Cast AI now supports Dynamic Resource Allocation (DRA) for Kubernetes 1.34+ on GKE and EKS, enabling intent-based GPU allocation instead of hardcoded GPU counts. DRA allows workloads to reference resource claims that describe their needs, while Cast AI's Autoscaler automatically provisions the right nodes, selects instance types, and finds available capacity. The system integrates with existing GPU sharing strategies, applies Spot optimization with automatic on-demand fallback, and uses OMNI to expand capacity search across regions and clouds when local supply is constrained.
Table of contents
Intent-based GPU allocation with DRAWhy DRA mattersHow Cast AI closes the loopFrom resource claim to running workloadStop managing GPUs. Start using them.Explore how Cast AI can optimize your GPU infrastructureSort: