diff --git a/IPython/terminal/interactiveshell.py b/IPython/terminal/interactiveshell.py
index eb49e67..0dda428 100644
--- a/IPython/terminal/interactiveshell.py
+++ b/IPython/terminal/interactiveshell.py
@@ -380,6 +380,16 @@ class TerminalInteractiveShell(InteractiveShell):
         def get_message():
             return PygmentsTokens(self.prompts.in_prompt_tokens())
 
+        if self.editing_mode == 'emacs':
+            # with emacs mode the prompt is (usually) static, so we call only
+            # the function once. With VI mode it can toggle between [ins] and
+            # [nor] so we can't precompute.
+            # here I'm going to favor the default keybinding which almost
+            # everybody uses to decrease CPU usage.
+            # if we have issues with users with custom Prompts we can see how to
+            # work around this.
+            get_message = get_message()
+
         return {
                 'complete_in_thread': False,
                 'lexer':IPythonPTLexer(),
diff --git a/docs/source/whatsnew/pr/prompt-update-modifications.rst b/docs/source/whatsnew/pr/prompt-update-modifications.rst
new file mode 100644
index 0000000..77d557b
--- /dev/null
+++ b/docs/source/whatsnew/pr/prompt-update-modifications.rst
@@ -0,0 +1,14 @@
+Prompt Rendering Performance improvements
+=========================================
+
+Pull Request :ghpull:`11933` introduced an optimisation in the prompt rendering
+logic that should decrease the resource usage of IPython when using the
+_default_ configuration but could potentially introduce a regression of
+functionalities if you are using a custom prompt.
+
+We know assume if you haven't changed the default keybindings that the prompt
+**will not change** during the duration of your input – which is for example
+not true when using vi insert mode that switches between `[ins]` and `[nor]`
+for the current mode.
+
+If you are experiencing any issue let us know.