diff --git a/argocd/manifests/nvidia-device-plugin/kustomization.yaml b/argocd/manifests/nvidia-device-plugin/kustomization.yaml
index 102127f..a46edf6 100644
--- a/argocd/manifests/nvidia-device-plugin/kustomization.yaml
+++ b/argocd/manifests/nvidia-device-plugin/kustomization.yaml
@@ -10,4 +10,4 @@ resources:
 
 images:
   - name: nvcr.io/nvidia/k8s-device-plugin
-    newTag: v0.18.2
+    newTag: v0.19.0
diff --git a/docs/changelog.d/+nvidia-device-plugin-v0.19.0.infra.md b/docs/changelog.d/+nvidia-device-plugin-v0.19.0.infra.md
new file mode 100644
index 0000000..95abf25
--- /dev/null
+++ b/docs/changelog.d/+nvidia-device-plugin-v0.19.0.infra.md
@@ -0,0 +1 @@
+Upgrade nvidia-device-plugin from v0.18.2 to v0.19.0
diff --git a/docs/reference/services/frigate.md b/docs/reference/services/frigate.md
index 46363bd..000d4ee 100644
--- a/docs/reference/services/frigate.md
+++ b/docs/reference/services/frigate.md
@@ -74,6 +74,7 @@ A separate **frigate-notify** pod polls Frigate's webapi every 15 seconds for de
 
 ## Related
 
+- [[nvidia-device-plugin]] - GPU device plugin enabling CUDA access
 - [[ntfy]] - Push notification delivery
 - [[sifaka]] - NAS storage for recordings
 - [[observability]] - Prometheus metrics at `/api/metrics`
diff --git a/docs/reference/services/nvidia-device-plugin.md b/docs/reference/services/nvidia-device-plugin.md
new file mode 100644
index 0000000..7eb28d9
--- /dev/null
+++ b/docs/reference/services/nvidia-device-plugin.md
@@ -0,0 +1,26 @@
+---
+title: NVIDIA Device Plugin
+modified: 2026-03-27
+tags:
+  - service
+  - gpu
+---
+
+# NVIDIA Device Plugin
+
+Kubernetes device plugin that exposes NVIDIA GPUs to pods on [[ringtail]]. Required for GPU workloads like [[frigate]] (object detection) and [[ollama]] (LLM inference).
+
+## Quick Reference
+
+| Property | Value |
+|----------|-------|
+| **Namespace** | `nvidia-device-plugin` |
+| **Image** | `nvcr.io/nvidia/k8s-device-plugin` |
+| **Upstream** | https://github.com/NVIDIA/k8s-device-plugin |
+| **Manifests** | [argocd/manifests/nvidia-device-plugin/](https://forge.eblu.me/eblume/blumeops/src/branch/main/argocd/manifests/nvidia-device-plugin) |
+
+## Architecture
+
+Runs as a DaemonSet with `privileged` security context, mounting the host's device-plugins socket, CDI specs, and NVIDIA driver libraries. A `RuntimeClass` named `nvidia` is defined for pods that need GPU access.
+
+Time-slicing is configured with 2 replicas per GPU, allowing two pods to share a single physical GPU.
diff --git a/service-versions.yaml b/service-versions.yaml
index ccc062e..6821488 100644
--- a/service-versions.yaml
+++ b/service-versions.yaml
@@ -46,8 +46,8 @@ services:
 
   - name: nvidia-device-plugin
     type: argocd
-    last-reviewed: 2026-02-19
-    current-version: "v0.18.2"
+    last-reviewed: 2026-03-27
+    current-version: "v0.19.0"
     upstream-source: https://github.com/NVIDIA/k8s-device-plugin/releases
     notes: DaemonSet + RuntimeClass on ringtail for GPU workloads