diff --git a/dspy/clients/lm.py b/dspy/clients/lm.py
index 349e653396..35c623deec 100644
--- a/dspy/clients/lm.py
+++ b/dspy/clients/lm.py
@@ -278,11 +278,11 @@ def func_cached(key: str, request: Dict[str, Any], *args, **kwargs):
         def wrapper(request: dict, *args, **kwargs):
             try:
                 key = cache_key(request)
-                return func_cached(key, request, *args, **kwargs)
             except Exception:
                 # If the cache key cannot be computed (e.g. because it contains a value that cannot
                 # be converted to JSON), bypass the cache and call the target function directly
                 return func(request, *args, **kwargs)
+            return func_cached(key, request, *args, **kwargs)
 
         return wrapper
 
diff --git a/tests/caching/test_caching.py b/tests/caching/test_caching.py
index 023fe45dcb..f890dfad36 100644
--- a/tests/caching/test_caching.py
+++ b/tests/caching/test_caching.py
@@ -150,3 +150,16 @@ def test_lm_calls_with_callables_are_cached_as_expected():
         lm_without_callable("Query")
 
         assert mock_completion.call_count == 2
+
+
+def test_lms_called_expected_number_of_times_for_cache_key_generation_failures():
+    with pytest.raises(Exception), patch("litellm.completion") as mock_completion:
+        mock_completion.side_effect = Exception("Mocked exception")
+        lm = dspy.LM(
+            model="openai/dspy-test-model",
+            api_base="fakebase",
+            api_key="fakekey",
+        )
+        lm("Do not retry")
+
+    assert mock_completion.call_count == 1