diff --git a/cluster-autoscaler/cluster_autoscaler.go b/cluster-autoscaler/cluster_autoscaler.go
index 1a6b84637b..aa0265495f 100644
--- a/cluster-autoscaler/cluster_autoscaler.go
+++ b/cluster-autoscaler/cluster_autoscaler.go
@@ -71,7 +71,8 @@ var (
 		"Node utilization level, defined as sum of requested resources divided by capacity, below which a node can be considered for scale down")
 	scaleDownTrialInterval = flag.Duration("scale-down-trial-interval", 1*time.Minute,
 		"How often scale down possiblity is check")
-	scanInterval = flag.Duration("scan-interval", 10*time.Second, "How often cluster is reevaluated for scale up or down")
+	scanInterval  = flag.Duration("scan-interval", 10*time.Second, "How often cluster is reevaluated for scale up or down")
+	maxNodesTotal = flag.Int("max-nodes-total", 0, "Maximum number of nodes in all node groups. Cluster autoscaler will not grow the cluster beyond this number.")
 
 	cloudProviderFlag = flag.String("cloud-provider", "gce", "Cloud provider type. Allowed values: gce")
 )
@@ -211,10 +212,13 @@ func run(_ <-chan struct{}) {
 
 				if len(unschedulablePodsToHelp) == 0 {
 					glog.V(1).Info("No unschedulable pods")
+				} else if *maxNodesTotal > 0 && len(nodes) >= *maxNodesTotal {
+					glog.V(1).Info("Max total nodes in cluster reached")
 				} else {
 					scaleUpStart := time.Now()
 					updateLastTime("scaleup")
-					scaledUp, err := ScaleUp(unschedulablePodsToHelp, nodes, cloudProvider, kubeClient, predicateChecker, recorder)
+					scaledUp, err := ScaleUp(unschedulablePodsToHelp, nodes, cloudProvider, kubeClient, predicateChecker, recorder,
+						*maxNodesTotal)
 
 					updateDuration("scaleup", scaleUpStart)
 
diff --git a/cluster-autoscaler/scale_up.go b/cluster-autoscaler/scale_up.go
index ba0c663ad3..48d309994e 100644
--- a/cluster-autoscaler/scale_up.go
+++ b/cluster-autoscaler/scale_up.go
@@ -36,9 +36,10 @@ type ExpansionOption struct {
 }
 
 // ScaleUp tries to scale the cluster up. Return true if it found a way to increase the size,
-// false if it didn't and error if an error occured.
+// false if it didn't and error if an error occured. Assumes that all nodes in the cluster are
+// ready and in sync with instance groups.
 func ScaleUp(unschedulablePods []*kube_api.Pod, nodes []*kube_api.Node, cloudProvider cloudprovider.CloudProvider, kubeClient *kube_client.Client,
-	predicateChecker *simulator.PredicateChecker, recorder kube_record.EventRecorder) (bool, error) {
+	predicateChecker *simulator.PredicateChecker, recorder kube_record.EventRecorder, maxNodesTotal int) (bool, error) {
 
 	// From now on we only care about unschedulable pods that were marked after the newest
 	// node became available for the scheduler.
@@ -122,6 +123,15 @@ func ScaleUp(unschedulablePods []*kube_api.Pod, nodes []*kube_api.Node, cloudPro
 			glog.V(1).Infof("Capping size to MAX (%d)", bestOption.nodeGroup.MaxSize())
 			newSize = bestOption.nodeGroup.MaxSize()
 		}
+
+		if maxNodesTotal > 0 && len(nodes)+(newSize-currentSize) > maxNodesTotal {
+			glog.V(1).Infof("Capping size to max cluster total size (%d)", maxNodesTotal)
+			newSize = maxNodesTotal - len(nodes) + currentSize
+			if newSize < currentSize {
+				return false, fmt.Errorf("max node total count already reached")
+			}
+		}
+
 		glog.V(1).Infof("Setting %s size to %d", bestOption.nodeGroup.Id(), newSize)
 
 		if err := bestOption.nodeGroup.IncreaseSize(newSize - currentSize); err != nil {