diff --git a/pyupgrade/_plugins/fstrings.py b/pyupgrade/_plugins/fstrings.py
index 3abf1fe6..01a07b25 100644
--- a/pyupgrade/_plugins/fstrings.py
+++ b/pyupgrade/_plugins/fstrings.py
@@ -4,6 +4,7 @@
 from typing import Iterable
 
 from tokenize_rt import Offset
+from tokenize_rt import parse_string_literal
 from tokenize_rt import Token
 from tokenize_rt import tokens_to_src
 
@@ -43,7 +44,12 @@ def _to_fstring(
 
     parts = []
     i = 0
-    for s, name, spec, conv in parse_format('f' + src):
+
+    # need to remove `u` prefix so it isn't invalid syntax
+    prefix, rest = parse_string_literal(src)
+    new_src = 'f' + prefix.translate({ord('u'): None, ord('U'): None}) + rest
+
+    for s, name, spec, conv in parse_format(new_src):
         if name is not None:
             k, dot, rest = name.partition('.')
             name = ''.join((params[k or str(i)], dot, rest))
diff --git a/tests/features/fstrings_test.py b/tests/features/fstrings_test.py
index 6c1212c2..51e9179c 100644
--- a/tests/features/fstrings_test.py
+++ b/tests/features/fstrings_test.py
@@ -64,6 +64,11 @@ def test_fix_fstrings_noop(s):
             r'f"\N{snowman} {a}"',
             id='named escape sequences',
         ),
+        pytest.param(
+            'u"foo{}".format(1)',
+            'f"foo{1}"',
+            id='u-prefixed format',
+        ),
     ),
 )
 def test_fix_fstrings(s, expected):