1. Packages
  2. Packages
  3. NVIDIA AI Cluster Runtime (AICR) Provider
Viewing docs for NVIDIA AI Cluster Runtime v0.1.6
published on Thursday, May 21, 2026 by Pulumi Labs

NVIDIA AI Cluster Runtime (AICR) Provider

nvidia-aicr logo
Viewing docs for NVIDIA AI Cluster Runtime v0.1.6
published on Thursday, May 21, 2026 by Pulumi Labs

    A Pulumi component provider that installs NVIDIA’s GPU software stack on Kubernetes. The ClusterStack component takes an accelerator, intent, and target platform, and installs the corresponding Helm charts in dependency order.

    Installation

    The AICR provider is available as a package in all Pulumi-supported languages:

    Example Usage

    The example below deploys an AICR recipe against an existing Kubernetes cluster (here, a local kind cluster suitable for development; replace service and provide a real kubeconfig for a cloud target).

    import * as aicr from "@pulumi-labs/nvidia-aicr";
    
    const stack = new aicr.ClusterStack("aicr", {
        accelerator: "h100",
        service: "kind",
        intent: "inference",
    });
    
    import pulumi_labs_nvidia_aicr as aicr
    
    aicr.ClusterStack(
        "aicr",
        accelerator="h100",
        service="kind",
        intent="inference",
    )
    
    package main
    
    import (
        aicr "github.com/pulumi-labs/pulumi-nvidia-aicr/sdk/go/nvidiaaicr"
        "github.com/pulumi/pulumi/sdk/v3/go/pulumi"
    )
    
    func main() {
        pulumi.Run(func(ctx *pulumi.Context) error {
            _, err := aicr.NewClusterStack(ctx, "aicr", &aicr.ClusterStackArgs{
                Accelerator: "h100",
                Service:     "kind",
                Intent:      "inference",
            })
            return err
        })
    }
    
    using Pulumi.Labs.NvidiaAicr;
    
    return await Pulumi.Deployment.RunAsync(() =>
    {
        var stack = new ClusterStack("aicr", new ClusterStackArgs
        {
            Accelerator = "h100",
            Service     = "kind",
            Intent      = "inference",
        });
    });
    
    import com.pulumi.Pulumi;
    import com.pulumi.labs.nvidiaaicr.ClusterStack;
    import com.pulumi.labs.nvidiaaicr.ClusterStackArgs;
    
    public final class App {
        public static void main(String[] args) {
            Pulumi.run(ctx -> {
                new ClusterStack("aicr", ClusterStackArgs.builder()
                    .accelerator("h100")
                    .service("kind")
                    .intent("inference")
                    .build());
            });
        }
    }
    
    resources:
      aicr:
        type: nvidia-aicr:index:ClusterStack
        properties:
          accelerator: h100
          service: kind
          intent: inference
    

    For cloud-targeted scenarios (EKS, AKS, GKE, OKE, CoreWeave), see the runnable per-language programs under examples/ in the provider repository.

    Configuration

    ClusterStack is a Pulumi component resource; the resolved set of Helm components is determined by the three criteria fields (accelerator, service, intent) plus the recipe overlay system. See the API Reference (left navigation) for the full input surface, including componentOverrides for per-component Helm value customization.

    nvidia-aicr logo
    Viewing docs for NVIDIA AI Cluster Runtime v0.1.6
    published on Thursday, May 21, 2026 by Pulumi Labs
      Try Pulumi Cloud free. Your team will thank you.