diff --git a/src/accelerate/parallelism_config.py b/src/accelerate/parallelism_config.py
index c4135f2f791..5ff4ebabfaf 100644
--- a/src/accelerate/parallelism_config.py
+++ b/src/accelerate/parallelism_config.py
@@ -236,8 +236,11 @@ def build_device_mesh(self, device_type: str):
         )
         if self.dp_dim_names:
             device_mesh[self.dp_dim_names]._flatten("dp")
-        if self.dp_shard_cp_dim_names:
-            device_mesh[self.dp_shard_cp_dim_names]._flatten("dp_shard_cp")
+        # Always create dp_shard_cp submesh from available dims in the mesh
+        # (dp_shard is always present after _get_mesh, even when size == 1)
+        shard_cp_dims = [d for d in ["dp_shard", "cp"] if d in mesh_dim_names]
+        if shard_cp_dims:
+            device_mesh[shard_cp_dims]._flatten("dp_shard_cp")
         if self.dp_cp_dim_names:
             device_mesh[self.dp_cp_dim_names]._flatten("dp_cp")
 
@@ -263,6 +266,11 @@ def _get_mesh(self) -> tuple[tuple[int, ...], tuple[str, ...]]:
         # Build mesh dimensions dictionary
         mesh_dims = {parallelism: self._sizes[parallelism] for parallelism in self.active_mesh_dims}
 
+        # Always include dp_shard for composable parallelism so that FSDP2 has
+        # a valid submesh even when dp_shard_size == 1 (no-op shard).
+        if mesh_dims and "dp_shard" not in mesh_dims:
+            mesh_dims["dp_shard"] = self.dp_shard_size
+
         # Apply canonical ordering
         mesh_order = ["dp_replicate", "dp_shard", "cp", "sp", "tp"]
         sorted_items = sorted(