AWS Cost Report, May 2026

$2,570/mo · $30,845/yr · 18 resources · confidence 70%

On-demand EC2/Fargate compute is $13,600/mo with only 18% of compute on a commitment. Covering ~70% of the on-demand baseline with a 1-yr no-upfront Compute Savings Plan (~27% off) saves about $2,570/mo. Verify the baseline is steady before committing.

Do this: Buy a 1-year no-upfront Compute Savings Plan sized to the steady baseline (check the 14-day usage floor first). Commit RDS separately with RDS Reserved Instances.

$1,416/mo · $16,992/yr · 16 resources · confidence 50%

NAT and cross-AZ transfer total $4,720/mo (NAT $3,320/mo). Routing AWS-bound traffic through VPC endpoints and co-locating chatty services in one AZ typically recovers ~30% (~$1,416/mo). Exact savings depend on the traffic mix.

Do this: Add free S3/DynamoDB gateway endpoints and interface endpoints for the busiest AWS services; check VPC flow logs for the top NAT talkers; co-locate high-chatter services in a single AZ.

$1,332/mo · $15,984/yr · 70 resources · confidence 55%

Snapshot spend is $3,330/mo against $4,400/mo of live volumes (76% ratio), well above the ~50% point where retention is usually unmanaged. Pruning the long tail typically reclaims ~40% (~$1,332/mo).

Do this: Apply a Data Lifecycle Manager policy (e.g. keep 7 daily / 4 weekly) and delete snapshots orphaned from deleted volumes. Confirm the prunable set against snapshot age before deleting.

$861/mo · $10,332/yr · 2 resources · confidence 65%

On-demand RDS/Aurora instance usage is $4,100/mo with 0% on a commitment. Reserving ~70% of the steady baseline with 1-yr no-upfront Reserved Instances (~30% off) saves about $861/mo. Compute Savings Plans do not cover RDS, so this spend is easy to miss.

Do this: Buy 1-year no-upfront RDS/Aurora Reserved Instances sized to the steady database baseline (match engine, class, and region). Confirm the instances run continuously before committing.

$640/mo · $7,680/yr · 29 resources · confidence 95%

29 gp2 volumes cost $3,200/mo. gp3 lists ~20% cheaper per GB-month at equal baseline performance, so the same capacity on gp3 saves about $640/mo.

Do this: Change the volume type from gp2 to gp3 (online, no downtime). gp3 also decouples IOPS/throughput from size if any volume needs more than the 3,000 IOPS baseline.

$227/mo · $2,724/yr · 18 resources · confidence 97%

18 Elastic IPs are billed as idle (allocated but not attached to a running instance), costing $227/mo. Idle EIPs are charged at the full hourly rate for doing nothing.

Do this: Release the idle addresses, or attach them to a running instance, NAT, or NLB if they are reserved for one. Most are leftovers from torn-down resources.

n/a/mo · n/a/yr · 12 resources · confidence 90%

$5,670/mo (14% of spend) has no team tag and cannot be charged back or owned. Top untagged services: S3 ($4,130), Athena ($820), VPC ($720).

Do this: Enforce a required team tag via tag policies and SCPs, backfill the top untagged resources, and add a budget alarm on untagged spend so it can't creep back up.