On November 9, 2026, KubeCon + CloudNativeCon North America will open its doors at the Salt Palace Convention Center in Salt Lake City. The flagship event of the Cloud Native Computing Foundation (CNCF) arrives at a moment when the cloud-native world is fundamentally realigning: AI has evolved from experiment to production workload, and the program is responding precisely to that. For the first time, there is a dedicated track for "AI Inference + Agentic" — a clear signal that Kubernetes is no longer just orchestrating containers, but is increasingly becoming the operations platform for inference systems and autonomous agents.

The New Track: AI Inference + Agentic

The CNCF had already published the full program schedule in August. According to the foundation, the new "AI Inference + Agentic" track is dedicated to operating production AI: GPU scheduling, model serving, dynamic routing, and observability across inference pipelines. Specifically mentioned are vLLM, KServe, Ray, and OpenTelemetry — projects that show how deeply cloud-native infrastructure is growing into AI operations.

CNCF Executive Director Jonathan Bryce articulated the direction in the announcement: AI is becoming one of the largest compute workloads in the industry, and the real engineering challenge today is no longer training but operations: "Kubernetes wasn't built for AI, but this community has spent years hardening it for exactly this kind of distributed, always-on workload."

One program item stands out: "Kubernetes Solutions for Agent-Shaped Problems" by Tim Hockin and Dmitry Berkovich (Google). Alongside it, a new project called "Agent Substrate" is in the program, which, according to the session description, aims to operate large numbers of agents securely and with extremely low latency on Kubernetes.

The Numbers Behind the Program

The focus is not an end in itself. According to the CNCF Annual Cloud Native Survey, 82 percent of container users run Kubernetes in production — and 66 percent of organizations with generative AI workloads rely on Kubernetes for that. The conference itself has become the largest open-source gathering in the world: At KubeCon EU 2026 in Amsterdam, the event counted more than 13,500 participants from 100 countries. For Salt Lake City, the Linux Foundation expects around 9,000 attendees, over 2,100 organizations, and more than 400 sessions over four days.

That the "industrialization of AI" is not just a buzzword was already demonstrated by Uber's appearance at KubeCon EU 2026: over 200 Kubernetes clusters, more than one million deployed workloads, around 20,000 models trained monthly, and over 30 million peak predictions per second — numbers that confirm the platform as a production system.

Platform Engineering and Security Remain Pillars

Alongside the AI track, the program focuses on established priorities. The Platform Engineering track covers internal developer platforms, self-service workflows, and automation — with projects like Backstage, Argo, and GitOps workflows. The Security track covers supply chain security, identity, runtime protection, and policy enforcement, among others with Cilium and eBPF-based observability; a session by ARMO and fusioncore.ai, for example, promotes "Detection as Code" to close the exploitable window after vulnerability discovery.

On Monday, November 9, the co-located events take place: Cloud Native AI + Inference Day, Agentics Day (MCP + Agents), ArgoCon, BackstageCon, CiliumCon, and WasmCon — a bundled deep dive that prepares the AI topics of the main program on the first day.

Uber Opens Michelangelo AI

One of the anticipated announcements comes from Uber: In a keynote on November 11, the company plans to release Michelangelo AI as open source — the Kubernetes-native ML platform that, according to the description, has been in production use for nearly a decade, runs more than 5,000 models, and delivers over 45 million predictions per second. The move follows the pattern established by many cloud-native projects: an internally proven platform becomes the foundation of an ecosystem. This is relevant for platform teams because Michelangelo AI could serve as a reference architecture for their own ML platforms.

Conclusion

KubeCon NA 2026 marks the transition from the question "Can Kubernetes do AI?" to the question "How do you operate AI reliably, securely, and cost-effectively on Kubernetes?" The new AI Inference track, Uber's open-source announcement, and the co-located events on agents show where the cloud-native landscape is heading: toward an infrastructure layer that treats inference workloads and agentic systems as standard loads. Those who cannot attend the conference do not have to miss out — recordings of keynotes and breakouts will be made available afterward via the CNCF YouTube channel. The topics of the week are likely to shape the Kubernetes roadmap for the coming months.

Sources