diff --git a/c/tests/CMakeLists.txt b/c/tests/CMakeLists.txt
index 6d52e5b174..4f65238841 100644
--- a/c/tests/CMakeLists.txt
+++ b/c/tests/CMakeLists.txt
@@ -1,6 +1,6 @@
 # =============================================================================
 # cmake-format: off
-# SPDX-FileCopyrightText: Copyright (c) 2021-2025, NVIDIA CORPORATION.
+# SPDX-FileCopyrightText: Copyright (c) 2021-2026, NVIDIA CORPORATION.
 # SPDX-License-Identifier: Apache-2.0
 # cmake-format: on
 # =============================================================================
@@ -49,6 +49,7 @@ function(ConfigureTest)
             GTest::gtest_main
             $<IF:$<BOOL:${CUVSC_STATIC_CUVS_LIBRARY}>,$<COMPILE_ONLY:cuvs::cuvs_static>,cuvs::cuvs>
             $<TARGET_NAME_IF_EXISTS:conda_env>
+            rmm::rmm
   )
   set_target_properties(
     ${TEST_NAME}
@@ -94,7 +95,7 @@ endif()
 
 ConfigureTest(NAME cuvs_c_headers PATH core/headers.c)
 ConfigureTest(NAME cuvs_c_test PATH core/c_api.c)
-target_link_libraries(cuvs_c_test PRIVATE CUDA::cudart)
+target_link_libraries(cuvs_c_test PRIVATE CUDA::cudart_static)
 ConfigureTest(NAME cuvs_c_neighbors_test PATH neighbors/c_api.c)
 
 # ##################################################################################################
diff --git a/ci/build_go.sh b/ci/build_go.sh
index 925dfb9153..80370048ff 100755
--- a/ci/build_go.sh
+++ b/ci/build_go.sh
@@ -1,5 +1,5 @@
 #!/bin/bash
-# SPDX-FileCopyrightText: Copyright (c) 2024-2025, NVIDIA CORPORATION.
+# SPDX-FileCopyrightText: Copyright (c) 2024-2026, NVIDIA CORPORATION.
 # SPDX-License-Identifier: Apache-2.0
 
 set -euo pipefail
@@ -31,7 +31,7 @@ set -eu
 rapids-print-env
 
 export CGO_CFLAGS="-I${CONDA_PREFIX}/include"
-export CGO_LDFLAGS="-L${CONDA_PREFIX}/lib -lcudart -lcuvs -lcuvs_c"
+export CGO_LDFLAGS="-L${CONDA_PREFIX}/lib -lcudart_static -ldl -lrt -lcuvs -lcuvs_c"
 export LD_LIBRARY_PATH="$CONDA_PREFIX/lib:$LD_LIBRARY_PATH"
 export CC=clang
 
