diff --git a/docs/source/admin/installation.rst b/docs/source/admin/installation.rst
index 83b7b85..8386973 100644
--- a/docs/source/admin/installation.rst
+++ b/docs/source/admin/installation.rst
@@ -67,7 +67,7 @@ Installing the DCGM exporter is best handled using NVIDIA's `gpu-operator <https
     nvidia-driver-daemonset-fvx9z                                     1/1     Running     0               9d
     nvidia-operator-validator-62dhx                                   1/1     Running     0               14d
 
-.. tip::
+.. warning::
 
     This guide currently works for on-prem bare metal deployments.
     We are still validating on how to deploy :code:`nvidia-dcgm-exporter` 
diff --git a/docs/source/cloud/getting_started.rst b/docs/source/cloud/getting_started.rst
index 12a6ca7..15d7f02 100644
--- a/docs/source/cloud/getting_started.rst
+++ b/docs/source/cloud/getting_started.rst
@@ -58,4 +58,8 @@ can be requested as.
             kueue.x-k8s.io/queue-name: user-queue # this is assigned by your admin
             kueue.x-k8s.io/priority-class: low-priority
 
+.. warning::
 
+    Trainy instances are ephemeral and will be autoscaled down in 10 minutes of idling. Be sure if you are
+    running stateful applications like model training to instrument your application to regularly
+    retrieve and back up to object storage (S3, GCS, Azure Blob, Cloudflare R2, etc.)