KubeCon + CloudNativeCon North America 2026 runs November 9–12 in Salt Lake City, with co-located events on Monday the 9th and the main conference from Tuesday through Thursday. The CNCF published the full schedule in August, and the headline change is a new AI Inference + Agentic track. For platform teams, the question is less “what is new” and more “which four days of sessions change a decision we have to make before the end of the year.”
This preview sorts the program by that test. Everything below is drawn from the CNCF’s published schedule announcement and its “your week in Salt Lake City” guide. Session lists change between now and the event, so treat specific talks as likely rather than guaranteed.
The Shape of the Week
The event is in person only. The CNCF says there is no virtual component, and breakout and keynote recordings go to the CNCF YouTube channel after the event. If your team is sending one person and sharing the rest from recordings, plan around that delay.
Pass type matters more than it used to. The All-Access pass includes Monday’s CNCF-hosted co-located events plus the three main days. The KubeCon + CloudNativeCon Only pass covers the main conference and Monday’s Project Lightning Talks, but not the co-located events. For a platform team, several of the most practical sessions of the week sit on Monday, so the cheaper pass can be a false economy. Check current pricing on the Linux Foundation registration page; standard registration closed on September 2, so the price you see now is not the price quoted in early coverage.
The main conference runs at the Salt Palace Convention Center downtown, according to exhibitor listings and trade coverage of the event.
Monday: The Co-Located Events Are the Platform Engineer’s Agenda
The CNCF’s lineup of hosted co-located events for November 9 reads like a platform team’s tooling inventory:
- Platform Engineering Day and BackstageCon — internal developer portals and self-service workflows
- ArgoCon and FluxCon — GitOps delivery, on both of the CNCF’s graduated GitOps projects
- OpenTofu Day — infrastructure as code after the Terraform license change
- CiliumCon — eBPF networking, network policy and service mesh without sidecars
- Observability Day and Open Source SecurityCon
- Cloud Native AI + Inference Day and Agentics Day: MCP + Agents
- Kubernetes on Edge Day
If you can only send one person to one Monday track, pick the one tied to a decision on your roadmap. A team choosing between Argo CD and Flux gets more from ArgoCon and FluxCon on the same day than from any keynote. A team still on Terraform with a renewal coming up should look hard at OpenTofu Day; our OpenTofu vs Terraform breakdown covers the background.
The New AI Inference + Agentic Track
The CNCF describes the new track as covering Kubernetes, AI inference, agentic workflows, GPU scheduling, model serving and observability for production AI systems. Projects named in the announcement include vLLM, KServe, Ray and OpenTelemetry. One highlighted session is “Kubernetes Solutions for Agent-Shaped Problems” from Tim Hockin and Dmitry Berkovich of Google.
The Wednesday Maintainer Track adds AI-aware networking, AI workload infrastructure, GPU and specialized device support, and AI/ML orchestration.
What to actually take away
Be skeptical of AI sessions that stay at the level of architecture diagrams. The useful ones answer three questions:
- How are GPUs shared and scheduled? Dynamic Resource Allocation has changed what is possible here; we covered the state of play in GPU scheduling on Kubernetes.
- What does inference cost per request, and how does autoscaling behave when a model takes tens of seconds to load?
- How is an agent’s behavior observed and constrained once it can call tools inside your cluster?
A talk that cannot answer at least one of these with numbers from a real deployment is a vendor pitch. If your organization is not running inference on Kubernetes today, this track is optional. If it is, this is the one area of the program most likely to change your next-quarter plan.
Platform Engineering and Security Tracks
The schedule keeps dedicated Platform Engineering and Security tracks. The CNCF frames platform engineering sessions around internal developer platforms, self-service workflows, automation and operational practice, and security sessions around supply chain security, identity, runtime protection, vulnerability management and policy enforcement.
Two practical filters help here.
Look for sessions with failure stories. The CNCF singled out EarnIn’s “When Infrastructure Breaks in Production” talk on testing across the CNCF stack. Postmortem-style talks are where you learn which tools hold up, which is worth more than another reference architecture.
Map security talks to a control you own. “Gone in 60 Minutes,” from ARMO and fusioncore.ai speakers, is about closing the exploitable window with detection as code. If your team owns runtime detection, it is directly relevant. If you do not, a policy-enforcement or supply-chain session tied to your admission control is a better use of the slot.
Maintainer Tracks and ContribFests
Maintainer Track sessions run all three main days and are where you hear about roadmap direction from the people writing the code. On the published schedule:
- Tuesday: Kubernetes autoscaling, device management, security, auth, developer experience, Helm, Flux, Kyverno, KubeVirt and others
- Wednesday: the AI infrastructure topics above, plus OpenCost, Cilium, etcd, Linkerd, Karmada and Kubernetes SIGs
- Thursday: SIG Storage, containerd, SPIFFE, WG Batch, Rook, SIG Multicluster, Open Policy Agent and Falco
Two items deserve a bookmark for cost-aware teams. The OpenCost maintainer session is the closest thing to a FinOps track at the event, and the SIG Multicluster session matters if you run more than a handful of clusters.
One Topic That Will Come Up Everywhere: Ingress-NGINX
The Kubernetes project retired Ingress-NGINX in March 2026. Existing installations still run, but there are no further releases or security fixes from the project. Expect the hallway track, the Gateway API sessions and many vendor booths to circle back to this. If your clusters still route through Ingress-NGINX, use the week to talk to the maintainers and implementers of Gateway API controllers, not just to collect brochures. Our Ingress-NGINX to Gateway API playbook covers the migration mechanics.
How to Get Value If You Are Not Going
Most teams will send nobody, or one person. In that case, the best return comes from preparation rather than attendance:
- Pick three sessions per engineer from the published schedule that map to a live decision, and watch the recordings within two weeks of posting.
- Assign one person to scan Maintainer Track recordings for deprecations and roadmap changes in projects you run in production.
- If someone does attend, give them a short list of questions to ask maintainers directly. The hallway conversation with the person who wrote the controller you depend on is the one thing a recording cannot replace.
KubeCon is most useful when you arrive with a decision to make. Write down the two or three choices your platform team has to settle by the end of Q1 — portal, GitOps tool, ingress controller, GPU strategy — and build the week, or the recording queue, around those. Everything else is optional.
Sources: CNCF schedule announcement (August 2026) and CNCF event guide “Your week in Salt Lake City” (September 2026) at cncf.io; Kubernetes project statement on Ingress-NGINX retirement at kubernetes.io.



