Kubernetes GPU Cost Attribution
April 3rd, 2026
GPU reservation costs are now attributed to the individual Kubernetes workloads that reserved them, instead of landing in unattributed cost pools. GPU requests (such as nvidia.com/gpu) are factored into Kubernetes cost attribution alongside CPU and memory, giving you accurate per-workload and per-team GPU spend.
This update is automatic for organizations running the CloudZero Kubernetes Agent v1.2.0 or later and requires no action.