diff --git a/doc/source/whatsnew/v1.3.0.rst b/doc/source/whatsnew/v1.3.0.rst
index 21e6f0ea57451..8b0fec2f2c6f9 100644
--- a/doc/source/whatsnew/v1.3.0.rst
+++ b/doc/source/whatsnew/v1.3.0.rst
@@ -682,7 +682,7 @@ Missing
 
 - Bug in :class:`Grouper` now correctly propagates ``dropna`` argument and :meth:`DataFrameGroupBy.transform` now correctly handles missing values for ``dropna=True`` (:issue:`35612`)
 - Bug in :func:`isna`, and :meth:`Series.isna`, :meth:`Index.isna`, :meth:`DataFrame.isna` (and the corresponding ``notna`` functions) not recognizing ``Decimal("NaN")`` objects (:issue:`39409`)
--
+- Bug in :meth:`DataFrame.fillna` not accepting dictionary for ``downcast`` keyword (:issue:`40809`)
 
 MultiIndex
 ^^^^^^^^^^
diff --git a/pandas/core/generic.py b/pandas/core/generic.py
index 4ef5aa1109074..611da0c1ef284 100644
--- a/pandas/core/generic.py
+++ b/pandas/core/generic.py
@@ -6540,11 +6540,13 @@ def fillna(
                     )
 
                 result = self if inplace else self.copy()
+                is_dict = isinstance(downcast, dict)
                 for k, v in value.items():
                     if k not in result:
                         continue
                     obj = result[k]
-                    obj.fillna(v, limit=limit, inplace=True, downcast=downcast)
+                    downcast_k = downcast if not is_dict else downcast.get(k)
+                    obj.fillna(v, limit=limit, inplace=True, downcast=downcast_k)
                 return result if not inplace else None
 
             elif not is_list_like(value):
diff --git a/pandas/tests/frame/methods/test_fillna.py b/pandas/tests/frame/methods/test_fillna.py
index 564481d01abc8..b827547b0753e 100644
--- a/pandas/tests/frame/methods/test_fillna.py
+++ b/pandas/tests/frame/methods/test_fillna.py
@@ -531,6 +531,13 @@ def test_fill_corner(self, float_frame, float_string_frame):
         # TODO(wesm): unused?
         result = empty_float.fillna(value=0)  # noqa
 
+    def test_fillna_downcast_dict(self):
+        # GH#40809
+        df = DataFrame({"col1": [1, np.nan]})
+        result = df.fillna({"col1": 2}, downcast={"col1": "int64"})
+        expected = DataFrame({"col1": [1, 2]})
+        tm.assert_frame_equal(result, expected)
+
 
 def test_fillna_nonconsolidated_frame():
     # https://github.com/pandas-dev/pandas/issues/36495