diff --git a/api/nvidia/v1/clusterpolicy_types.go b/api/nvidia/v1/clusterpolicy_types.go
index 12f8e9dcde..5fdac040dc 100644
--- a/api/nvidia/v1/clusterpolicy_types.go
+++ b/api/nvidia/v1/clusterpolicy_types.go
@@ -176,6 +176,14 @@ type ServiceMonitorConfig struct {
 	// +operator-sdk:gen-csv:customresourcedefinitions.specDescriptors=true
 	// +operator-sdk:gen-csv:customresourcedefinitions.specDescriptors.displayName="Relabelings allows to rewrite labels on metric sets"
 	Relabelings []*promv1.RelabelConfig `json:"relabelings,omitempty"`
+
+	// APIGroup is the API group used for ServiceMonitor CRDs.
+	// Defaults to monitoring.coreos.com. Set to azmonitoring.coreos.com on AKS
+	// clusters using Azure Managed Prometheus.
+	// +operator-sdk:gen-csv:customresourcedefinitions.specDescriptors=true
+	// +operator-sdk:gen-csv:customresourcedefinitions.specDescriptors.displayName="API group for ServiceMonitor CRDs"
+	// +operator-sdk:gen-csv:customresourcedefinitions.specDescriptors.x-descriptors="urn:alm:descriptor:com.tectonic.ui:text"
+	APIGroup string `json:"apiGroup,omitempty"`
 }
 
 // The Alias for backward compatibility
@@ -2446,6 +2454,21 @@ func (sm *ServiceMonitorConfig) IsEnabled() bool {
 	return *sm.Enabled
 }
 
+// GetAPIGroup returns the API group for ServiceMonitor CRDs.
+func (sm *ServiceMonitorConfig) GetAPIGroup() string {
+	if sm != nil && sm.APIGroup != "" {
+		return sm.APIGroup
+	}
+	return DefaultMonitoringAPIGroup
+}
+
+const (
+	// DefaultMonitoringAPIGroup is the default Prometheus Operator API group.
+	DefaultMonitoringAPIGroup = "monitoring.coreos.com"
+	// AzureMonitoringAPIGroup is used by Azure Managed Prometheus on AKS.
+	AzureMonitoringAPIGroup = "azmonitoring.coreos.com"
+)
+
 // IsNLSEnabled returns true if NLS should be used for licensing the driver
 func (l *DriverLicensingConfigSpec) IsNLSEnabled() bool {
 	if l.NLSEnabled == nil {
diff --git a/config/rbac/role.yaml b/config/rbac/role.yaml
index 7a631b9e3b..844f9498f6 100644
--- a/config/rbac/role.yaml
+++ b/config/rbac/role.yaml
@@ -117,6 +117,18 @@ rules:
   - patch
   - update
   - watch
+- apiGroups:
+  - azmonitoring.coreos.com
+  resources:
+  - servicemonitors
+  verbs:
+  - create
+  - delete
+  - get
+  - list
+  - patch
+  - update
+  - watch
 - apiGroups:
   - node.k8s.io
   resources:
diff --git a/controllers/object_controls.go b/controllers/object_controls.go
index 299da2a45b..cde6db64c9 100644
--- a/controllers/object_controls.go
+++ b/controllers/object_controls.go
@@ -39,6 +39,8 @@ import (
 	apiextensionsv1 "k8s.io/apiextensions-apiserver/pkg/apis/apiextensions/v1"
 	apierrors "k8s.io/apimachinery/pkg/api/errors"
 	metav1 "k8s.io/apimachinery/pkg/apis/meta/v1"
+	"k8s.io/apimachinery/pkg/apis/meta/v1/unstructured"
+	"k8s.io/apimachinery/pkg/runtime"
 	"k8s.io/apimachinery/pkg/runtime/schema"
 	"k8s.io/apimachinery/pkg/types"
 	"k8s.io/apimachinery/pkg/util/intstr"
@@ -5049,6 +5051,59 @@ func applyServiceMonitorCustomEdits(desiredState *gpuv1.ServiceMonitorConfig, cu
 	}
 }
 
+func serviceMonitorCRDName(apiGroup string) string {
+	if apiGroup == "" {
+		apiGroup = gpuv1.DefaultMonitoringAPIGroup
+	}
+	return fmt.Sprintf("servicemonitors.%s", apiGroup)
+}
+
+func serviceMonitorGVK(apiGroup string) schema.GroupVersionKind {
+	if apiGroup == "" {
+		apiGroup = gpuv1.DefaultMonitoringAPIGroup
+	}
+	return schema.GroupVersionKind{Group: apiGroup, Version: "v1", Kind: "ServiceMonitor"}
+}
+
+func serviceMonitorConfigForState(n ClusterPolicyController, state string) *gpuv1.ServiceMonitorConfig {
+	switch state {
+	case "state-dcgm-exporter":
+		if n.singleton.Spec.DCGMExporter.ServiceMonitor != nil {
+			return n.singleton.Spec.DCGMExporter.ServiceMonitor
+		}
+	case "state-operator-metrics":
+		if n.singleton.Spec.Operator.Metrics.ServiceMonitor != nil {
+			return n.singleton.Spec.Operator.Metrics.ServiceMonitor
+		}
+	}
+	return nil
+}
+
+func serviceMonitorAPIGroupForState(n ClusterPolicyController, state string) string {
+	if sm := serviceMonitorConfigForState(n, state); sm != nil {
+		return sm.GetAPIGroup()
+	}
+	return gpuv1.DefaultMonitoringAPIGroup
+}
+
+func toUnstructuredServiceMonitor(sm *promv1.ServiceMonitor, apiGroup string) (*unstructured.Unstructured, error) {
+	objMap, err := runtime.DefaultUnstructuredConverter.ToUnstructured(sm)
+	if err != nil {
+		return nil, err
+	}
+	u := &unstructured.Unstructured{Object: objMap}
+	u.SetGroupVersionKind(serviceMonitorGVK(apiGroup))
+	return u, nil
+}
+
+func deleteServiceMonitor(ctx context.Context, c client.Client, sm *promv1.ServiceMonitor, apiGroup string) error {
+	u, err := toUnstructuredServiceMonitor(sm, apiGroup)
+	if err != nil {
+		return err
+	}
+	return c.Delete(ctx, u)
+}
+
 // ServiceMonitor creates ServiceMonitor object
 func ServiceMonitor(n ClusterPolicyController) (gpuv1.State, error) {
 	ctx := n.ctx
@@ -5058,8 +5113,11 @@ func ServiceMonitor(n ClusterPolicyController) (gpuv1.State, error) {
 
 	logger := n.logger.WithValues("ServiceMonitor", obj.Name, "Namespace", obj.Namespace)
 
+	apiGroup := serviceMonitorAPIGroupForState(n, n.stateNames[state])
+	crdName := serviceMonitorCRDName(apiGroup)
+
 	// Check if ServiceMonitor is a valid kind
-	serviceMonitorCRDExists, err := crdExists(n, ServiceMonitorCRDName)
+	serviceMonitorCRDExists, err := crdExists(n, crdName)
 	if err != nil {
 		return gpuv1.NotReady, err
 	}
@@ -5069,7 +5127,7 @@ func ServiceMonitor(n ClusterPolicyController) (gpuv1.State, error) {
 		if !serviceMonitorCRDExists {
 			return gpuv1.Ready, nil
 		}
-		err := n.client.Delete(ctx, obj)
+		err := deleteServiceMonitor(ctx, n.client, obj, apiGroup)
 		if err != nil && !apierrors.IsNotFound(err) {
 			logger.Info("Couldn't delete", "Error", err)
 			return gpuv1.NotReady, err
@@ -5084,7 +5142,7 @@ func ServiceMonitor(n ClusterPolicyController) (gpuv1.State, error) {
 			if !serviceMonitorCRDExists {
 				return gpuv1.Ready, nil
 			}
-			err := n.client.Delete(ctx, obj)
+			err := deleteServiceMonitor(ctx, n.client, obj, apiGroup)
 			if err != nil && !apierrors.IsNotFound(err) {
 				logger.Info("Couldn't delete", "Error", err)
 				return gpuv1.NotReady, err
@@ -5138,11 +5196,17 @@ func ServiceMonitor(n ClusterPolicyController) (gpuv1.State, error) {
 		return gpuv1.NotReady, err
 	}
 
-	found := &promv1.ServiceMonitor{}
+	desired, err := toUnstructuredServiceMonitor(obj, apiGroup)
+	if err != nil {
+		return gpuv1.NotReady, err
+	}
+
+	found := &unstructured.Unstructured{}
+	found.SetGroupVersionKind(serviceMonitorGVK(apiGroup))
 	err = n.client.Get(ctx, types.NamespacedName{Namespace: obj.Namespace, Name: obj.Name}, found)
 	if err != nil && apierrors.IsNotFound(err) {
 		logger.Info("Not found, creating...")
-		err = n.client.Create(ctx, obj)
+		err = n.client.Create(ctx, desired)
 		if err != nil {
 			logger.Info("Couldn't create", "Error", err)
 			return gpuv1.NotReady, err
@@ -5153,9 +5217,9 @@ func ServiceMonitor(n ClusterPolicyController) (gpuv1.State, error) {
 	}
 
 	logger.Info("Found Resource, updating...")
-	obj.ResourceVersion = found.ResourceVersion
+	desired.SetResourceVersion(found.GetResourceVersion())
 
-	err = n.client.Update(ctx, obj)
+	err = n.client.Update(ctx, desired)
 	if err != nil {
 		logger.Info("Couldn't update", "Error", err)
 		return gpuv1.NotReady, err
diff --git a/deployments/gpu-operator/templates/role.yaml b/deployments/gpu-operator/templates/role.yaml
index dc4674c575..f233bb3e88 100644
--- a/deployments/gpu-operator/templates/role.yaml
+++ b/deployments/gpu-operator/templates/role.yaml
@@ -73,6 +73,7 @@ rules:
   - delete
 - apiGroups:
   - monitoring.coreos.com
+  - azmonitoring.coreos.com
   resources:
   - servicemonitors
   - prometheusrules
diff --git a/deployments/gpu-operator/values.yaml b/deployments/gpu-operator/values.yaml
index 90efb9fe1a..018cda5cec 100644
--- a/deployments/gpu-operator/values.yaml
+++ b/deployments/gpu-operator/values.yaml
@@ -322,6 +322,7 @@ dcgmExporter:
     internalTrafficPolicy: Cluster
   serviceMonitor:
     enabled: true
+    # apiGroup: azmonitoring.coreos.com  # required for Azure Managed Prometheus on AKS
     interval: 15s
     scrapeTimeout: 10s
     honorLabels: false