On-Demand Kubernetes

Hyperstack Kubernetes clusters are built on GPU-enabled worker flavours and NVIDIA-optimised images, making them a strong fit for AI, ML and other compute-intensive workloads. Clusters come ready for training jobs, fine-tuning pipelines, inference services and general cloud-native applications.

Every cluster can be deployed on GPU worker flavours and includes NVIDIA-optimised drivers and base images. This ensures seamless execution of AI/ML workloads without manual GPU configuration, driver installation or compatibility troubleshooting.

Launch a fully configured Kubernetes infrastructure on demand in minutes with minimal setup. Hyperstack automates node provisioning, networking, OS configuration and driver installation, so teams can focus on building, training and scaling applications rather than managing infrastructure.

Clusters run on high-speed, low-latency networking designed for distributed computing and AI workloads. Bastion and load balancer nodes are publicly accessible, while master and worker nodes stay on an internal network. Firewall rules and bastion access provide a secure pattern for managing SSH access, traffic flow and exposed services.

Hyperstack On-Kubernetes supports node groups, a collection of worker nodes that share the same flavor and configuration. This lets you run diverse workloads within a single cluster by creating multiple node groups for different GPU or CPU flavors. Teams can isolate training, inference and general-purpose workloads efficiently, with node group creation and scaling available via the API.