NVIDIA announced on November 10, 2025, the availability of NVIDIA Grove within NVIDIA Dynamo as a modular component. NVIDIA Grove is a Kubernetes API designed to streamline modern Machine Learning (ML) inference workloads on Kubernetes clusters. Grove addresses the challenges of managing complex, multi-component AI inference systems by enabling users to scale multinode inference deployments, supporting tens of thousands of GPUs. Grove is fully open source and available on the ai-dynamo/grove GitHub repo. Key features of NVIDIA Grove include:

Multilevel Autoscaling: Scales individual components, related component groups, and entire service replicas.
System-Level Lifecycle Management: Manages recovery and updates on complete service instances.
Flexible Hierarchical Gang Scheduling: Supports policies that enforce minimum viable component combinations while enabling flexible scaling.
Topology-Aware Scheduling: Optimizes component placement based on network topology.
Role-Aware Orchestration: Enforces role-specific configuration and dependency for reliable initialization.

Grove orchestrates multicomponent AI workloads using three hierarchical custom resources in its Workload API:

PodCliques: Represent groups of Kubernetes pods with specific roles and independent configuration.
PodCliqueScalingGroups: Bundle tightly coupled PodCliques that scale together.
PodCliqueSets: Define the entire multicomponent workload, specifying startup ordering and scaling policies.

To get started with Grove, NVIDIA provides a step-by-step guide using Dynamo to deploy a disaggregated serving architecture with a KV-routing deployer, utilizing the Qwen3 0.6B model. Grove is fully open source and the company welcomes contributions, pull requests, and feedback from the community. NVIDIA will showcase Grove at KubeCon 2025 in Atlanta at Booth.