Amazon Elastic Kubernetes Service is AWS's managed Kubernetes offering. EKS provisions and manages the Kubernetes control plane while leaving worker nodes (where pods actually run) under customer control. The service integrates tightly with AWS IAM (authentication), VPC (networking), EBS (storage), and CloudWatch (observability).
EKS is one of three major managed Kubernetes options in the public cloud market (alongside GKE on Google Cloud and AKS on Azure). The choice between them is driven primarily by which cloud the rest of the workload runs in. Cross-cloud Kubernetes is possible but introduces substantial operational complexity.
For AI workloads, EKS is widely used to orchestrate GPU-based training jobs and inference services. The operational pattern — scheduling workloads across GPU-attached nodes, managing model loading and caching, handling autoscaling based on inference demand — has become a standard enterprise capability.