NVIDIA at KubeCon and CloudNativeCon Europe

NVIDIA is donating its Dynamic Resource Allocation Driver for GPUs to the Cloud Native Computing Foundation, enabling developers to manage high-performance AI workloads on Kubernetes with greater efficiency and scale. This open source contribution makes enterprise AI infrastructure more accessible through smarter GPU resource sharing, dynamic reconfiguration, and enhanced security for confidential computing.

Disaggregated serving architectures like DeepSeek-R1 split inference into specialized roles with different scaling needs, pushing beyond what Kubernetes primitives handle well. This guide compares deploying aggregated and disaggregated models using LeaderWorkerSet versus Grove, showing when to use each and how Grove eliminates orchestration complexity.

Discover NVIDIA’s open-source AI Cluster Runtime (AICR), which generates deterministic, version-locked Kubernetes recipes so you can standardize GPU clusters, streamline upgrades, and validate performance across any environment

Rolling coverage from San Jose, including NVIDIA CEO Jensen Huang’s keynote, breakout highlights, live demos and on‑the‑ground color through March 19.

The "Trust Gap" is holding back AI's move to production on shared infrastructure. Discover NVIDIA's Confidential Containers (CoCo) reference architecture, which uses hardware-backed TEEs to cryptographically protect proprietary Model IP and sensitive data during inference, maintaining Kubernetes agility.

From the data center and cloud to the desktop and edge, NVIDIA Cloud-Native Technologies provide the ability to run deep learning, machine learning, and other GPU-accelerated workloads managed by Kubernetes on systems with NVIDIA GPUs. They also allow the seamless deployment and development of containerized software on enterprise cloud-native management frameworks.

Discover how NVIDIA Dynamo 1.0 delivers low‑latency, high‑throughput, distributed inference for generative AI, making it easier to serve large models across multi‑GPU and multi‑node environments in production.

NVIDIA Dynamo is an open-source inference software for accelerating AI model deployment at AI-factory scale. Using disaggregated serving, Dynamo breaks inference tasks into smaller components, dynamically routing and rerouting workloads to the most optimal compute resources available at that moment.

KAI Scheduler is the open-source Kubernetes scheduler designed to manage large-scale GPU clusters and optimize resource allocation from interactive jobs to large-scale training and inference workloads. Built with advanced features like gang scheduling, hierarchical queues, and workload consolidation, it maximizes GPU utilization while maintaining fairness across teams.

NVIDIA drives open source innovation by releasing their technologies to the community. This equips developers—solo builders or scaling companies—with tools to create breakthrough applications using accelerated computing.