diff --git a/CMakeLists.txt b/CMakeLists.txt
index d3d99f6e7c8..2c25a70ccfc 100644
--- a/CMakeLists.txt
+++ b/CMakeLists.txt
@@ -1,4 +1,4 @@
-cmake_minimum_required(VERSION 3.18)
+cmake_minimum_required(VERSION 3.24)
 
 #
 # Setting a cmake_policy to OLD is deprecated by definition and will raise a
@@ -24,16 +24,6 @@ if(POLICY CMP0127)
     cmake_policy(SET CMP0127 NEW)
 endif()
 
-#
-# CMake 3.18+: CMAKE_CUDA_ARCHITECTURES
-# https://cmake.org/cmake/help/latest/policy/CMP0104.html
-# We have to migrate there, but maybe the new "native" option (CMake 3.24+)
-# is what we want to wait for:
-# https://cmake.org/cmake/help/v3.24/prop_tgt/CUDA_ARCHITECTURES.html
-if(POLICY CMP0104)
-    cmake_policy(SET CMP0104 OLD)
-endif()
-
 #
 # Prevent in-source builds
 #
diff --git a/Tools/CMake/AMReXCUDAOptions.cmake b/Tools/CMake/AMReXCUDAOptions.cmake
index cce0306f4b9..79b06caee90 100644
--- a/Tools/CMake/AMReXCUDAOptions.cmake
+++ b/Tools/CMake/AMReXCUDAOptions.cmake
@@ -17,16 +17,7 @@ endmacro ()
 #
 message(STATUS "Enabled CUDA options:")
 
-set(AMReX_CUDA_ARCH_DEFAULT "Auto")
-if(DEFINED CMAKE_CUDA_ARCHITECTURES)
-   set(AMReX_CUDA_ARCH_DEFAULT "${CMAKE_CUDA_ARCHITECTURES}")
-endif ()
-if(DEFINED ENV{AMREX_CUDA_ARCH})
-   set(AMReX_CUDA_ARCH_DEFAULT "$ENV{AMREX_CUDA_ARCH}")
-endif()
-set(AMReX_CUDA_ARCH ${AMReX_CUDA_ARCH_DEFAULT} CACHE STRING "CUDA architecture (Use 'Auto' for automatic detection)")
-
-option(AMReX_CUDA_FASTMATH "Enable CUDA fastmath" ON)  # Note: inconsistent with AMReX_FASTMATH defaults
+option(AMReX_CUDA_FASTMATH "Enable CUDA fastmath" ON)
 cuda_print_option( AMReX_CUDA_FASTMATH )
 
 set(AMReX_CUDA_MAXREGCOUNT "255" CACHE STRING
diff --git a/Tools/CMake/AMReXParallelBackends.cmake b/Tools/CMake/AMReXParallelBackends.cmake
index 4119b1b2678..83d9d490814 100644
--- a/Tools/CMake/AMReXParallelBackends.cmake
+++ b/Tools/CMake/AMReXParallelBackends.cmake
@@ -109,8 +109,14 @@ if (  AMReX_GPU_BACKEND STREQUAL "CUDA"
           )
   endforeach()
 
-   # Take care of cuda archs
-   set_cuda_architectures(AMReX_CUDA_ARCH)
+   # default to building for the architecture of the available GPU 
+   set(AMREX_CUDA_ARCHS native)
+
+   # but, if the user specifies architectures explicitly then use that instead
+   if (DEFINED CMAKE_CUDA_ARCHITECTURES) 
+      set(AMREX_CUDA_ARCHS ${CMAKE_CUDA_ARCHITECTURES})
+   endif()
+
    foreach(D IN LISTS AMReX_SPACEDIM)
        set_target_properties(amrex_${D}d
           PROPERTIES