diff --git a/tensorflow/core/kernels/bincount_op_gpu.cu.cc b/tensorflow/core/kernels/bincount_op_gpu.cu.cc
index 6074b3e1f6f..7d09e9b8208 100644
--- a/tensorflow/core/kernels/bincount_op_gpu.cu.cc
+++ b/tensorflow/core/kernels/bincount_op_gpu.cu.cc
@@ -17,7 +17,7 @@ limitations under the License.
 
 #define EIGEN_USE_GPU
 
-#include "external/cub_archive/cub/device/device_histogram.cuh"
+#include "third_party/cub/device/device_histogram.cuh"
 #include "tensorflow/core/framework/op_kernel.h"
 #include "tensorflow/core/framework/register_types.h"
 #include "tensorflow/core/framework/tensor.h"
diff --git a/tensorflow/core/kernels/depthwise_conv_op_gpu.cu.cc b/tensorflow/core/kernels/depthwise_conv_op_gpu.cu.cc
index 2a254591949..76afd6f18c2 100644
--- a/tensorflow/core/kernels/depthwise_conv_op_gpu.cu.cc
+++ b/tensorflow/core/kernels/depthwise_conv_op_gpu.cu.cc
@@ -17,7 +17,7 @@ limitations under the License.
 #define EIGEN_USE_GPU
 
 #include "third_party/eigen3/unsupported/Eigen/CXX11/Tensor"
-#include "external/cub_archive/cub/util_ptx.cuh"
+#include "third_party/cub/util_ptx.cuh"
 #include "tensorflow/core/framework/op_kernel.h"
 #include "tensorflow/core/kernels/depthwise_conv_op.h"
 #include "tensorflow/core/platform/types.h"
diff --git a/tensorflow/core/kernels/dynamic_partition_op_gpu.cu.cc b/tensorflow/core/kernels/dynamic_partition_op_gpu.cu.cc
index 862a97723fd..e7882acc80e 100644
--- a/tensorflow/core/kernels/dynamic_partition_op_gpu.cu.cc
+++ b/tensorflow/core/kernels/dynamic_partition_op_gpu.cu.cc
@@ -35,10 +35,10 @@ limitations under the License.
 
 #define EIGEN_USE_GPU
 
-#include "external/cub_archive/cub/device/device_radix_sort.cuh"
-#include "external/cub_archive/cub/device/device_reduce.cuh"
-#include "external/cub_archive/cub/iterator/constant_input_iterator.cuh"
-#include "external/cub_archive/cub/thread/thread_operators.cuh"
+#include "third_party/cub/device/device_radix_sort.cuh"
+#include "third_party/cub/device/device_reduce.cuh"
+#include "third_party/cub/iterator/constant_input_iterator.cuh"
+#include "third_party/cub/thread/thread_operators.cuh"
 #include "tensorflow/core/common_runtime/gpu/gpu_event_mgr.h"
 #include "tensorflow/core/framework/op_kernel.h"
 #include "tensorflow/core/framework/register_types.h"
diff --git a/tensorflow/core/kernels/histogram_op_gpu.cu.cc b/tensorflow/core/kernels/histogram_op_gpu.cu.cc
index a88e9b0ddcd..374a05850eb 100644
--- a/tensorflow/core/kernels/histogram_op_gpu.cu.cc
+++ b/tensorflow/core/kernels/histogram_op_gpu.cu.cc
@@ -18,7 +18,7 @@ limitations under the License.
 #define EIGEN_USE_GPU
 
 #include "third_party/eigen3/unsupported/Eigen/CXX11/Tensor"
-#include "external/cub_archive/cub/device/device_histogram.cuh"
+#include "third_party/cub/device/device_histogram.cuh"
 #include "tensorflow/core/framework/op_kernel.h"
 #include "tensorflow/core/framework/register_types.h"
 #include "tensorflow/core/framework/tensor.h"
diff --git a/tensorflow/core/kernels/reduction_gpu_kernels.cu.h b/tensorflow/core/kernels/reduction_gpu_kernels.cu.h
index 88b3c2ac760..bb8254eaacf 100644
--- a/tensorflow/core/kernels/reduction_gpu_kernels.cu.h
+++ b/tensorflow/core/kernels/reduction_gpu_kernels.cu.h
@@ -21,11 +21,11 @@ limitations under the License.
 #define EIGEN_USE_GPU
 