diff --git a/conda/environments/all_cuda-129_arch-aarch64.yaml b/conda/environments/all_cuda-129_arch-aarch64.yaml
index 74f13bad68..d3a6d4dd8d 100644
--- a/conda/environments/all_cuda-129_arch-aarch64.yaml
+++ b/conda/environments/all_cuda-129_arch-aarch64.yaml
@@ -11,7 +11,6 @@ dependencies:
 - clang-tools==20.1.4
 - clang==20.1.4
 - cmake>=3.30.4
-- cuda-cudart-dev
 - cuda-nvcc
 - cuda-nvtx-dev
 - cuda-profiler-api
diff --git a/conda/environments/all_cuda-129_arch-x86_64.yaml b/conda/environments/all_cuda-129_arch-x86_64.yaml
index f726fbd93f..7da3405cec 100644
--- a/conda/environments/all_cuda-129_arch-x86_64.yaml
+++ b/conda/environments/all_cuda-129_arch-x86_64.yaml
@@ -11,7 +11,6 @@ dependencies:
 - clang-tools==20.1.4
 - clang==20.1.4
 - cmake>=3.30.4
-- cuda-cudart-dev
 - cuda-nvcc
 - cuda-nvtx-dev
 - cuda-profiler-api
diff --git a/conda/environments/all_cuda-131_arch-aarch64.yaml b/conda/environments/all_cuda-131_arch-aarch64.yaml
index d35632064f..94129ae55a 100644
--- a/conda/environments/all_cuda-131_arch-aarch64.yaml
+++ b/conda/environments/all_cuda-131_arch-aarch64.yaml
@@ -11,7 +11,6 @@ dependencies:
 - clang-tools==20.1.4
 - clang==20.1.4
 - cmake>=3.30.4
-- cuda-cudart-dev
 - cuda-nvcc
 - cuda-nvtx-dev
 - cuda-profiler-api
diff --git a/conda/environments/all_cuda-131_arch-x86_64.yaml b/conda/environments/all_cuda-131_arch-x86_64.yaml
index 367cf0ec6e..23150f3ba4 100644
--- a/conda/environments/all_cuda-131_arch-x86_64.yaml
+++ b/conda/environments/all_cuda-131_arch-x86_64.yaml
@@ -11,7 +11,6 @@ dependencies:
 - clang-tools==20.1.4
 - clang==20.1.4
 - cmake>=3.30.4
-- cuda-cudart-dev
 - cuda-nvcc
 - cuda-nvtx-dev
 - cuda-profiler-api
diff --git a/conda/environments/bench_ann_cuda-129_arch-aarch64.yaml b/conda/environments/bench_ann_cuda-129_arch-aarch64.yaml
index 26df7c96c9..5dd1abf183 100644
--- a/conda/environments/bench_ann_cuda-129_arch-aarch64.yaml
+++ b/conda/environments/bench_ann_cuda-129_arch-aarch64.yaml
@@ -10,7 +10,6 @@ dependencies:
 - clang==20.1.4
 - click
 - cmake>=3.30.4
-- cuda-cudart-dev
 - cuda-nvcc
 - cuda-nvtx-dev
 - cuda-profiler-api
diff --git a/conda/environments/bench_ann_cuda-129_arch-x86_64.yaml b/conda/environments/bench_ann_cuda-129_arch-x86_64.yaml
index d63791f084..75d36d1f33 100644
--- a/conda/environments/bench_ann_cuda-129_arch-x86_64.yaml
+++ b/conda/environments/bench_ann_cuda-129_arch-x86_64.yaml
@@ -10,7 +10,6 @@ dependencies:
 - clang==20.1.4
 - click
 - cmake>=3.30.4
-- cuda-cudart-dev
 - cuda-nvcc
 - cuda-nvtx-dev
 - cuda-profiler-api
diff --git a/conda/environments/bench_ann_cuda-131_arch-aarch64.yaml b/conda/environments/bench_ann_cuda-131_arch-aarch64.yaml
index d1e3a5bae4..3758fab63b 100644
--- a/conda/environments/bench_ann_cuda-131_arch-aarch64.yaml
+++ b/conda/environments/bench_ann_cuda-131_arch-aarch64.yaml
@@ -10,7 +10,6 @@ dependencies:
 - clang==20.1.4
 - click
 - cmake>=3.30.4
-- cuda-cudart-dev
 - cuda-nvcc
 - cuda-nvtx-dev
 - cuda-profiler-api
diff --git a/conda/environments/bench_ann_cuda-131_arch-x86_64.yaml b/conda/environments/bench_ann_cuda-131_arch-x86_64.yaml
index 4b09d1b885..5f93dbb946 100644
--- a/conda/environments/bench_ann_cuda-131_arch-x86_64.yaml
+++ b/conda/environments/bench_ann_cuda-131_arch-x86_64.yaml
@@ -10,7 +10,6 @@ dependencies:
 - clang==20.1.4
 - click
 - cmake>=3.30.4
-- cuda-cudart-dev
 - cuda-nvcc
 - cuda-nvtx-dev
 - cuda-profiler-api
