EKS 1.31 extended support ends on November 26, 2026. If a cluster is still on 1.31 after that, AWS upgrades it for you. I reread the AWS docs this morning to check what that upgrade does, and two details are easy to miss.
It lands on 1.32, and 1.32 is already on extended support
AWS says the cluster is auto-upgraded to the oldest version still in extended support. Today that's 1.32. Standard support for 1.32 ended on March 23, 2026, so the cluster keeps paying $0.60 per hour. It doesn't go back to $0.10.
1.33 is on extended support too, and 1.34 joins it on December 2. After that date the first version at the standard price is 1.35. From 1.32 that's three more upgrades, because EKS moves one minor version at a time.
Only the control plane moves
The AWS FAQ says the auto-upgrade covers the control plane only. Managed node groups and self-managed nodes stay on the previous version. (EKS Auto Mode nodes may update on their own.)
So after the forced upgrade the API server is on 1.32 and the kubelets are still on 1.31. That's allowed. The Kubernetes version skew policy lets a kubelet be up to three minor versions older than the API server. Do the math and 1.31 nodes are supported under a 1.32, 1.33 or 1.34 control plane, and unsupported at 1.35. The nodes have to move before that hop.
One command
kubectl get nodes -o json | jq -r '.items[] | "\(.metadata.name) \(.status.nodeInfo.kubeletVersion)"' It's read-only and prints one line per node, the node name and its kubelet version. Compare the minor version with the cluster version shown in the EKS console. If the nodes are one or two versions behind, plan the node group upgrade before the next control plane hop, not after the last one.
If you're on 1.31 today, I'd do the 1.32 hop yourself before the 26th, on your own schedule, and upgrade the nodes right after.
kube-upgrade-check (free, open source) runs the same skew check against a target version if you don't want to eyeball it.