 #include "third_party/eigen3/unsupported/Eigen/CXX11/Tensor"
-#include "external/cub_archive/cub/device/device_reduce.cuh"
-#include "external/cub_archive/cub/device/device_segmented_reduce.cuh"
-#include "external/cub_archive/cub/iterator/counting_input_iterator.cuh"
-#include "external/cub_archive/cub/iterator/transform_input_iterator.cuh"
-#include "external/cub_archive/cub/warp/warp_reduce.cuh"
+#include "third_party/cub/device/device_reduce.cuh"
+#include "third_party/cub/device/device_segmented_reduce.cuh"
+#include "third_party/cub/iterator/counting_input_iterator.cuh"
+#include "third_party/cub/iterator/transform_input_iterator.cuh"
+#include "third_party/cub/warp/warp_reduce.cuh"
 #include "cuda/include/cuComplex.h"
 #include "tensorflow/core/kernels/reduction_ops.h"
 #include "tensorflow/core/lib/core/bits.h"
diff --git a/tensorflow/core/kernels/topk_op_gpu.cu.cc b/tensorflow/core/kernels/topk_op_gpu.cu.cc
index ca296d5aa04..2fbe1fe7cbb 100644
--- a/tensorflow/core/kernels/topk_op_gpu.cu.cc
+++ b/tensorflow/core/kernels/topk_op_gpu.cu.cc
@@ -20,9 +20,9 @@ limitations under the License.
 #include <cmath>
 #include <vector>
 #include "third_party/eigen3/unsupported/Eigen/CXX11/Tensor"
-#include "external/cub_archive/cub/device/device_segmented_radix_sort.cuh"
-#include "external/cub_archive/cub/iterator/counting_input_iterator.cuh"
-#include "external/cub_archive/cub/iterator/transform_input_iterator.cuh"
+#include "third_party/cub/device/device_segmented_radix_sort.cuh"
+#include "third_party/cub/iterator/counting_input_iterator.cuh"
+#include "third_party/cub/iterator/transform_input_iterator.cuh"
 #include "tensorflow/core/framework/op_kernel.h"
 #include "tensorflow/core/framework/register_types.h"
 #include "tensorflow/core/framework/tensor.h"
diff --git a/tensorflow/core/kernels/where_op_gpu.cu.h b/tensorflow/core/kernels/where_op_gpu.cu.h
index 8879d9dd4c7..2255597651f 100644
--- a/tensorflow/core/kernels/where_op_gpu.cu.h
+++ b/tensorflow/core/kernels/where_op_gpu.cu.h
@@ -21,10 +21,10 @@ limitations under the License.
 #define EIGEN_USE_GPU
 
 #include "third_party/eigen3/unsupported/Eigen/CXX11/Tensor"
-#include "external/cub_archive/cub/device/device_reduce.cuh"
-#include "external/cub_archive/cub/device/device_select.cuh"
-#include "external/cub_archive/cub/iterator/counting_input_iterator.cuh"
-#include "external/cub_archive/cub/iterator/transform_input_iterator.cuh"
+#include "third_party/cub/device/device_reduce.cuh"
+#include "third_party/cub/device/device_select.cuh"
+#include "third_party/cub/iterator/counting_input_iterator.cuh"
+#include "third_party/cub/iterator/transform_input_iterator.cuh"
 #include "tensorflow/core/framework/register_types.h"
 #include "tensorflow/core/framework/tensor_types.h"
 #include "tensorflow/core/kernels/bounds_check.h"
diff --git a/third_party/cub.BUILD b/third_party/cub.BUILD
index 29159c9dad3..a04347b21ee 100644
--- a/third_party/cub.BUILD
+++ b/third_party/cub.BUILD
@@ -20,6 +20,7 @@ filegroup(
 cc_library(
     name = "cub",
     hdrs = if_cuda([":cub_header_files"]),
+    include_prefix = "third_party",
     deps = [
         "@local_config_cuda//cuda:cuda_headers",
     ],