DOP 344: KubeCon EU 2026 Review

DevOps Paradox Episode 344 reviews KubeCon EU 2026, a landmark event where Kubernetes solidified its transformation from a container orchestrator into a full-fledged AI platform. The hosts, Darren and Victor, explore how AI workloads—especially inference—are now central to Kubernetes' evolution, with major contributions from NVIDIA (KAI scheduler, DRA drivers), Google (TPU DRA), and Red Hat (LLM-D for disaggregated inference). Key developments include the Gateway API becoming the de facto standard, replacing Ingress, and new sandbox projects like Highgrass (AI-native gateway) and Valero (Kubernetes backup). The episode dives into the implications of running AI workloads in micro VMs instead of containers due to security and escape concerns, and questions whether containers still matter. A major theme is the rise of agents as first-class users of platforms, with tools like Microsoft’s Holmes GPT and SUSE’s Rancher AI Crew enabling agent-driven infrastructure management. The hosts debate the maturity of agent autonomy, the need for deterministic guardrails, and the cultural shift required for platform engineering to scale. They also reflect on the conference’s scale (13.5k attendees), the absence of major AI vendors like OpenAI and Anthropic on the expo floor, and speculate that EU may now be the primary hub for major Kubernetes announcements due to geopolitical factors. Despite some nostalgia for the early Kubernetes excitement, the hosts agree KubeCon EU 2026 was one of the best, driven by the rapid pace of AI innovation and the platform’s maturation. Key takeaways include: 1) Kubernetes is now an AI platform first, with inference workloads reshaping infrastructure design; 2) The Gateway API is mature enough to replace Ingress, but deprecation timelines are lacking; 3) Agents are becoming the primary users of platforms, requiring new design paradigms; 4) Confidential computing and micro VMs are gaining traction, especially in the EU; 5) The future of platform engineering lies in enabling agents, not just humans; 6) The speed of AI innovation is accelerating, with companies like Anthropic releasing features daily; 7) The mainframe may outlive AI, highlighting the enduring nature of foundational tech. The episode ends on a note of cautious optimism: while AI is transforming Kubernetes, the real challenge remains cultural and organizational, not just technical.