diff --git a/mlx/backend/cuda/CMakeLists.txt b/mlx/backend/cuda/CMakeLists.txt
index 8c9a40d03..c991c2094 100644
--- a/mlx/backend/cuda/CMakeLists.txt
+++ b/mlx/backend/cuda/CMakeLists.txt
@@ -17,6 +17,8 @@ target_sources(
           ${CMAKE_CURRENT_SOURCE_DIR}/utils.cpp
           ${CMAKE_CURRENT_SOURCE_DIR}/worker.cpp)
 
+target_compile_definitions(mlx PRIVATE MLX_USE_CUDA)
+
 # Enable defining device lambda functions.
 target_compile_options(mlx
                        PRIVATE "$<$<COMPILE_LANGUAGE:CUDA>:--extended-lambda>")
diff --git a/mlx/backend/gpu/primitives.cpp b/mlx/backend/gpu/primitives.cpp
index cd9296075..938923977 100644
--- a/mlx/backend/gpu/primitives.cpp
+++ b/mlx/backend/gpu/primitives.cpp
@@ -5,9 +5,17 @@
 #include "mlx/backend/gpu/copy.h"
 #include "mlx/backend/gpu/slicing.h"
 
+#if defined(MLX_USE_CUDA)
+#include <nvtx3/nvtx3.hpp>
+#endif
+
 #include <cassert>
 
+#if defined(MLX_USE_CUDA)
+#define MLX_PROFILER_RANGE(message) nvtx3::scoped_range r(message)
+#else
 #define MLX_PROFILER_RANGE(message)
+#endif
 
 namespace mlx::core {