Rafay Achieves CNCF Kubernetes AI Conformance for v1.35 | Rafay

Why It Matters

The certification gives enterprises a vendor‑neutral benchmark for AI‑ready Kubernetes, reducing lock‑in risk and ensuring workloads can move across clouds with confidence. It also signals that Rafay will stay aligned with evolving AI‑specific Kubernetes primitives.

Key Takeaways

• Rafay MKS certified CNCF AI Conformance for Kubernetes v1.35.
• Certification validates DRA, gang scheduling, GPU autoscaling, observability, security.
• AI Conformance ensures workload portability across certified Kubernetes platforms.
• 31 platforms now certified, a 70% increase since program launch.
• Automated testing slated for 2026 will make conformance baseline.

Pulse Analysis

The rapid rise of AI and machine‑learning workloads has exposed a fragmentation gap in Kubernetes ecosystems. While the core platform excels at container orchestration, GPU‑intensive jobs demand fine‑grained accelerator allocation, gang scheduling, and hardware‑level metrics that many distributions implement inconsistently. To address this, the Cloud Native Computing Foundation introduced the Certified Kubernetes AI Conformance program in November 2025, defining a concrete checklist—known as the Kubernetes AI Requirements (KARs)—that any platform must satisfy to be deemed AI‑ready. The initiative builds on the existing Kubernetes Conformance suite, extending it with tests for dynamic resource allocation, secure GPU sharing, and observability standards aligned with OpenTelemetry.

Version 1.35 of the program, released at KubeCon Europe in March 2026, raised the bar with stricter validation of agentic workloads, stable in‑place pod resizing, and workload‑aware scheduling. Rafay’s Managed Kubernetes Service successfully met every mandatory KAR, demonstrating native support for Dynamic Resource Allocation, reliable gang scheduling via Kueue or Volcano, GPU‑aware autoscaling, and per‑accelerator telemetry exposed in Prometheus format. The public submission in the CNCF repository provides transparent evidence of these capabilities, reinforcing Rafay’s claim of delivering an enterprise‑grade, multi‑tenant AI platform that can run on bare‑metal or virtualized infrastructure without custom glue code.

For businesses, the certification translates into tangible benefits: portable AI workloads, reduced engineering overhead, and a clear procurement metric when comparing cloud providers or private clouds. As the CNCF prepares to roll out automated conformance testing in 2026, the certification will evolve from a self‑assessment to a continuously verified guarantee, making AI‑ready status a baseline requirement rather than a differentiator. Enterprises planning GPU‑as‑a‑Service offerings, sovereign clouds, or internal AI platforms should therefore prioritize vendors with AI Conformance certification, as it ensures alignment with the latest Kubernetes primitives and future‑proofs investments against the rapidly shifting AI infrastructure landscape.