diff --git a/scripts/launch_triton_server.py b/scripts/launch_triton_server.py
index 8a9bc3e0..ae03cda8 100644
--- a/scripts/launch_triton_server.py
+++ b/scripts/launch_triton_server.py
@@ -159,4 +159,5 @@ def get_cmd(world_size, tritonserver, grpc_port, http_port, metrics_port,
     if args.multi_model:
         assert args.world_size == 1, 'World size must be 1 when using multi-model. Processes will be spawned automatically to run the multi-GPU models'
         env['TRTLLM_ORCHESTRATOR'] = '1'
-    subprocess.Popen(cmd, env=env)
+    child = subprocess.Popen(cmd, env=env)
+    child.communicate()