Kubernetes in the Age of AI: GenAI & agentic workloads move to K8s

Kubernetes is emerging as the core platform for running AI systems in production, shifting from container orchestration to AI infrastructure. The CNCF reports that among container users, 82% use Kubernetes in production in 2025 (up from 66% in 2023). It also says Kubernetes is now a de facto AI platform: it provides a unified orchestration layer for both traditional workloads and compute-intensive AI tasks. For generative AI, the article highlights that model operations are challenging once teams move from experimentation to production—cost, scalability, resilience, security, and deployment complexity. Red Hat engineers note that LLMs are resource intensive and that Kubernetes fits well for pretraining, fine-tuning, deployment, and prompt engineering. CNCF data cited here shows that by 2025, 66% of organizations run generative AI workloads on Kubernetes. Named users include OpenAI (testing/experiments), Tesla (KServe for LLM inference), and Adobe, alongside Uber, Intuit, and Google. For agentic AI, Kubernetes is described as hosting ML pipelines and evolving agents from demos into fleets of agents. Examples of related tooling mentioned include Kagent, K8sGPT, Sympozium, and Agent Sandbox. Finally, the piece stresses fundamentals—especially Kubernetes networking. It references a new CNCF certification for Kubernetes network engineers, reflecting how mission-critical AI training and regulated workloads increase demand for networking expertise. Overall, the takeaway is that Kubernetes for AI is becoming operationally necessary, not optional.