EKS path, priced
Auto Mode, Karpenter, or node groups — which EKS path costs less?
Side-by-side monthly cost for MNG, Karpenter, Auto Mode, and Fargate using your pod vCPU, memory, utilization, and Spot-eligible share. us-east-1 list pricing.
- No signup
- Instant results
- ~3 min
Rates reviewed
Cluster & demand
us-east-1list rates, pricing as of 2026-10-17. Source: AWS public pricing.
Enter total pod requests across the fleet (not node capacity). Bin-pack efficiency differs by path.
Sum of pod CPU requests.
Utilization & Spot
Lower utilization means more headroom (and more nodes) on EC2 paths.
Applied as ~70% off On-Demand for that share (EC2 Spot / Fargate Spot).
Your EKS Path Comparison
Cheapest path
Lowest Monthly Path
live resultCost Breakdown
Full four-path table unlocks after email
See the Full Path Comparison
Control plane + compute for all four paths, ranked — emailed to you.
Path comparison unlocked!
Full ranking and notes below.
All Paths (cheapest first)
How to read this
- MNG: DIY nodes — more wasted capacity (~65% pack).
- Karpenter / Auto Mode: Better consolidation (~85% pack); Auto Mode is managed.
- Fargate: No nodes — wins for spiky or low-utilization pod shapes.
- Extended Support: Adds $0.60/cluster-hour on every path.
Want this applied to your real account?
EKS path choice is half rate card, half ops load. We size Auto Mode vs Karpenter vs Fargate against real CUR and utilization — then implement the winner. See our managed EKS decision guide for the ops trade-offs.
Talk to Platform Experts →Who This Tool Is For
Platform engineers and CTOs choosing between managed node groups, Karpenter, EKS Auto Mode, and Fargate — and needing a cost number before the architecture review.
Why We Built This Tool
Each path meters differently: node hours with packing waste vs pod vCPU/GB-hours. This model uses us-east-1 list rates as of 2026-10-17 with explicit bin-pack assumptions so the comparison is honest.
What Problem It Solves
- Four paths, one shape. Same pod demand priced four ways.
- Spot blend. Interruptible share applies ~70% off to that portion.
- Extended Support tax. Surfaced on every path so it does not hide.
- Ops vs dollars. Auto Mode and Karpenter can tie on compute — pick on ops.
Read the Auto Mode vs Karpenter decision guide or use the EKS pricing calculator.
Frequently Asked Questions
What instance type is assumed for EC2 paths?
A reference m6i.large ($0.096/hr On-Demand in us-east-1) with 2 vCPU / 8 GiB. Node count is derived from your vCPU demand, utilization, and path bin-pack efficiency.
Why is Karpenter cheaper than managed node groups?
This model assumes ~85% packing efficiency for Karpenter/Auto Mode vs ~65% for typical MNG ASGs — fewer wasted cores for the same pod demand. Your real gap depends on consolidation settings and pod topology.
Does EKS Auto Mode cost extra beyond EC2?
You still pay the EKS control plane and the EC2 instances Auto Mode launches. This tool models Auto Mode compute at Karpenter-class packing on list EC2 rates; the product value is managed operations, not a separate vCPU surcharge in this estimate.
When does Fargate win?
When utilization is low, demand is spiky, or you want zero node ops. Dense, steady fleets usually favor EC2 paths with Spot.
How is Spot modeled?
The interruptible percentage is billed at ~70% off On-Demand (EC2 Spot / Fargate Spot planning rate). Diversify pools in production.
How accurate is this?
Directional for architecture decisions using us-east-1 list rates as of 2026-10-17. It excludes EBS, data transfer, and add-ons (ALB Controller, monitoring).
