How Much Are You Overpaying on Cloud Databases?
Most teams overpay 30–40% on RDS, Aurora, Snowflake, and BigQuery. Move the sliders and see exactly where your money is going — and how much is recoverable.
Your Database Setup
Total monthly cost — compute, storage, I/O, and data transfer
Your database instance's average CPU over the last 30 days
% of your database spend committed to 1-year or 3-year pricing
CPU utilization from AWS CloudWatch: CPUUtilization metric, 30-day average. RI coverage from Cost Explorer → "Reservation Coverage" report.
Estimated monthly overspend
$1K
$1,402/month — $17K/year
CPU utilization health
Severely over-provisionedWhere the savings come from
Instance right-sizing
CPU at 25% — healthy range is 60–80%. Potential: go down one instance class.
Reserved Instance / Savings Plan gap
20% RI coverage. On-demand premium vs. 1-year RI: ~35% discount.
Storage and I/O optimization
gp2 → gp3 migration, orphaned snapshots, unused PIOPS provisioning.
Before vs After Optimization
What your bill looks like after right-sizing, RI commitment, and storage cleanup.
What the calculator measures
Instance right-sizing
Most teams provision for peak load and never revisit it. If your average CPU is under 40%, you're paying for capacity that sits idle 24/7. Going down one instance class typically saves 30–40% of compute cost.
Reserved Instance gap
On-demand pricing carries a 30–40% premium over 1-year Reserved Instances or Savings Plans. Most teams run steady-state workloads that qualify for commitments but never make them.
Storage and I/O waste
gp2 volumes cost 16% more per GB than gp3. Orphaned snapshots accumulate silently. Provisioned IOPS on lightly-loaded databases add hundreds per month for no benefit.
Not included in this estimate: Multi-AZ redundancy optimization, data transfer costs, read replica sizing, and query-level I/O reduction from index improvements. A full audit typically finds 35–45% total savings, not the 25–35% this calculator shows conservatively.
Frequently Asked Questions
How much do companies typically overspend on cloud databases?
Industry-wide, Flexera's State of the Cloud research puts average cloud waste at around 27% of total spend, with database overprovisioning cited as one of the largest contributors. In our own engagements, teams running on-demand instances without active cost governance typically overpay 30–40%, consistent with that broader industry figure.
What CPU utilization is healthy for an RDS or Aurora instance?
60–80% average CPU is generally considered the healthy range for a production database instance. Below 30% average utilization over 30 days signals significant over-provisioning — you're paying for compute capacity that mostly sits idle. Above 85–90% sustained is a genuine capacity risk, not just a cost issue.
How much do Reserved Instances actually save vs. on-demand pricing?
AWS RDS Reserved Instances can save around 30% with no upfront payment on a 1-year term, up to roughly 60% with partial upfront on a 3-year term, and up to about 69% with all-upfront 3-year commitments, depending on the database engine. The order matters: right-size your instance class first, then commit to Reserved Instance pricing — committing on an oversized instance just locks in the oversized rate at a discount.
Is gp3 storage always cheaper than gp2?
For nearly all workloads, yes. gp3 costs about 20% less per GB-month than gp2 (roughly $0.08 vs. $0.10 per GB in most US regions) and includes a 3,000 IOPS / 125 MB/s baseline at no extra charge, versus gp2's IOPS-scales-with-size model. Migrating from gp2 to gp3 is applied online with no downtime.
What isn't included in this calculator's savings estimate?
Multi-AZ redundancy optimization, data transfer costs, read replica right-sizing, and query-level I/O reduction from indexing improvements aren't modeled here — they require looking at your actual query patterns and topology. A full audit typically finds 35–45% total recoverable spend, higher than the conservative estimate this calculator shows.