diff --git a/news/13986.feature.rst b/news/13986.feature.rst
new file mode 100644
index 0000000000..70686f847c
--- /dev/null
+++ b/news/13986.feature.rst
@@ -0,0 +1 @@
+Improve package candidate collection performance by optimizing URL parsing.
diff --git a/src/pip/_internal/models/link.py b/src/pip/_internal/models/link.py
index 200ec34c56..ccfa5bf1e9 100644
--- a/src/pip/_internal/models/link.py
+++ b/src/pip/_internal/models/link.py
@@ -13,6 +13,7 @@
 from dataclasses import dataclass
 from typing import (
     Any,
+    Final,
     NamedTuple,
 )
 
@@ -140,6 +141,11 @@ def _clean_file_url_path(part: str) -> str:
 # percent-encoded:                   /
 _reserved_chars_re = re.compile("(@|%2F)", re.IGNORECASE)
 
+# Characters that survive a quote(unquote(part)) round-trip unchanged in a
+# URL path: quote()'s always-safe alphabet plus '/' (the default of the safe
+# argument).
+_UNSAFE_URL_PATH_CHARS_RE: Final[re.Pattern[str]] = re.compile(r"[^A-Za-z0-9_./~\-]")
+
 
 def _clean_url_path(path: str, is_local_path: bool) -> str:
     """
@@ -169,6 +175,17 @@ def _ensure_quoted_url(url: str) -> str:
     For example, if ' ' occurs in the URL, it will be replaced with "%20",
     and without double-quoting other characters.
     """
+    # Fast path: skip quoting round-trip if the path component of a http(s):// link
+    # only contains characters that quote() would leave untouched AND has no %-escapes
+    # for unquote().
+    #
+    # NOTE: we check everything after the scheme because calling urlsplit() to get just
+    # the path component is too costly here.
+    if url.startswith(("https://", "http://")):
+        url_no_scheme = url.removeprefix("https:").removeprefix("http:")
+        if _UNSAFE_URL_PATH_CHARS_RE.search(url_no_scheme) is None:
+            return url
+
     # Split the URL into parts according to the general structure
     # `scheme://netloc/path?query#fragment`.
     result = urllib.parse.urlsplit(url)
diff --git a/tests/unit/test_collector.py b/tests/unit/test_collector.py
index b84fbdb0b3..eaa4f0d017 100644
--- a/tests/unit/test_collector.py
+++ b/tests/unit/test_collector.py
@@ -424,6 +424,36 @@ def test_ensure_quoted_url(url: str, clean_url: str) -> None:
     assert _ensure_quoted_url(url) == clean_url
 
 
+@pytest.mark.parametrize(
+    "url",
+    [
+        pytest.param(
+            "https://files.pythonhosted.org/packages/12/34/somepackage-1.2.3-py3-none-any.whl",
+            id="typical-pypi-wheel",
+        ),
+        pytest.param(
+            "https://files.pythonhosted.org/packages/12/34/somepackage-1.2.3-py3-none-any.whl#sha256=abc",
+            id="pypi-wheel-with-fragment",
+        ),
+        pytest.param(
+            "http://localhost:8181/simple/foo/",
+            id="http-localhost-with-port",
+        ),
+        pytest.param(
+            "https://example.com/path/to/file.tar.gz?build=1",
+            id="https-with-query",
+        ),
+    ],
+)
+def test_ensure_quoted_url_idempotent_for_clean_urls(url: str) -> None:
+    """http(s):// URLs that are pure ASCII with no whitespace and no
+    %-escapes already pass through urlsplit() + urlunsplit() unchanged.
+    The function MUST return the input unchanged for them so callers can rely
+    on it as an identity (the implementation may take a fast path here).
+    """
+    assert _ensure_quoted_url(url) == url
+
+
 def _test_parse_links_data_attribute(
     anchor_html: str, attr: str, expected: str | None
 ) -> Link: