diff --git a/sklearn/decomposition/_dict_learning.py b/sklearn/decomposition/_dict_learning.py
index f0e070d485e81..2da32c9b0d907 100644
--- a/sklearn/decomposition/_dict_learning.py
+++ b/sklearn/decomposition/_dict_learning.py
@@ -893,7 +893,8 @@ def dict_learning_online(
         dictionary = dictionary[:n_components, :]
     else:
         dictionary = np.r_[
-            dictionary, np.zeros((n_components - r, dictionary.shape[1]))
+            dictionary,
+            np.zeros((n_components - r, dictionary.shape[1]), dtype=dictionary.dtype),
         ]
 
     if verbose == 1:
@@ -905,25 +906,23 @@ def dict_learning_online(
     else:
         X_train = X
 
-    # Fortran-order dict better suited for the sparse coding which is the
-    # bottleneck of this algorithm.
-    dictionary = check_array(
-        dictionary, order="F", dtype=[np.float64, np.float32], copy=False
-    )
-    dictionary = np.require(dictionary, requirements="W")
-
     X_train = check_array(
         X_train, order="C", dtype=[np.float64, np.float32], copy=False
     )
 
+    # Fortran-order dict better suited for the sparse coding which is the
+    # bottleneck of this algorithm.
+    dictionary = check_array(dictionary, order="F", dtype=X_train.dtype, copy=False)
+    dictionary = np.require(dictionary, requirements="W")
+
     batches = gen_batches(n_samples, batch_size)
     batches = itertools.cycle(batches)
 
     # The covariance of the dictionary
     if inner_stats is None:
-        A = np.zeros((n_components, n_components))
+        A = np.zeros((n_components, n_components), dtype=X_train.dtype)
         # The data approximation
-        B = np.zeros((n_features, n_components))
+        B = np.zeros((n_features, n_components), dtype=X_train.dtype)
     else:
         A = inner_stats[0].copy()
         B = inner_stats[1].copy()
diff --git a/sklearn/decomposition/tests/test_dict_learning.py b/sklearn/decomposition/tests/test_dict_learning.py
index b7ee66d0e78cb..456379eee6d08 100644
--- a/sklearn/decomposition/tests/test_dict_learning.py
+++ b/sklearn/decomposition/tests/test_dict_learning.py
@@ -740,6 +740,10 @@ def test_dictionary_learning_dtype_match(
     assert dict_learner.components_.dtype == expected_type
     assert dict_learner.transform(X.astype(data_type)).dtype == expected_type
 
+    if dictionary_learning_transformer is MiniBatchDictionaryLearning:
+        assert dict_learner.inner_stats_[0].dtype == expected_type
+        assert dict_learner.inner_stats_[1].dtype == expected_type
+
 
 @pytest.mark.parametrize("method", ("lars", "cd"))
 @pytest.mark.parametrize(