diff --git a/cpp/include/cugraph/legacy/functions.hpp b/cpp/include/cugraph/legacy/functions.hpp
index e5e5915e112..2ccb676c23f 100644
--- a/cpp/include/cugraph/legacy/functions.hpp
+++ b/cpp/include/cugraph/legacy/functions.hpp
@@ -1,5 +1,5 @@
 /*
- * SPDX-FileCopyrightText: Copyright (c) 2020-2026, NVIDIA CORPORATION.
+ * SPDX-FileCopyrightText: Copyright (c) 2020-2026, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
  * SPDX-License-Identifier: Apache-2.0
  */
 #pragma once
@@ -35,6 +35,9 @@ namespace CUGRAPH_EXPORT cugraph {
  * @param[in]  graph          cuGraph graph in coordinate format
  * @param[in]  mr             Memory resource used to allocate the returned graph
  *
+ * @note Synchronizes the internal CUDA stream before returning so callers may use the
+ *       returned CSR on a different stream without additional ordering.
+ *
  * @return                    Unique pointer to generate Compressed Sparse Row graph
  *
  */
diff --git a/cpp/src/converters/legacy/COOtoCSR.cuh b/cpp/src/converters/legacy/COOtoCSR.cuh
index ad194eca019..0f8f0bd6532 100644
--- a/cpp/src/converters/legacy/COOtoCSR.cuh
+++ b/cpp/src/converters/legacy/COOtoCSR.cuh
@@ -1,5 +1,5 @@
 /*
- * SPDX-FileCopyrightText: Copyright (c) 2019-2026, NVIDIA CORPORATION.
+ * SPDX-FileCopyrightText: Copyright (c) 2019-2026, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
  * SPDX-License-Identifier: Apache-2.0
  */
 /*
@@ -154,6 +154,11 @@ std::unique_ptr<legacy::GraphCSR<VT, ET, WT>> coo_to_csr(
     std::move(coo_contents.dst_indices),
     std::move(coo_contents.edge_data)};
 
+  // All conversion work runs on stream_view, which callers do not share (for example
+  // raft::handle_t defaults to cudaStreamPerThread). Synchronize before returning so
+  // downstream algorithms can safely consume the CSR buffers on any stream.
+  stream_view.synchronize();
+
   return std::make_unique<legacy::GraphCSR<VT, ET, WT>>(std::move(csr_contents));
 }