diff --git a/conda/recipes/cuvs/recipe.yaml b/conda/recipes/cuvs/recipe.yaml
index dcce17cb13..690a2cf1f9 100644
--- a/conda/recipes/cuvs/recipe.yaml
+++ b/conda/recipes/cuvs/recipe.yaml
@@ -91,7 +91,6 @@ requirements:
     - if: cuda_major == "12"
       then: cuda-python >=12.9.2,<13.0
       else: cuda-python >=13.0.1,<14.0
-    - cuda-cudart
   ignore_run_exports:
     by_name:
       - cuda-version
diff --git a/conda/recipes/libcuvs/recipe.yaml b/conda/recipes/libcuvs/recipe.yaml
index abd3031a94..4453b18393 100644
--- a/conda/recipes/libcuvs/recipe.yaml
+++ b/conda/recipes/libcuvs/recipe.yaml
@@ -129,7 +129,6 @@ outputs:
         - libraft-headers =${{ minor_version }}
         - librmm =${{ minor_version }}
         - nccl
-        - cuda-cudart
         - libcublas
         - libcurand
         - libcusolver
@@ -198,7 +197,6 @@ outputs:
         - libraft-headers =${{ minor_version }}
         - librmm =${{ minor_version }}
         - nccl
-        - cuda-cudart
         - libcublas
         - libcurand
         - libcusolver
@@ -265,7 +263,6 @@ outputs:
         - libraft-headers =${{ minor_version }}
         - librmm =${{ minor_version }}
         - nccl
-        - cuda-cudart
         - libcublas
         - libcurand
         - libcusolver
@@ -328,7 +325,6 @@ outputs:
         - ${{ pin_compatible("cuda-version", upper_bound="x", lower_bound="x") }}
         - libraft-headers =${{ minor_version }}
         - nccl
-        - cuda-cudart
         - libcublas
         - libcurand
         - libcusolver
@@ -431,7 +427,6 @@ outputs:
         - ${{ pin_subpackage("libcuvs", exact=True) }}
         - ${{ pin_compatible("cuda-version", upper_bound="x", lower_bound="x") }}
         - nccl
-        - cuda-cudart
         - libcublas
         - libcurand
         - libcusolver
@@ -506,7 +501,6 @@ outputs:
         - ${{ pin_subpackage("libcuvs-headers", exact=True) }}
         - ${{ pin_subpackage("libcuvs", exact=True) }}
         - ${{ pin_compatible("cuda-version", upper_bound="x", lower_bound="x") }}
-        - cuda-cudart
         - libcublas
         - libcurand
         - libcusolver
diff --git a/cpp/CMakeLists.txt b/cpp/CMakeLists.txt
index d4e9f2c6b4..d91bf0e30a 100644
--- a/cpp/CMakeLists.txt
+++ b/cpp/CMakeLists.txt
@@ -57,7 +57,6 @@ option(CUDA_ENABLE_KERNELINFO "Enable kernel resource usage info" OFF)
 option(CUDA_ENABLE_LINEINFO
        "Enable the -lineinfo option for nvcc (useful for cuda-memcheck / profiler)" OFF
 )
-option(CUDA_STATIC_RUNTIME "Statically link the CUDA runtime" OFF)
 option(CUDA_STATIC_MATH_LIBRARIES "Statically link the CUDA math libraries" OFF)
 option(CUVS_STATIC_RAPIDS_LIBRARIES "Build and statically link RAPIDS libraries" OFF)
 option(CUDA_LOG_COMPILE_TIME "Write a log of compilation times to nvcc_compile_log.csv" OFF)
@@ -116,7 +115,6 @@ message(VERBOSE "cuVS: Disable OpenMP: ${DISABLE_OPENMP}")
 message(VERBOSE "cuVS: Enable kernel resource usage info: ${CUDA_ENABLE_KERNELINFO}")
 message(VERBOSE "cuVS: Enable lineinfo in nvcc: ${CUDA_ENABLE_LINEINFO}")
 message(VERBOSE "cuVS: Enable nvtx markers: ${CUVS_NVTX}")
-message(VERBOSE "cuVS: Statically link the CUDA runtime: ${CUDA_STATIC_RUNTIME}")
 message(VERBOSE "cuVS: Statically link the CUDA math libraries: ${CUDA_STATIC_MATH_LIBRARIES}")
 message(VERBOSE "cuVS: Build and statically link RAPIDS libraries: ${CUVS_STATIC_RAPIDS_LIBRARIES}")
 
@@ -153,7 +151,7 @@ endif()
 
 if(NOT BUILD_CPU_ONLY)
   # CUDA runtime
-  rapids_cuda_init_runtime(USE_STATIC ${CUDA_STATIC_RUNTIME})
+  rapids_cuda_init_runtime(USE_STATIC ON)
   # * find CUDAToolkit package
   # * determine GPU architectures
   # * enable the CMake CUDA language
diff --git a/cpp/cmake/thirdparty/get_faiss.cmake b/cpp/cmake/thirdparty/get_faiss.cmake
index 75a4473276..dfdc1a4d31 100644
--- a/cpp/cmake/thirdparty/get_faiss.cmake
+++ b/cpp/cmake/thirdparty/get_faiss.cmake
@@ -48,7 +48,7 @@ function(find_and_configure_faiss)
     "FAISS_ENABLE_CUVS ${PKG_ENABLE_GPU}"
     "FAISS_ENABLE_PYTHON OFF"
     "FAISS_OPT_LEVEL ${CUVS_FAISS_OPT_LEVEL}"
-    "FAISS_USE_CUDA_TOOLKIT_STATIC ${CUDA_STATIC_RUNTIME}"
+    "FAISS_USE_CUDA_TOOLKIT_STATIC ON"
     "BUILD_TESTING OFF"
     "CMAKE_MESSAGE_LOG_LEVEL VERBOSE"
     )
diff --git a/dependencies.yaml b/dependencies.yaml
index e855a93f3e..df6159084b 100644
--- a/dependencies.yaml
+++ b/dependencies.yaml
@@ -96,6 +96,7 @@ files:
       - cuda_version
       - rapids_build
       - rust
+      - depends_on_cudart
       - depends_on_libcuvs
       - depends_on_libraft
       - depends_on_nccl
@@ -110,6 +111,7 @@ files:
       - cuda_version
       - rapids_build
       - go
+      - depends_on_cudart
       - depends_on_libcuvs
       - depends_on_libraft
       - depends_on_nccl
@@ -319,7 +321,6 @@ dependencies:
       - output_types: [conda]
         packages:
           - cuda-nvtx-dev
-          - cuda-cudart-dev
           - cuda-profiler-api
           - libcublas-dev
           - libcurand-dev
@@ -357,6 +358,11 @@ dependencies:
           - matrix:
             packages:
               - cuda-toolkit[cublas,curand,cusolver,cusparse,nvjitlink]>=12,<14
+  depends_on_cudart:
+    common:
+      - output_types: conda
+        packages:
+          - cuda-cudart-dev
   depends_on_cupy:
     common:
       - output_types: conda
diff --git a/docs/source/build.rst b/docs/source/build.rst
index dcef10c96b..5e863e40f4 100644
--- a/docs/source/build.rst
+++ b/docs/source/build.rst
@@ -205,7 +205,7 @@ After building the C and C++ libraries, the Golang library can be built with the
 
     export CUDA_HOME="/usr/local/cuda" # or wherever your CUDA installation is.
     export CGO_CFLAGS="-I${CONDA_PREFIX}/include -I${CUDA_HOME}/include"
-    export CGO_LDFLAGS="-L${CONDA_PREFIX}/lib -lcudart -lcuvs -lcuvs_c"
+    export CGO_LDFLAGS="-L${CONDA_PREFIX}/lib -lcuvs -lcuvs_c"
     export LD_LIBRARY_PATH="$CONDA_PREFIX/lib:$LD_LIBRARY_PATH"
     export CC=clang
 
@@ -259,11 +259,6 @@ cuVS has the following configurable cmake flags available:
    - OFF
    - Enable the `-lineinfo` option for nvcc
 
- * - CUDA_STATIC_RUNTIME
-   - ON, OFF
-   - OFF
-   - Statically link the CUDA runtime
-
  * - CUDA_STATIC_MATH_LIBRARIES
    - ON, OFF
    - OFF
diff --git a/examples/c/CMakeLists.txt b/examples/c/CMakeLists.txt
index feb7a03309..61092834c7 100644
--- a/examples/c/CMakeLists.txt
+++ b/examples/c/CMakeLists.txt
@@ -34,29 +34,29 @@ set(CMAKE_C_FLAGS "${CMAKE_C_FLAGS} -Wall -Wextra -Werror")
 add_executable(CAGRA_C_EXAMPLE src/cagra_c_example.c)
 target_include_directories(CAGRA_C_EXAMPLE PUBLIC "$<BUILD_INTERFACE:${DLPACK_INCLUDE_DIR}>")
 target_link_libraries(
-  CAGRA_C_EXAMPLE PRIVATE cuvs::c_api CUDA::cudart $<TARGET_NAME_IF_EXISTS:conda_env>
+  CAGRA_C_EXAMPLE PRIVATE cuvs::c_api CUDA::cudart_static $<TARGET_NAME_IF_EXISTS:conda_env>
 )
 
 add_executable(L2_C_EXAMPLE src/L2_c_example.c)
 target_include_directories(L2_C_EXAMPLE PUBLIC "$<BUILD_INTERFACE:${DLPACK_INCLUDE_DIR}>")
 target_link_libraries(
-  L2_C_EXAMPLE PRIVATE cuvs::c_api CUDA::cudart $<TARGET_NAME_IF_EXISTS:conda_env>
+  L2_C_EXAMPLE PRIVATE cuvs::c_api CUDA::cudart_static $<TARGET_NAME_IF_EXISTS:conda_env>
 )
 
 add_executable(IVF_FLAT_C_EXAMPLE src/ivf_flat_c_example.c)
 target_include_directories(IVF_FLAT_C_EXAMPLE PUBLIC "$<BUILD_INTERFACE:${DLPACK_INCLUDE_DIR}>")
 target_link_libraries(
-  IVF_FLAT_C_EXAMPLE PRIVATE cuvs::c_api CUDA::cudart $<TARGET_NAME_IF_EXISTS:conda_env>
+  IVF_FLAT_C_EXAMPLE PRIVATE cuvs::c_api CUDA::cudart_static $<TARGET_NAME_IF_EXISTS:conda_env>
 )
 
