Kubernetes autoscaling mechanisms optimize for resource utilization and not for economics. Horizontal Pod Autoscalers react to metrics, and node provisioners such as Karpenter react to scheduling pressure. But neither system has any concept of financial constraints, budget ceilings, or business priorities. In real-world production environments, this creates an architectural gap: infrastructure scales correctly, yet cost overruns still occur.

This talk explores the design patterns behind introducing a budget-aware control layer on top of Kubernetes. We will examine why cost signals are invisible to the control plane, how external financial state can be modeled as a control loop, and what tradeoffs emerge when economic governance interacts with scheduling decisions.
Topics: The architectural limitations of metric-driven autoscaling, Designing external controllers that enforce budget policies, Stability vs. savings tradeoffs, Priority-aware workload reduction strategies
Attendees will leave with practical architectural patterns and a deeper understanding of where financial governance fits, and does not fit, in cloud-native systems.