diff --git a/python/quadrants/lang/ast/ast_transformers/function_def_transformer.py b/python/quadrants/lang/ast/ast_transformers/function_def_transformer.py
index 4c9bd5115b..da56f9918d 100644
--- a/python/quadrants/lang/ast/ast_transformers/function_def_transformer.py
+++ b/python/quadrants/lang/ast/ast_transformers/function_def_transformer.py
@@ -28,6 +28,7 @@
 )
 from quadrants.lang.buffer_view import BufferView
 from quadrants.lang.exception import (
+    QuadrantsCompilationError,
     QuadrantsSyntaxError,
 )
 from quadrants.lang.matrix import MatrixType
@@ -211,10 +212,6 @@ def _predeclare_struct_ndarrays(ctx: ASTTransformerFuncContext) -> None:
         ``ctx.global_context.struct_ndarray_launch_info`` so the launch path can populate the corresponding slots in the
         launch context.
         """
-        from quadrants.lang.util import cook_dtype  # pylint: disable=C0415
-
-        cache = ctx.global_context.ndarray_to_any_array
-        launch_info = ctx.global_context.struct_ndarray_launch_info
 
         def _walk_obj(obj, arg_idx, path):
             if dataclasses.is_dataclass(obj) and not isinstance(obj, type):
@@ -234,25 +231,16 @@ def _walk_obj(obj, arg_idx, path):
                         _register_ndarray(attr_val, arg_idx, (*path, attr_name))
 
         def _register_ndarray(nd, arg_idx, attr_chain):
-            key = id(nd)
-            if key in cache:
-                return
-            from quadrants._lib import core as _qd_core  # pylint: disable=C0415
-
-            element_type = cook_dtype(nd.element_type)
-            ndim = len(nd._physical_shape)
-            needs_grad = nd.grad is not None
-            layout = getattr(nd, "_qd_layout", None)
-            name = f"__qd_struct_nd_{key}"
-            arg_id_vec = impl.get_runtime().compiling_callable.insert_ndarray_param(
-                element_type, ndim, name, needs_grad
-            )
-            arr = any_array.AnyArray(
-                _qd_core.make_external_tensor_expr(element_type, ndim, arg_id_vec, needs_grad, BoundaryMode.UNSAFE),
-                _qd_layout=layout,
+            param_name = ctx.func.arg_metas[arg_idx].name
+            attr_path = ".".join(attr_chain)
+            raise QuadrantsCompilationError(
+                f"Kernel parameter '{param_name}' is annotated as qd.template(), but "
+                f"'{param_name}.{attr_path}' is a qd.ndarray. Passing ndarrays through "
+                f"template structs is not supported because it bypasses argument pruning "
+                f"and degrades launch performance. Use a concrete struct annotation "
+                f"(e.g. a @dataclass type hint) instead of qd.template() for struct "
+                f"parameters that contain ndarrays."
             )
-            cache[key] = arr
-            launch_info.append((arg_id_vec[0], arg_idx, attr_chain))
 
         assert ctx.py_args is not None
         for i, arg_meta in enumerate(ctx.func.arg_metas):
diff --git a/tests/python/test_tensor_annotation_in_func.py b/tests/python/test_tensor_annotation_in_func.py
index 059885d7dd..874a1c870c 100644
--- a/tests/python/test_tensor_annotation_in_func.py
+++ b/tests/python/test_tensor_annotation_in_func.py
@@ -13,6 +13,7 @@
 import pytest
 
 import quadrants as qd
+from quadrants.lang.exception import QuadrantsCompilationError
 
 from tests import test_utils
 
@@ -80,8 +81,12 @@ def fill(st: qd.template()):
         for i in range(N):
             st.vals[i] = i * 3
 
-    fill(s)
-    np.testing.assert_array_equal(t.to_numpy(), np.arange(N) * 3)
+    if backend == qd.Backend.NDARRAY:
+        with pytest.raises(QuadrantsCompilationError, match="qd.template.*qd.ndarray"):
+            fill(s)
+    else:
+        fill(s)
+        np.testing.assert_array_equal(t.to_numpy(), np.arange(N) * 3)
 
 
 # ---------------------------------------------------------------------------
@@ -114,8 +119,12 @@ def run(st: qd.template()):
             st.vals[i] = i
         inc_all(st)
 
-    run(s)
-    np.testing.assert_array_equal(t.to_numpy(), np.arange(N) + 10)
+    if backend == qd.Backend.NDARRAY:
+        with pytest.raises(QuadrantsCompilationError, match="qd.template.*qd.ndarray"):
+            run(s)
+    else:
+        run(s)
+        np.testing.assert_array_equal(t.to_numpy(), np.arange(N) + 10)
 
 
 # ---------------------------------------------------------------------------
@@ -148,8 +157,12 @@ def run(st: qd.template()):
         for i in range(N):
             write(st, i, i * 7)
 
-    run(s)
-    np.testing.assert_array_equal(t.to_numpy(), np.arange(N) * 7)
+    if backend == qd.Backend.NDARRAY:
+        with pytest.raises(QuadrantsCompilationError, match="qd.template.*qd.ndarray"):
+            run(s)
+    else:
+        run(s)
+        np.testing.assert_array_equal(t.to_numpy(), np.arange(N) * 7)
 
 
 # ---------------------------------------------------------------------------
@@ -181,8 +194,12 @@ def run(st: qd.template()):
         for i in range(N):
             scaled_fill(st, i)
 
-    run(s)
-    np.testing.assert_array_equal(t.to_numpy(), np.arange(N) * 5)
+    if backend == qd.Backend.NDARRAY:
+        with pytest.raises(QuadrantsCompilationError, match="qd.template.*qd.ndarray"):
+            run(s)
+    else:
+        run(s)
+        np.testing.assert_array_equal(t.to_numpy(), np.arange(N) * 5)
 
 
 # ---------------------------------------------------------------------------
@@ -254,10 +271,14 @@ def fill(st: qd.template()):
             for j in qd.static(range(3)):
                 st.vals[i][j] = i * 10.0 + j
 
-    fill(s)
-    arr = t.to_numpy()
-    assert arr.shape == (N, 3)
-    assert arr[2, 1] == pytest.approx(21.0)
+    if backend == qd.Backend.NDARRAY:
+        with pytest.raises(QuadrantsCompilationError, match="qd.template.*qd.ndarray"):
+            fill(s)
+    else:
+        fill(s)
+        arr = t.to_numpy()
+        assert arr.shape == (N, 3)
+        assert arr[2, 1] == pytest.approx(21.0)
 
 
 # ---------------------------------------------------------------------------