EKS Auto Mode: Kubernetes Without DevOps on AWS (2026 guide) | OptimyCloud

EKS Auto Mode: Kubernetes has never been this simple on AWS

March 11, 2026 12 min read Alexandre Gillon
Kubernetes EKS Auto Mode on AWS

If you've ever run a Kubernetes cluster in production, you know the pain: managing nodes, updating AMIs, configuring Karpenter, tuning the VPC CNI, installing the EBS CSI driver... All of that before you even deploy your first application.

With EKS Auto Mode, launched by AWS in late 2024 and mature in 2026, this complexity disappears. AWS automatically manages the nodes, scaling, networking and storage. You only have to worry about your workloads. It's the same philosophy as serverless on AWS: delegate the infrastructure to focus on value. Kubernetes has finally delivered on its promise: focus on applications, not infrastructure.

In short

EKS Auto Mode = EKS + Karpenter + VPC CNI + EBS CSI + CoreDNS + kube-proxy, all managed by AWS. Zero add-ons to install, zero node groups to configure. You create a cluster, you deploy your pods.

The problem: Kubernetes is powerful, but it's heavy

A classic EKS cluster requires managing a considerable stack of tools just to get the infrastructure working:

  • Managed Node Groups or Karpenter to provision the EC2 instances
  • VPC CNI add-on for pod networking
  • EBS CSI driver for persistent volumes
  • CoreDNS and kube-proxy to keep up to date
  • Regular AMI updates + node rotations
  • IAM roles for each component (IRSA, Pod Identity)

The result: entire teams spend their time maintaining the platform instead of delivering value. SMEs and startups that want to use Kubernetes are often put off by this complexity — all the more so when they're also looking to integrate generative AI into their processes.

EKS Auto Mode: AWS does the work for you

EKS Auto Mode is a new operating mode for Amazon EKS where AWS manages the entire data plane. Concretely, when you enable Auto Mode:

Compute

AWS provisions and scales the nodes automatically with the built-in Karpenter technology. Zero node groups to configure.

Networking

VPC CNI, kube-proxy and CoreDNS are managed and updated by AWS. Pods get native VPC IPs.

Storage

The EBS CSI driver is pre-installed. PersistentVolumeClaims work immediately with the auto StorageClass.

Load Balancing

The AWS Load Balancer Controller is built in. Ingress and Services of type LoadBalancer work natively.

Create an EKS Auto Mode cluster in 5 minutes

This is where it gets impressive. A working EKS Auto Mode cluster is created with a single command:

# Create an EKS Auto Mode cluster with eksctl
eksctl create cluster \
  --name mon-cluster-auto \
  --region eu-west-3 \
  --enable-auto-mode
# That's it. No node group, no add-ons.
# Deploy an app directly:
kubectl apply -f mon-deployment.yaml

With Terraform, it's just as simple:

# main.tf - EKS Auto Mode cluster
resource
"aws_eks_cluster"
"main"
{
name = "mon-cluster-auto"
role_arn = aws_iam_role.eks.arn
version = "1.32"
vpc_config {
subnet_ids = var.subnet_ids
}
compute_config {
enabled = true
node_pools = ["general-purpose", "system"]
}
kubernetes_network_config {
elastic_network_interfaces {
enabled = true
}
}
storage_config {
block_storage {
enabled = true
}
}
}

Comparison

A classic EKS cluster with Karpenter, VPC CNI, EBS CSI and ALB Controller requires ~300 lines of Terraform and 4-5 IAM roles. With Auto Mode: ~30 lines and 1 role.

Classic EKS vs EKS Auto Mode

Feature Classic EKS EKS Auto Mode
Node provisioning Managed Node Groups or Karpenter Automatic
AMI updates Manual or rolling updates Automatic
VPC CNI Add-on to install + IAM Included and managed
EBS CSI Add-on + IAM + StorageClass Included and managed
Load Balancer Controller Helm chart + IAM + IngressClass Included and managed
CoreDNS / kube-proxy Versions to manage manually Automatic
Scaling Cluster Autoscaler or Karpenter Built-in Karpenter
Compute cost EC2 on-demand/spot/RI ~40% more expensive (vCPU/h)
Setup time 2-5 days 15 minutes

The real math: total cost vs EC2 cost

Yes, EKS Auto Mode compute is more expensive than raw EC2. But the real math has to include the total cost of ownership (TCO):

Classic EKS (hidden costs)

  • 2-5 days of initial setup (~EUR 2,000-5,000)
  • 1-2 days/month of maintenance (AMI, add-ons)
  • Scaling incidents, zombie nodes
  • Over-provisioning for safety

EKS Auto Mode (all included)

  • Setup in 15 minutes
  • Zero node/add-on maintenance
  • Optimal bin-packing = less waste
  • Team focused on business value

My field experience

For a startup with 5-20 microservices, EKS Auto Mode often works out cheaper overall than a classic EKS, because the over-provisioning of a poorly tuned cluster costs more than the per-vCPU premium. Not to mention the engineering days saved.

Migrating an existing cluster to Auto Mode

Good news: you don't need to recreate your cluster. The migration is done gradually, with no downtime. Last month we migrated a classic 47-pod EKS cluster to Auto Mode for a SaaS client: setup time dropped from 3 days to 2 hours, and the monthly compute bill fell by 15% thanks to optimized bin-packing.

1

Enable Auto Mode on the existing cluster

Via the AWS console or Terraform. The existing node groups keep running.

2

Migrate the workloads gradually

Cordon + drain the old nodes. The pods are rescheduled onto the Auto Mode nodes.

3

Remove the old node groups and add-ons

Once all the workloads are migrated, delete the managed node groups, Karpenter, and the obsolete add-ons.

4

Clean up the unused IAM roles

The IRSA roles for VPC CNI, EBS CSI and ALB Controller are no longer needed.

# Step-by-step migration
kubectl cordon <old-node>
kubectl drain <old-node> --ignore-daemonsets --delete-emptydir-data
# The pods restart on the Auto Mode nodes
eksctl delete nodegroup --cluster mon-cluster --name old-ng

The limitations to know about

EKS Auto Mode is not a universal solution. Here are the cases where classic mode is still relevant:

  • No custom DaemonSets on the managed nodes — if you use Datadog, Falco or a specific agent as a DaemonSet, you'll need to adapt (sidecar or mixed node groups)
  • No SSH on the nodes — debugging is done via kubectl debug, CloudWatch Logs and Container Insights
  • Limited GPU support — heavy ML/AI workloads with specific GPUs still require classic node groups
  • No Bottlerocket or custom AMI — AWS uses its own optimized AMIs, which can't be modified

Who is EKS Auto Mode for?

Startups

Production Kubernetes without a dedicated platform team

SMEs

Migrating to Kubernetes without blowing up engineering costs

Dev teams

Focus on the apps, not on the Kubernetes infra

Frequently asked questions

What is EKS Auto Mode?

EKS Auto Mode is an Amazon EKS feature that fully automates the management of nodes, scaling, networking and storage. AWS manages the Kubernetes infrastructure for you.

How much does EKS Auto Mode cost?

0.0842 USD per vCPU/hour, about 40% more than EC2 on-demand. But the TCO is often lower thanks to the elimination of maintenance and over-provisioning.

Can you migrate an existing cluster?

Yes, by enabling Auto Mode on the existing cluster and gradually migrating the workloads. Managed node groups run in parallel during the transition.

Does EKS Auto Mode replace Karpenter?

EKS Auto Mode uses Karpenter under the hood, but you no longer have to install or configure it. The behavior is similar but fully managed.

What are the limitations?

No custom DaemonSets, no SSH on the nodes, limited GPU support, and no custom AMI. For these cases, classic mode remains recommended.

Read also

See our pricing →

Simplify your Kubernetes infrastructure

Migration to EKS Auto Mode, audit of your existing cluster or setup from scratch — I'll support you every step of the way.

Let's talk about your project