The challenge
This media and entertainment leader runs a global business on a Kubernetes estate that spans multiple clouds. Like most large estates, it was significantly overprovisioned. Engineers padded resource requests to protect workloads. Waste compounded across clusters.
The hard part was not knowing waste existed. It was knowing where. Which clusters. Which nodes. By how much. First generation tooling could not answer those questions with the precision the platform team needed to act.
The solution
The customer deployed Kubex AI Agent and Kubex Kubernetes Rightsizing across its estate.
Kubex analyzed actual workload utilization, not self reported requests, and gave the team a clear line of sight into exactly where overprovisioning was hiding. For each cluster, Kubex identified the right node types based on observed behavior. Recommendations were not speculative. They were specific and executable.
The platform team stopped tuning and started shipping changes with confidence.
The results
- $1.2M in annual infrastructure savings
- Approximately 6,000 vCPU cores reclaimed
- Continuous, cluster level visibility into overprovisioning across a multi cloud Kubernetes estate
- Foundation in place to extend optimization to AI and GPU workloads next
In their words
“We’ve been testing the Kubex MCP server in our Kubernetes environments, and it promises to transform how our team accesses optimization recommendations. Having Kubex’s precision available to anyone simply by asking questions is a game changer and makes its power available to a much broader set of users.”
Sr. Director of Cloud Architecture, global media and entertainment leader
