diff --git a/xarray/computation/nanops.py b/xarray/computation/nanops.py
index a28078540bb..c5a9d9bcc04 100644
--- a/xarray/computation/nanops.py
+++ b/xarray/computation/nanops.py
@@ -29,7 +29,9 @@ def _maybe_null_out(result, axis, mask, min_count=1):
         dtype, fill_value = dtypes.maybe_promote(result.dtype)
         result = where(null_mask, fill_value, astype(result, dtype))
 
-    elif getattr(result, "dtype", None) not in dtypes.NAT_TYPES:
+    elif (dtype := getattr(result, "dtype", None)) and getattr(
+        dtype, "kind", None
+    ) not in {"m", "M"}:
         null_mask = mask.size - duck_array_ops.sum(mask)
         result = where(null_mask < min_count, np.nan, result)
 
diff --git a/xarray/core/dtypes.py b/xarray/core/dtypes.py
index 40334d58e68..bb2fe26d727 100644
--- a/xarray/core/dtypes.py
+++ b/xarray/core/dtypes.py
@@ -10,6 +10,7 @@
 from xarray.compat import array_api_compat, npcompat
 from xarray.compat.npcompat import HAS_STRING_DTYPE
 from xarray.core import utils
+from xarray.core.types import PDDatetimeUnitOptions
 
 if TYPE_CHECKING:
     from typing import Any
@@ -88,7 +89,11 @@ def maybe_promote(dtype: T_dtype) -> tuple[T_dtype, Any]:
         # See https://github.com/numpy/numpy/issues/10685
         # np.timedelta64 is a subclass of np.integer
         # Check np.timedelta64 before np.integer
-        fill_value = np.timedelta64("NaT")
+        unit, _ = np.datetime_data(dtype)
+        # np.datetime_data returns a generic str for the unit so we need to
+        # cast it to a valid time unit for mypy purposes.
+        unit = cast(PDDatetimeUnitOptions, unit)
+        fill_value = np.timedelta64("NaT", unit)
         dtype_ = dtype
     elif isdtype(dtype, "integral"):
         dtype_ = np.float32 if dtype.itemsize <= 2 else np.float64
@@ -97,8 +102,12 @@ def maybe_promote(dtype: T_dtype) -> tuple[T_dtype, Any]:
         dtype_ = dtype
         fill_value = np.nan + np.nan * 1j
     elif np.issubdtype(dtype, np.datetime64):
+        unit, _ = np.datetime_data(dtype)
+        # np.datetime_data returns a generic str for the unit so we need to
+        # cast it to a valid time unit for mypy purposes.
+        unit = cast(PDDatetimeUnitOptions, unit)
         dtype_ = dtype
-        fill_value = np.datetime64("NaT")
+        fill_value = np.datetime64("NaT", unit)
     else:
         dtype_ = object
         fill_value = np.nan
@@ -108,9 +117,6 @@ def maybe_promote(dtype: T_dtype) -> tuple[T_dtype, Any]:
     return dtype_out, fill_value
 
 
-NAT_TYPES = {np.datetime64("NaT").dtype, np.timedelta64("NaT").dtype}
-
-
 def get_fill_value(dtype):
     """Return an appropriate fill value for this dtype.
 
diff --git a/xarray/plot/utils.py b/xarray/plot/utils.py
index 167718fe9d8..59d1c5d3c57 100644
--- a/xarray/plot/utils.py
+++ b/xarray/plot/utils.py
@@ -189,6 +189,15 @@ def _determine_cmap_params(
     else:
         mpl = attempt_import("matplotlib")
 
+    if plot_data.dtype.kind == "m":
+        unit, _ = np.datetime_data(plot_data.dtype)
+        zero = np.timedelta64(0, unit)
+    elif plot_data.dtype.kind == "M":
+        unit, _ = np.datetime_data(plot_data.dtype)
+        zero = np.datetime64(0, unit)
+    else:
+        zero = 0.0
+
     if isinstance(levels, Iterable):
         levels = sorted(levels)
 
@@ -197,7 +206,7 @@ def _determine_cmap_params(
     # Handle all-NaN input data gracefully
     if calc_data.size == 0:
         # Arbitrary default for when all values are NaN
-        calc_data = np.array(0.0)
+        calc_data = np.array(zero)
 
     # Setting center=False prevents a divergent cmap
     possibly_divergent = center is not False
@@ -205,7 +214,7 @@ def _determine_cmap_params(
     # Set center to 0 so math below makes sense but remember its state
     center_is_none = False
     if center is None:
-        center = 0
+        center = zero
         center_is_none = True
 
     # Setting both vmin and vmax prevents a divergent cmap
@@ -240,10 +249,10 @@ def _determine_cmap_params(
 
     if possibly_divergent:
         levels_are_divergent = (
-            isinstance(levels, Iterable) and levels[0] * levels[-1] < 0
+            isinstance(levels, Iterable) and levels[0] * levels[-1] < zero
         )
         # kwargs not specific about divergent or not: infer defaults from data
-        divergent = (vmin < 0 < vmax) or not center_is_none or levels_are_divergent
+        divergent = (vmin < zero < vmax) or not center_is_none or levels_are_divergent
     else:
         divergent = False
 
diff --git a/xarray/tests/test_dtypes.py b/xarray/tests/test_dtypes.py
index a1bdd9082ef..4ed66509725 100644
--- a/xarray/tests/test_dtypes.py
+++ b/xarray/tests/test_dtypes.py
@@ -102,8 +102,8 @@ def test_inf(obj) -> None:
         ("I", (np.float64, "nan")),  # dtype('uint32')
         ("l", (np.float64, "nan")),  # dtype('int64')
         ("L", (np.float64, "nan")),  # dtype('uint64')
-        ("m", (np.timedelta64, "NaT")),  # dtype('<m8')
-        ("M", (np.datetime64, "NaT")),  # dtype('<M8')
+        ("<m8[ns]", (np.dtype("<m8[ns]"), "NaT")),  # dtype('<m8[ns]')
+        ("<M8[ns]", (np.dtype("<M8[ns]"), "NaT")),  # dtype('<M8[ns]')
         ("O", (np.dtype("O"), "nan")),  # dtype('O')
         ("p", (np.float64, "nan")),  # dtype('int64')
         ("P", (np.float64, "nan")),  # dtype('uint64')
@@ -123,12 +123,6 @@ def test_maybe_promote(kind, expected) -> None:
     assert str(actual[1]) == expected[1]
 
 
-def test_nat_types_membership() -> None:
-    assert np.datetime64("NaT").dtype in dtypes.NAT_TYPES
-    assert np.timedelta64("NaT").dtype in dtypes.NAT_TYPES
-    assert np.float64 not in dtypes.NAT_TYPES
-
-
 @pytest.mark.parametrize(
     ["dtype", "kinds", "xp", "expected"],
     (