diff --git a/doc/source/whatsnew/v2.0.0.rst b/doc/source/whatsnew/v2.0.0.rst
index 90a00cd01d665..909e2454b7cb8 100644
--- a/doc/source/whatsnew/v2.0.0.rst
+++ b/doc/source/whatsnew/v2.0.0.rst
@@ -900,6 +900,7 @@ Performance improvements
 - Performance improvements to :func:`read_sas` (:issue:`47403`, :issue:`47405`, :issue:`47656`, :issue:`48502`)
 - Memory improvement in :meth:`RangeIndex.sort_values` (:issue:`48801`)
 - Performance improvement in :meth:`Series.to_numpy` if ``copy=True`` by avoiding copying twice (:issue:`24345`)
+- Performance improvement in :meth:`Series.rename` with :class:`MultiIndex` (:issue:`21055`)
 - Performance improvement in :class:`DataFrameGroupBy` and :class:`SeriesGroupBy` when ``by`` is a categorical type and ``sort=False`` (:issue:`48976`)
 - Performance improvement in :class:`DataFrameGroupBy` and :class:`SeriesGroupBy` when ``by`` is a categorical type and ``observed=False`` (:issue:`49596`)
 - Performance improvement in :func:`read_stata` with parameter ``index_col`` set to ``None`` (the default). Now the index will be a :class:`RangeIndex` instead of :class:`Int64Index` (:issue:`49745`)
@@ -1020,6 +1021,7 @@ Indexing
 - Bug in :meth:`DataFrame.iloc` raising ``IndexError`` when indexer is a :class:`Series` with numeric extension array dtype (:issue:`49521`)
 - Bug in :func:`~DataFrame.describe` when formatting percentiles in the resulting index showed more decimals than needed (:issue:`46362`)
 - Bug in :meth:`DataFrame.compare` does not recognize differences when comparing ``NA`` with value in nullable dtypes (:issue:`48939`)
+- Bug in :meth:`Series.rename` with :class:`MultiIndex` losing extension array dtypes (:issue:`21055`)
 - Bug in :meth:`DataFrame.isetitem` coercing extension array dtypes in :class:`DataFrame` to object (:issue:`49922`)
 - Bug in :class:`BusinessHour` would cause creation of :class:`DatetimeIndex` to fail when no opening hour was included in the index (:issue:`49835`)
 -
diff --git a/pandas/core/indexes/base.py b/pandas/core/indexes/base.py
index b05b7cae8042e..2948bb81d0b6a 100644
--- a/pandas/core/indexes/base.py
+++ b/pandas/core/indexes/base.py
@@ -6042,15 +6042,13 @@ def _transform_index(self, func, *, level=None) -> Index:
         Only apply function to one level of the MultiIndex if level is specified.
         """
         if isinstance(self, ABCMultiIndex):
-            if level is not None:
-                # Caller is responsible for ensuring level is positional.
-                items = [
-                    tuple(func(y) if i == level else y for i, y in enumerate(x))
-                    for x in self
-                ]
-            else:
-                items = [tuple(func(y) for y in x) for x in self]
-            return type(self).from_tuples(items, names=self.names)
+            values = [
+                self.get_level_values(i).map(func)
+                if i == level or level is None
+                else self.get_level_values(i)
+                for i in range(self.nlevels)
+            ]
+            return type(self).from_arrays(values)
         else:
             items = [func(x) for x in self]
             return Index(items, name=self.name, tupleize_cols=False)
diff --git a/pandas/tests/series/methods/test_rename.py b/pandas/tests/series/methods/test_rename.py
index d0392929cb082..93c4fbb7f3c46 100644
--- a/pandas/tests/series/methods/test_rename.py
+++ b/pandas/tests/series/methods/test_rename.py
@@ -136,6 +136,25 @@ def test_rename_series_with_multiindex(self):
 
         tm.assert_series_equal(result, series_expected)
 
+    def test_rename_series_with_multiindex_keeps_ea_dtypes(self):
+        # GH21055
+        arrays = [
+            Index([1, 2, 3], dtype="Int64").astype("category"),
+            Index([1, 2, 3], dtype="Int64"),
+        ]
+        mi = MultiIndex.from_arrays(arrays, names=["A", "B"])
+        ser = Series(1, index=mi)
+        result = ser.rename({1: 4}, level=1)
+
+        arrays_expected = [
+            Index([1, 2, 3], dtype="Int64").astype("category"),
+            Index([4, 2, 3], dtype="Int64"),
+        ]
+        mi_expected = MultiIndex.from_arrays(arrays_expected, names=["A", "B"])
+        expected = Series(1, index=mi_expected)
+
+        tm.assert_series_equal(result, expected)
+
     def test_rename_error_arg(self):
         # GH 46889
         ser = Series(["foo", "bar"])