March 2025 was when everything went to hell. IBM bought HashiCorp and immediately killed the free tier that 500,000+ developers relied on. Now you pay for every single resource, every hour, whether you're actively developing or sleeping.
What \"Resource Under Management\" Actually Means
Each resource in your Terraform state costs money. I counted 63 resources in my personal dev environment yesterday - that's $38.47/month just for basic AWS infrastructure I barely use. Every data source lookup? That's a resource. Each one cost me $0.10/month to look up the same fucking Ubuntu AMI.
Here's what shocked me: data sources count as resources. Run terraform state list | grep data.
on your state - I had 47 resources I counted last week, but terraform state list | wc -l showed 63. The difference? Data sources for AMI lookups, VPC references, and availability zones.
The Terraform 1.6.0 Bug That Costs Money
There's a known issue where Terraform 1.6.0+ creates duplicate state entries for certain AWS resources. Each duplicate counts as a billable resource. I found this when my resource count jumped from 45 to 67 overnight after a routine update. The fix requires manual state editing:
terraform state rm aws_instance.duplicate_entry
terraform import aws_instance.web i-1234567890abcdef0
Pulumi's \"Credit\" System is Misleading
Pulumi markets 150k free credits like it's generous. One credit = managing one resource for one hour. Sounds like a lot until you do math:
- 150,000 credits ÷ 730 hours/month = 205 resources maximum
- But credits expire if unused, unlike actual storage
- Exceed your limit? They bill at $0.000685/hour immediately
CloudFormation's Deceptive \"Free\"
AWS CloudFormation appears free, but creates expensive surprises. The tool itself doesn't charge, but every resource it provisions bills at full AWS rates. The trap: CloudFormation's error recovery creates then destroys resources multiple times during failed deployments.
I watched a failed EKS cluster deployment burn through $47 in terminated instances before the stack finally errored out. CloudFormation kept trying to fix network policies by recreating NAT gateways ($45/month each) until hitting resource limits.
The Real Architecture Impact
I'm literally designing worse infrastructure to save on tool fees. Instead of proper IAM roles per service (12 resources), I reuse one role across everything (1 resource). Security best practices cost extra money now.
My team consolidated three environments into one because paying for dev, staging, and prod state management was $180/month before touching AWS costs. We're debugging production issues more because we can't afford proper testing environments.