I ran this exact audit for a Series B SaaS company last quarter. In 14 minutes, we identified $47,000 in annual waste. It wasn't hidden. It wasn't complex. They just hadn't looked.
Cloud waste isn't a mystery. It's a visibility problem. And by the end of this post, you'll have the visibility you need.
27% of cloud spend is wasted on average (Flexera 2025). 84% of organizations struggle to manage cloud spend. The projected infrastructure cloud waste for 2025? $44.5 billion.
The Real Numbers
Before we dive into the audit, let's establish what's actually happening in the industry:
- 33% of enterprises now spend over $12 million annually on public cloud (Flexera)
- Cloud spend increased 28% year-over-year in 2024
- Organizations are exceeding their cloud budgets by 17% on average
- 66% of organizations report wasted spend due to idle or underused resources
But here's what most people miss: only 43% of developers have access to real-time data on idle cloud resources, and just 39% can see unused or orphaned resources. The people provisioning resources can't see the waste they're creating.
The 15-Minute Cloud Cost Audit
Open your cloud billing console. Set the view to "Last 30 Days" and "Group by Service." Now work through these five checks.
Step 1 (2 minutes): The Growth Spike Check
Look at month-over-month growth for each service. Any line item growing more than 20% month-over-month without a corresponding business event (product launch, customer surge) is suspect.
Red flag example: Your S3 storage grew 45% last month but your user base grew 5%. You're likely storing logs, backups, or temp files that should have been deleted.
Step 2 (3 minutes): The Compute Sizing Audit
Navigate to your compute instances (EC2, Compute Engine, or VMs). Sort by cost, descending. For each of your top 5 most expensive instances:
1. Check CPU utilization over the past 2 weeks
2. If average utilization is under 20%, you're overprovisioned
3. If peak utilization never exceeds 60%, you're significantly overprovisioned
Typical finding: 30-40% of instances run at under 20% utilization. Downsizing one m5.xlarge to m5.large saves $7,000/year per instance.
Step 3 (3 minutes): The Reserved Instance Gap
Look at your on-demand compute spend. For any workload running 24/7 (databases, core services, always-on applications), calculate:
On-demand monthly cost × 12 × 0.40 = Your potential annual savings
Reserved Instances and Savings Plans offer 30-72% discounts for 1-3 year commitments. If you have $10,000/month in stable on-demand compute, you're leaving $36,000-$86,000 on the table annually.
Quick check: If you've been running the same instance types for 6+ months without Reserved Instances, this is found money.
Step 4 (4 minutes): The Storage Graveyard
Navigate to block storage (EBS volumes, persistent disks). Sort by unattached/available status.
Every unattached volume is pure waste. I've seen companies with $8,000/month in unattached storage—volumes from deleted instances, failed experiments, or forgotten backups.
Next, check snapshot age. Snapshots over 90 days old with no associated AMIs or active volumes are candidates for deletion or archival to cold storage (60-90% cheaper).
Step 5 (3 minutes): The Environment Zombie Hunt
Filter resources by environment tags: dev, staging, test. Check runtime hours.
Non-production environments running 24/7 represent 10-15% of total cloud waste. Dev and staging should typically run business hours only (40 hours/week vs 168 hours/week = 76% savings).
Quick math: A dev environment costing $2,000/month running 24/7 costs $24,000/year. Scheduled to business hours only: $5,760/year. One policy change, $18,240 saved.
The Five Biggest Waste Categories (With Dollar Signs)
Here's where waste actually shows up, based on analysis of hundreds of cloud estates:
Idle Resources
Compute instances with near-zero utilization, databases with no connections, load balancers pointing to nothing. The silent majority of waste.
Overprovisioning
Instances sized for peak loads that rarely materialize. That c5.4xlarge "just to be safe" when a c5.xlarge would handle 95% of your traffic.
Orphaned Resources
Unattached volumes, abandoned snapshots, forgotten test environments, IP addresses from deleted instances. The digital equivalent of leaving lights on.
On-Demand vs. Reserved
Paying sticker price for predictable workloads. Like buying a car one day at a time instead of leasing it for a year.
Storage Tiers
Hot storage for cold data. Logs in standard S3 that should be in Glacier. EBS gp3 volumes for data that belongs in object storage.
Real Savings Benchmarks
To make this concrete, here are actual savings I've helped clients achieve:
- $340,000/year: 2,400-person enterprise. Primary issue: 340 unattached EBS volumes from a decommissioned legacy system, plus zero Reserved Instance coverage on $70,000/month in stable compute.
- $89,000/year: Mid-market SaaS. Primary issue: All dev/staging environments running 24/7, oversized production instances (averaging 12% CPU), and 18-month-old RDS snapshots in standard storage.
- $12,400/year: Small startup ($4,000/month cloud spend). Primary issue: Unused NAT Gateways in 3 VPCs ($3,600/year each), forgotten load balancers, and oversized RDS instances.
Notice the pattern: None of these required engineering work. No architecture changes. No migration projects. Just visibility and action.
Why This Keeps Happening
If cloud waste is so obvious, why does it persist? Three reasons:
1. The Developer-FinOps Disconnect
Developers provision resources to solve problems. They're incentivized on velocity, not cost. And as we saw, fewer than half have real-time visibility into whether their resources are actually being used.
When you don't see the bill, you don't optimize. It's not malice; it's human nature.
2. The "We'll Clean It Up Later" Fallacy
That temporary environment for testing? Still running. The spike-provisioned instances for Black Friday? Still at full scale in February. The experiment that didn't pan out? Resources orphaned, not deleted.
"Later" never comes. The backlog always has customer-facing work ahead of cleanup.
3. Fear of Breaking Something
This is the legitimate concern. Nobody wants to be the person who deleted the wrong volume. So caution wins, and waste accumulates.
The fix: Start with obvious wins. Unattached volumes. Stopped instances. Test environments everyone agrees are abandoned. Build confidence, then tackle the harder calls.
Your Next Steps
You now have the exact audit I use. Here's what to do with it:
- Schedule 15 minutes this week. Block it on your calendar. The audit takes less time than most standups.
- Document your findings. Screenshot the waste you find. Quantify it in dollars. This builds the case for action.
- Pick the top 3 fixes. Don't try to solve everything. The three highest-impact, lowest-risk items are your targets.
- Set a recurring reminder. Monthly for the first quarter, then quarterly. Waste creeps back in.
The average company can cut 20-30% of their cloud spend with zero performance impact. That's not cutting corners—that's stopping the bleeding.
For a company spending $50,000/month on cloud, a 25% reduction is $150,000 annually. That's a senior engineer's salary. That's product development budget. That's runway extension.
The money's already there. You just need to claim it.
Want a second pair of eyes on your cloud architecture?