diff --git a/docs/GPU.md b/docs/GPU.md
index 975af95..1e08064 100644
--- a/docs/GPU.md
+++ b/docs/GPU.md
@@ -148,7 +148,7 @@ The NVIDIA GPU Operator does not work on WSL2 as of version v24.3.0 (see [issue]
 
 To get around this issue, the recommended course of action is to install UDS K3d without the `cuda` flavor, and then deploy the NVIDIA Device Plugin separately. Below are the steps for doing so:
 
-1. Run `uds run default` or `uds zarf package deploy oci://justinthelaw/uds-k3d:${VERSION} --confirm`
+1. Run `uds run default --set K3D_EXTRA_ARGS="--gpus=all"` or `uds zarf package deploy oci://justinthelaw/uds-k3d:${VERSION} --confirm --set K3D_EXTRA_ARGS="--gpus=all"`
 2. Create an `nvidia-device-plugin.yaml` manifest like the one below, and a deploy it with `uds zarf tools kubectl apply -f nvidia-device-plugin.yaml`
 
   ```yaml
diff --git a/zarf.yaml b/zarf.yaml
index 0cd4c9e..d9cb542 100644
--- a/zarf.yaml
+++ b/zarf.yaml
@@ -56,7 +56,7 @@ components:
             setVariables:
               - name: K3D_IMAGE
 
-  - name: enable-cuda
+  - name: inject-cuda-image
     required: true
     only:
       flavor: cuda
@@ -69,6 +69,19 @@ components:
             setVariables:
               - name: K3D_IMAGE
 
+  - name: expose-gpus
+    required: true
+    only:
+      flavor: cuda
+    description: "Adds the extra K3d argument for exposing host GPUs to the cluster"
+    actions:
+      onDeploy:
+        before:
+          - cmd: |
+              echo "${ZARF_VAR_K3D_EXTRA_ARGS} --gpus=all"
+            setVariables:
+              - name: K3D_EXTRA_ARGS
+
   - name: create-cluster
     required: true
     description: "Create the k3d cluster"