple workloads (compute, databases, object storage). The cost savings can be substantial: 50-88% on commodity compute. Keep your core application on the Big 3 provider you depend on, and offload simple, standard workloads to cheaper providers.