diff --git a/cpp/cmake/patches/cutlass/Support-both-CUDA-12-and-13-cccl-header-locations.patch b/cpp/cmake/patches/cutlass/Support-both-CUDA-12-and-13-cccl-header-locations.patch
new file mode 100644
index 0000000000..489ac69f93
--- /dev/null
+++ b/cpp/cmake/patches/cutlass/Support-both-CUDA-12-and-13-cccl-header-locations.patch
@@ -0,0 +1,31 @@
+From 661c7e679ac72926d619da46834d09f52a727f5e Mon Sep 17 00:00:00 2001
+From: Robert Maynard <rmaynard@nvidia.com>
+Date: Tue, 5 Aug 2025 15:05:57 -0400
+Subject: [PATCH] Support both CUDA 12 and 13 cccl header locations
+
+---
+ CMakeLists.txt | 8 +++++++-
+ 1 file changed, 7 insertions(+), 1 deletion(-)
+
+diff --git a/CMakeLists.txt b/CMakeLists.txt
+index 38dcca9f..4088b71f 100755
+--- a/CMakeLists.txt
++++ b/CMakeLists.txt
+@@ -704,8 +704,14 @@ target_include_directories(
+   CUTLASS
+   SYSTEM INTERFACE
+   $<BUILD_INTERFACE:${CUDA_TOOLKIT_ROOT_DIR}/include>
+-  $<BUILD_INTERFACE:${CUDA_TOOLKIT_ROOT_DIR}/include/cccl>
+   )
++if(CUDA_VERSION VERSION_GREATER_EQUAL 13.0)
++  target_include_directories(
++    CUTLASS
++    SYSTEM INTERFACE
++    $<BUILD_INTERFACE:${CUDA_TOOLKIT_ROOT_DIR}/include/cccl>
++    )
++endif()
+
+ install(
+   DIRECTORY
+--
+2.39.5 (Apple Git-154)
diff --git a/cpp/cmake/patches/cutlass/build-export.patch b/cpp/cmake/patches/cutlass/build-export.patch
deleted file mode 100644
index 31bbd25102..0000000000
--- a/cpp/cmake/patches/cutlass/build-export.patch
+++ /dev/null
@@ -1,26 +0,0 @@
-From e0a9597946257a01ae8444200f836ee51d5597ba Mon Sep 17 00:00:00 2001
-From: Kyle Edwards <kyedwards@nvidia.com>
-Date: Wed, 20 Nov 2024 16:37:38 -0500
-Subject: [PATCH] Remove erroneous include directories
-
-These directories are left over from when CuTe was a separate
-CMake project. Remove them.
----
- CMakeLists.txt | 2 --
- 1 file changed, 2 deletions(-)
-
-diff --git a/CMakeLists.txt b/CMakeLists.txt
-index 7419bdf5e..545384d82 100755
---- a/CMakeLists.txt
-+++ b/CMakeLists.txt
-@@ -665,8 +665,6 @@ target_include_directories(
-   $<INSTALL_INTERFACE:include>
-   $<BUILD_INTERFACE:${CUTLASS_INCLUDE_DIR}>
-   $<BUILD_INTERFACE:${CMAKE_CURRENT_BINARY_DIR}/include>
--  $<BUILD_INTERFACE:${cute_SOURCE_DIR}/include>
--  $<BUILD_INTERFACE:${cute_SOURCE_DIR}/examples>
-   )
-
- # Mark CTK headers as system to supress warnings from them
---
-2.34.1
diff --git a/cpp/cmake/patches/cutlass_override.json b/cpp/cmake/patches/cutlass_override.json
index 7bf818987f..bf9791ebd1 100644
--- a/cpp/cmake/patches/cutlass_override.json
+++ b/cpp/cmake/patches/cutlass_override.json
@@ -1,13 +1,13 @@
 {
   "packages" : {
     "cutlass" : {
-      "version": "3.5.1",
+      "version": "4.1.0",
       "git_url": "https://github.com/NVIDIA/cutlass.git",
       "git_tag": "v${version}",
       "patches" : [
         {
-          "file" : "${current_json_dir}/cutlass/build-export.patch",
-          "issue" : "Fix build directory export",
+          "file" : "${current_json_dir}/cutlass/Support-both-CUDA-12-and-13-cccl-header-locations.patch",
+          "issue" : "Support CUDA 12 CTK layout[https://github.com/NVIDIA/cutlass/pull/2543]",
           "fixed_in" : ""
         }
       ]
diff --git a/cpp/cmake/thirdparty/get_cutlass.cmake b/cpp/cmake/thirdparty/get_cutlass.cmake
index 2a12d63f22..8ad6da994b 100644
--- a/cpp/cmake/thirdparty/get_cutlass.cmake
+++ b/cpp/cmake/thirdparty/get_cutlass.cmake
@@ -39,19 +39,14 @@ function(find_and_configure_cutlass)
   include("${rapids-cmake-dir}/cpm/package_override.cmake")
   rapids_cpm_package_override("${CMAKE_CURRENT_FUNCTION_LIST_DIR}/../patches/cutlass_override.json")
 
-  include("${rapids-cmake-dir}/cpm/detail/package_details.cmake")
-  rapids_cpm_package_details(cutlass version repository tag shallow exclude)
-
-  include("${rapids-cmake-dir}/cpm/detail/generate_patch_command.cmake")
-  rapids_cpm_generate_patch_command(cutlass ${version} patch_command build_patch_only)
+  include("${rapids-cmake-dir}/cpm/detail/package_info.cmake")
+  rapids_cpm_package_info(cutlass VERSION_VAR version FIND_VAR find_args CPM_VAR cpm_find_info
+                          TO_INSTALL_VAR to_install)
 
   rapids_cpm_find(
-    NvidiaCutlass ${version} ${build_patch_only}
+    NvidiaCutlass ${version} ${find_args}
     GLOBAL_TARGETS nvidia::cutlass::cutlass
-    CPM_ARGS
-    GIT_REPOSITORY ${repository}
-    GIT_TAG ${tag}
-    GIT_SHALLOW ${shallow} ${patch_command}
+    CPM_ARGS ${cpm_find_info}
     OPTIONS "CUDAToolkit_ROOT ${CUDAToolkit_LIBRARY_DIR}"
   )
 
@@ -67,7 +62,6 @@ function(find_and_configure_cutlass)
       NAMESPACE nvidia::cutlass::
     )
   endif()
-  # endif()
 
   # We generate the cutlass-config files when we built cutlass locally, so always do
   # `find_dependency`