Scale AI workloads on any hardware with Cast AI support for TPUs

Cast AI now supports automated provisioning and scale-to-zero for Google Cloud TPU v5e and v5p slices on GKE. Teams can treat TPUs as standard compute resources managed through a single control plane alongside CPUs, GPUs, and AWS Trainium/Neuron. The integration requires no additional device plugins — pod specs with standard node selectors and tolerations trigger automatic TPU slice provisioning. Current limitations include single-host slices only and GKE exclusivity at launch, with broader topology support planned throughout the year.