For the past decade, the cloud native community has been the backbone of the world’s largest internet-scale systems. But what happens now that AI is taking over the workload?
Catching Up with Chris Aniszczyk
I had the absolute privilege of catching up with Chris Aniszczyk, CTO of the CNCF, at KubeCon in Amsterdam to talk about the evolution of our ecosystem.
With over 13,500 people gathered here, the energy is incredible. Chris highlighted a fascinating shift: while the last ten years were about helping companies like Uber, Booking.com, and major airlines scale their massive distributed systems, the frontier has officially changed.
We are now deeply in the era of Agentic AI.
The Infrastructure Shift
The new challenge is fundamentally different from what cloud native has tackled before:
- Previous decade — scaling stateless microservices, service mesh, observability, GitOps for web-scale applications
- Current frontier — running massive-scale, open-source AI inference and LLMs efficiently on Kubernetes
This requires solving entirely new architectural problems:
- GPU scheduling and multi-tenancy — efficiently sharing expensive accelerators across teams and workloads, something I covered in my KubeCon talk on multi-tenant GPUs on bare metal
- Model serving at scale — frameworks like NVIDIA Dynamo and NIM for disaggregated inference
- Agent orchestration — durable AI agents that can reliably execute multi-step workflows
- Observability for AI workloads — right-sizing inference and monitoring token economics
- Token economics and cost management — the inference gold rush is creating new financial models
But the community’s approach remains the same — open, collaborative, and decentralized.
End-Users Driving Innovation
What I loved most about Chris’s perspective is who is driving this innovation right now. It is not just vendors pushing products. It is a massive wave of end-users and practitioners coming together to share real-world ideas, lessons learned, and open-source solutions.
This was visible everywhere at KubeCon:
- Engineers from organizations like DUO exploring open-source networking solutions
- Companies like Redpill Linpro building sovereign cloud infrastructure on Kubernetes
- Practitioners like Clemens Scholz using AI as a daily development companion
- Startups like Stack8s rethinking multi-cloud control planes
The CNCF ecosystem now includes projects that directly address the AI infrastructure challenge:
- Kubernetes itself with device plugins and dynamic resource allocation for GPUs
- KServe for model serving
- Kubeflow for ML pipelines
- Kueue for job scheduling and quota management
- The new AI Conformance program defining what “AI-ready Kubernetes” means
13,500 and Growing
The sheer scale of KubeCon EU 2026 — 13,500 attendees, hundreds of sponsors, dozens of co-located events — reflects that cloud native is not slowing down. If anything, the AI wave is accelerating adoption as organizations realize that Kubernetes is the natural platform for running inference workloads.
A huge thank you to Chris and the entire CNCF team for organizing such a monumental event and continuing to foster this incredible ecosystem.
Related Posts
- KubeCon EU 2026 in Numbers
- Packed Room: Multi-Tenant GPUs on Bare Metal
- NVIDIA Dynamo: Open Source Inference Framework
- The Inference Gold Rush
- AI on Kubernetes: The First 90 Days
About the Author
I am Luca Berton, AI and Cloud Advisor. I help enterprises run AI inference on Kubernetes at scale. Book a consultation.
