DEV Community

Hive80-lab
Hive80-lab

Posted on

Quick Tip: How to Cut AWS Bills by 40% in One Weekend (5 Practical Steps)

Quick Tip: How to Cut AWS Bills by 40% in One Weekend (5 Practical Steps)

Problem: Cloud bills are eating your startup's runway

Starting a business is hard enough. Watching your AWS bill eat into your runway is worse.

I worked with a company that was spending $12,000/month on infrastructure and couldn't explain why. After a 3-hour audit, we cut their AWS bill by 40% with zero downtime.

Here are the 5 steps I used — you can do them this weekend:


1. Find the invisible costs with AWS Cost Explorer

Go to Cost Explorer → create a custom date range (last 30 days).

Look for:

  • Cost by Service (what's consuming the most?)
  • Usage by Instance Type (are you paying for expensive instances?)
  • Free Tier Expiration Dates (you're paying for what you should get free)

Quick fix: If you see "Reserved Instances" with 0% utilization, turn them off. You're wasting money.


2. Kill the zombie EC2 instances

# Find running instances older than 7 days
aws ec2 describe-instances --query 'Reservations[].Instances[?LaunchTime<`7d ago`].InstanceId' --output text
Enter fullscreen mode Exit fullscreen mode

3 quick ways to identify zombies:

  1. Launch Time filter in EC2 console (older than 3 days = likely dead)
  2. Tag all instances with "Environment" = "production/staging/dev"
  3. CPU Utilization check: anything under 5% for 48+ hours = kill it

Rule: If you haven't touched it in 7 days, turn it off. If you need it, you'll notice.


3. Optimize storage (EBS volumes are sneakier than you think)

Check your EBS costs:

aws ec2 describe-volumes --query 'Volumes[?Attached==`true` && UsagePrice>0].{VolumeId:VolumeId,Type:VolumeType,Usage:UsagePrice}' --output table
Enter fullscreen mode Exit fullscreen mode

3 optimization moves:

  1. Stop in-use volumes that aren't accessed (S3 already saves your data)
  2. Snapshot policy: Auto-snapshot daily, keep 7 days (cheaper than keeping volumes attached 24/7)
  3. Volume types: If you have IOPS < 3,000, switch from Provisioned IOPS (I3) to General Purpose (GP3) — up to 63% cheaper

4. Enable Auto Scaling (downtime doesn't cost you money)

Most companies set a fixed number of instances, then panic when traffic spikes.

Fix: Set minimum/maximum instances based on historical traffic patterns.

Example (Node.js API serving 10K req/min):

  • 3 instances minimum (baseline)
  • 10 instances maximum (peak)
  • Auto Scaling rules: scale out when CPU > 70%, scale in when CPU < 30%

Result: You pay for what you need, not for idle capacity.


5. Enable AWS Budgets and set alerts

Create a budget:

  • Go to AWS Billing → Budgets → Create Budget
  • Set your monthly limit (e.g., $3,000 for a $12K bill)
  • Add an alert: if you spend 80% of the budget, get an email

Golden rule: Never wait for a bill you can't afford. Budgets give you 2 weeks' runway to adjust.


Total savings: $4,800/month (40%)

If you're spending $12K/month on infrastructure, this weekend fix = $4,800/month in runway.

That's one less time you need to raise capital.


Tools I recommend

Goal Tool
Auto-savings rules AWS Trusted Advisor
Cost monitoring CloudWatch Cost Explorer + CloudZero
Enforce budgets AWS Budgets (set alerts)
Power usage Cloud Custodian (for bulk tag updates)
Negotiating AWS Discounts Manager (reserved instances)

Quick checklist for this weekend:

  • [ ] Run Cost Explorer (last 30 days)
  • [ ] Identify zombie EC2 instances (turn off 3-5)
  • [ ] Review EBS volumes (stop unused ones)
  • [ ] Enable Auto Scaling if you're manually scaling
  • [ ] Set up AWS Budgets alert at 80%

Time spent: 3-5 hours

Money saved: $1,500 - $5,000/month (depends on your scale)


Want to turn this into a reusable playbook? I've built a complete AWS cost audit checklist — grab it from my profile.

P.S. The startup I worked with also reduced their bill by another 20% after we enabled **Reserved Instances* for workloads that run 24/7. Let me know if you want that step too.*

follow @hive80lab on Dev.to for more ops automation playbooks.


I also keep a free copy of the first-30-minutes incident runbook (what to check, log, and escalate when an unattended agent breaks) here: https://hive80lab.gumroad.com/l/first-30-minutes

Top comments (0)