Flexera's 2024 State of the Cloud report put a number on cloud waste: 32% of all cloud spend is wasted. For a company spending $500K/month on AWS, that's $160K/month — $1.92M per year — generating zero business value.
The sources of waste are consistent across organizations: over-provisioned instances, idle resources that nobody turned off, suboptimal storage choices, and data egress costs nobody models until they appear on a bill.
The fixes are well-documented, widely available, and regularly ignored until a billing alarm triggers a crisis.
Where Cloud Money Actually Goes
Understanding the waste categories requires understanding the spend categories. For a typical AWS-heavy SaaS company:
- EC2 (compute): 35–50% of total cloud spend
- RDS / database services: 15–25%
- S3 / storage: 5–15%
- Data transfer / egress: 5–15%
- Supporting services (CloudFront, Lambda, ECS, etc.): 15–25%
Waste concentrates most heavily in compute and databases — the categories where over-provisioning is easiest and most common.
Rightsizing: The Highest-Return Action
Most teams provision instances based on peak-load assumptions, then forget to resize when actual usage proves lower. The result: servers running at 15–30% CPU utilization on average, charged at 100% of instance cost.
| Original Instance | Avg. CPU Utilization | Rightsized To | Monthly Savings | Savings % |
|---|---|---|---|---|
| m5.2xlarge ($0.384/hr) | 18% | m5.large ($0.096/hr) | $208 | 75% |
| r5.4xlarge ($1.008/hr) | 22% | r5.xlarge ($0.252/hr) | $551 | 75% |
| c5.4xlarge ($0.680/hr) | 35% | c5.2xlarge ($0.340/hr) | $245 | 50% |
| db.r5.2xlarge RDS ($0.480/hr) | 25% | db.r5.large ($0.240/hr) | $173 | 50% |
| m5.xlarge × 10 instances ($0.192/hr each) | 20% avg | m5.large × 10 ($0.096/hr each) | $700 | 50% |
AWS Compute Optimizer and CloudWatch metrics make this analysis largely automated — it just requires someone to act on the recommendations.
Across a typical SaaS company, rightsizing alone reduces compute costs by 20–40%.
Reserved Instances and Savings Plans: The On/Off Switch
On-demand pricing is convenient and expensive. AWS Reserved Instances and Savings Plans offer 30–72% discounts for committing to 1- or 3-year terms.
For workloads with predictable, stable compute needs:
| Commitment Type | Term | Discount vs On-Demand | Flexibility |
|---|---|---|---|
| On-Demand | None | 0% | Full |
| Compute Savings Plan | 1-year | 33–40% | High (applies across instance families) |
| EC2 Savings Plan | 1-year | 40–50% | Medium (specific region/family) |
| Standard Reserved Instance | 1-year | 40–55% | Low (specific instance type) |
| Standard Reserved Instance | 3-year | 55–72% | Low |
The most actionable starting point: Compute Savings Plans for baseline load, with on-demand for burst capacity. Almost every organization with stable monthly cloud spend should have some portion in Savings Plans.
Idle Resources: The Easiest Wins
Idle resources — instances running but producing no business value — are the purest form of waste. Common examples:
- Development and staging environments running 24/7 when engineers work 8–10 hours/day
- Snapshot and AMI accumulation — old snapshots and machine images that nobody deletes
- Unattached EBS volumes left behind after instances are terminated
- Idle load balancers pointing to no active targets
- Forgotten Lambda functions with provisioned concurrency enabled
A scheduled shutdown of non-production environments for 14 hours/day (8pm–10am) and weekends (60+ hours/week off) reduces dev/staging compute costs by 35–40%.
A quarterly audit of unattached EBS volumes and old snapshots at a typical mid-size company routinely reveals $5,000–$20,000/month in accumulated waste.
Data Egress: The Hidden Bill
Data transfer out of AWS (to the internet or to other cloud providers) is priced per GB, and it adds up in ways that aren't visible until too late:
- S3 to internet: $0.09/GB (first 10TB/month)
- EC2 to internet: $0.09/GB (first 10TB/month)
- Cross-region transfer: $0.02/GB
A company with 100TB/month of S3-to-internet egress pays $9,000/month in egress fees alone — not compute, not storage, just moving data out.
Egress cost reduction strategies:
- CloudFront CDN caches frequently accessed content at edge; egress from CloudFront to internet is $0.0085–$0.085/GB vs. $0.09 from S3 directly
- S3 Transfer Acceleration for inbound; minimize cross-region replication
- VPC endpoints for AWS-internal traffic (free; avoids internet routing costs)
The 40% Reduction Roadmap
Organizations that systematically tackle cloud waste typically achieve 30–45% cost reduction over 90 days:
- Week 1–2: Enable AWS Cost Explorer, set up billing alerts, pull Compute Optimizer report
- Week 3–4: Rightsize top 10 over-provisioned instances (highest spend × lowest utilization)
- Month 2: Purchase 1-year Compute Savings Plans for stable baseline load
- Month 2: Schedule automatic shutdown of non-production environments
- Month 3: Audit and delete idle resources (EBS volumes, old snapshots, idle load balancers)
- Month 3: Implement CloudFront for high-egress S3 buckets
This is not advanced cloud engineering. It's operational hygiene — and the ROI is among the highest of any engineering investment a company can make.
Calculate It Yourself
The Cloud Cost Calculator models your monthly AWS/GCP/Azure spend across compute, storage, and data transfer — and estimates your savings potential from rightsizing, reserved pricing, and idle resource elimination. Run it against your current bill.