Kubernetes clusters commonly run at 10-12% CPU utilization due to architectural decisions prioritizing flexibility over efficiency. The platform's default scheduler spreads pods for resilience rather than density, autoscalers (VPA/HPA) conflict and cause restart loops, and static resource requests lead to overprovisioning. Tools like Cluster Autoscaler and Karpenter improve bin packing but remain reactive. Core constraints—no pod migration, request-based scheduling, independent autoscalers—prevent native optimization. Effective solutions require combining continuous rightsizing, predictive scaling, and intelligent consolidation beyond Kubernetes' current capabilities.