 add_executable(IVF_PQ_C_EXAMPLE src/ivf_pq_c_example.c)
 target_include_directories(IVF_PQ_C_EXAMPLE PUBLIC "$<BUILD_INTERFACE:${DLPACK_INCLUDE_DIR}>")
 target_link_libraries(
-  IVF_PQ_C_EXAMPLE PRIVATE cuvs::c_api CUDA::cudart $<TARGET_NAME_IF_EXISTS:conda_env>
+  IVF_PQ_C_EXAMPLE PRIVATE cuvs::c_api CUDA::cudart_static $<TARGET_NAME_IF_EXISTS:conda_env>
 )
 
 add_executable(BRUTEFORCE_C_EXAMPLE src/bruteforce_c_example.c)
 target_include_directories(BRUTEFORCE_C_EXAMPLE PUBLIC "$<BUILD_INTERFACE:${DLPACK_INCLUDE_DIR}>")
 target_link_libraries(
-  BRUTEFORCE_C_EXAMPLE PRIVATE cuvs::c_api CUDA::cudart $<TARGET_NAME_IF_EXISTS:conda_env>
+  BRUTEFORCE_C_EXAMPLE PRIVATE cuvs::c_api CUDA::cudart_static $<TARGET_NAME_IF_EXISTS:conda_env>
 )
diff --git a/examples/go/README.md b/examples/go/README.md
index 798a024de4..819b05c700 100644
--- a/examples/go/README.md
+++ b/examples/go/README.md
@@ -17,7 +17,7 @@ You may prefer to use `mamba`, as it provides significant speedup over `conda`.
 1. Set up the required environment variables:
 ```bash
 export CGO_CFLAGS="-I${CONDA_PREFIX}/include"
-export CGO_LDFLAGS="-L${CONDA_PREFIX}/lib -lcudart -lcuvs -lcuvs_c"
+export CGO_LDFLAGS="-L${CONDA_PREFIX}/lib -lcudart_static -ldl -lrt -lcuvs -lcuvs_c"
 export LD_LIBRARY_PATH="$CONDA_PREFIX/lib:$LD_LIBRARY_PATH"
 export CC=clang
 ```
diff --git a/go/dlpack.go b/go/dlpack.go
index 6fe619fd35..fcb9632586 100644
--- a/go/dlpack.go
+++ b/go/dlpack.go
@@ -3,6 +3,7 @@ package cuvs
 // #include <stdlib.h>
 // #include <dlpack/dlpack.h>
 // #include <cuvs/core/c_api.h>
+// #include <cuda_runtime_api.h>
 import "C"
 
 import (
diff --git a/python/libcuvs/CMakeLists.txt b/python/libcuvs/CMakeLists.txt
index 7f5d100587..bac8373cf7 100644
--- a/python/libcuvs/CMakeLists.txt
+++ b/python/libcuvs/CMakeLists.txt
@@ -1,6 +1,6 @@
 # =============================================================================
 # cmake-format: off
-# SPDX-FileCopyrightText: Copyright (c) 2025, NVIDIA CORPORATION.
+# SPDX-FileCopyrightText: Copyright (c) 2025-2026, NVIDIA CORPORATION.
 # SPDX-License-Identifier: Apache-2.0
 # cmake-format: on
 # =============================================================================
@@ -31,7 +31,6 @@ endif()
 unset(cuvs_FOUND)
 
 # --- CUDA --- #
-set(CUDA_STATIC_RUNTIME ON)
 set(CUDA_STATIC_MATH_LIBRARIES OFF)
 
 # --- RAFT ---#
diff --git a/rust/cuvs-sys/build.rs b/rust/cuvs-sys/build.rs
index 99e811eeb9..cec80eb736 100644
--- a/rust/cuvs-sys/build.rs
+++ b/rust/cuvs-sys/build.rs
@@ -15,8 +15,11 @@ fn main() {
         "cargo:rustc-link-search=native={}/lib",
         cuvs_build.display()
     );
+    if let Ok(conda_prefix) = env::var("CONDA_PREFIX") {
+        println!("cargo:rustc-link-search=native={}/lib", conda_prefix);
+    }
     println!("cargo:rustc-link-lib=dylib=cuvs_c");
-    println!("cargo:rustc-link-lib=dylib=cudart");
+    println!("cargo:rustc-link-lib=static=cudart_static");
 
     // we need some extra flags both to link against cuvs, and also to run bindgen
     // specifically we need to: