diff --git a/pandas/_libs/tslibs/tzconversion.pyx b/pandas/_libs/tslibs/tzconversion.pyx
index 030549f2528ed..575088b8ab2cc 100644
--- a/pandas/_libs/tslibs/tzconversion.pyx
+++ b/pandas/_libs/tslibs/tzconversion.pyx
@@ -557,30 +557,25 @@ cdef int64_t[:] _tz_convert_dst(
         ndarray[int64_t] trans
         int64_t[:] deltas
         int64_t v
-        bint tz_is_local
 
-    tz_is_local = is_tzlocal(tz)
+    # tz is assumed _not_ to be tzlocal; that should go
+    #  through _tz_convert_tzlocal_utc
 
-    if not tz_is_local:
-        # get_dst_info cannot extract offsets from tzlocal because its
-        # dependent on a datetime
-        trans, deltas, _ = get_dst_info(tz)
-        if not to_utc:
-            # We add `offset` below instead of subtracting it
-            deltas = -1 * np.array(deltas, dtype='i8')
+    trans, deltas, _ = get_dst_info(tz)
+    if not to_utc:
+        # We add `offset` below instead of subtracting it
+        deltas = -1 * np.array(deltas, dtype='i8')
 
-        # Previously, this search was done pointwise to try and benefit
-        # from getting to skip searches for iNaTs. However, it seems call
-        # overhead dominates the search time so doing it once in bulk
-        # is substantially faster (GH#24603)
-        pos = trans.searchsorted(values, side='right') - 1
+    # Previously, this search was done pointwise to try and benefit
+    # from getting to skip searches for iNaTs. However, it seems call
+    # overhead dominates the search time so doing it once in bulk
+    # is substantially faster (GH#24603)
+    pos = trans.searchsorted(values, side='right') - 1
 
     for i in range(n):
         v = values[i]
         if v == NPY_NAT:
             result[i] = v
-        elif tz_is_local:
-            result[i] = _tz_convert_tzlocal_utc(v, tz, to_utc=to_utc)
         else:
             if pos[i] < 0:
                 raise ValueError('First time before start of DST info')