diff --git a/python-package/xgboost/data.py b/python-package/xgboost/data.py
index 7a07be4d2d20..687c1117877c 100644
--- a/python-package/xgboost/data.py
+++ b/python-package/xgboost/data.py
@@ -475,29 +475,26 @@ def _is_arrow(data) -> bool:
         return False
 
 
-class RecordBatchDataIter:
-    """Data iterator used to ingest Arrow columnar record batches. We are not
-    using class DataIter because it is only intended for building Device DMatrix.
+def record_batch_data_iter(data_iter: Iterator) -> Callable:
+    """Data iterator used to ingest Arrow columnar record batches. We are not using
+    class DataIter because it is only intended for building Device DMatrix and external
+    memory DMatrix.
 
     """
+    from pyarrow.cffi import ffi
 
-    def __init__(self, data_iter: Iterator) -> None:
-        from pyarrow.cffi import ffi
-
-        self.data_iter = data_iter  # an iterator for Arrow record batches
-        self.c_schemas: List[ffi.CData] = []
-        self.c_arrays: List[ffi.CData] = []
+    c_schemas: List[ffi.CData] = []
+    c_arrays: List[ffi.CData] = []
 
-    def next(self, data_handle: int) -> int:
-        "Fetch the next batch."
+    def _next(data_handle: int) -> int:
         from pyarrow.cffi import ffi
 
         try:
-            batch = next(self.data_iter)
-            self.c_schemas.append(ffi.new("struct ArrowSchema*"))
-            self.c_arrays.append(ffi.new("struct ArrowArray*"))
-            ptr_schema = int(ffi.cast("uintptr_t", self.c_schemas[-1]))
-            ptr_array = int(ffi.cast("uintptr_t", self.c_arrays[-1]))
+            batch = next(data_iter)
+            c_schemas.append(ffi.new("struct ArrowSchema*"))
+            c_arrays.append(ffi.new("struct ArrowArray*"))
+            ptr_schema = int(ffi.cast("uintptr_t", c_schemas[-1]))
+            ptr_array = int(ffi.cast("uintptr_t", c_arrays[-1]))
             # pylint: disable=protected-access
             batch._export_to_c(ptr_array, ptr_schema)
             _check_call(
@@ -511,6 +508,8 @@ def next(self, data_handle: int) -> int:
         except StopIteration:
             return 0
 
+    return _next
+
 
 def _from_arrow(
     data,
@@ -538,8 +537,8 @@ def _from_arrow(
         # use_async=True to workaround pyarrow 6.0.1 hang,
         # see Modin-3982 and ARROW-15362
         rb_iter = iter(data.to_batches(use_async=True))
-    it = RecordBatchDataIter(rb_iter)
-    next_callback = ctypes.CFUNCTYPE(ctypes.c_int, ctypes.c_void_p)(it.next)
+    it = record_batch_data_iter(rb_iter)
+    next_callback = ctypes.CFUNCTYPE(ctypes.c_int, ctypes.c_void_p)(it)
     handle = ctypes.c_void_p()
 
     config = bytes(json.dumps({"missing": missing, "nthread": nthread}), "utf-8")