diff --git a/numpy_financial/_financial.py b/numpy_financial/_financial.py
index 08220b2..2a38531 100644
--- a/numpy_financial/_financial.py
+++ b/numpy_financial/_financial.py
@@ -112,6 +112,15 @@ def _ufunc_like(array: np.generic | npt.NDArray[Any]) -> Any:
         # Otherwise, return entire array
         return array.squeeze()
 
+
+def _float_power_base(base: Any) -> Any:
+    # Integer ``**`` wraps around on overflow and rejects negative
+    # exponents, so evaluate integer and boolean bases in float64.
+    # Other bases, including Decimal and object arrays, are unchanged.
+    if isinstance(base, (np.ndarray, np.generic)) and base.dtype.kind in "biu":
+        return base.astype(np.float64)
+    return base
+
 @overload
 def _convert_when(when: _ArrayT) -> _ArrayT: ...
 @overload
@@ -253,7 +262,7 @@ def fv(rate, nper, pmt, pv, when: _When = 'end'):
     fv_array[zero] = -(pv[zero] + pmt[zero] * nper[zero])
 
     rate_nonzero = rate[nonzero]
-    temp = (1 + rate_nonzero) ** nper[nonzero]
+    temp = _float_power_base(1 + rate_nonzero) ** nper[nonzero]
     fv_array[nonzero] = (
             - pv[nonzero] * temp
             - pmt[nonzero] * (1 + rate_nonzero * when[nonzero]) / rate_nonzero
@@ -393,7 +402,7 @@ def pmt(rate, nper, pv, fv: Any = 0, when: _When = 'end'):
     """
     when = _convert_when(when)
     (rate, nper, pv, fv, when) = map(np.array, [rate, nper, pv, fv, when])
-    temp = (1 + rate) ** nper
+    temp = _float_power_base(1 + rate) ** nper
     mask = (rate == 0)
     masked_rate = np.where(mask, 1, rate)
     fact = np.where(mask != 0, nper,
@@ -880,7 +889,7 @@ def pv(rate, nper, pmt, fv: Any = 0, when: _When = 'end'):
     """
     when = _convert_when(when)
     rate, nper, pmt, fv, when = map(np.asarray, [rate, nper, pmt, fv, when])
-    temp = (1 + rate) ** nper
+    temp = _float_power_base(1 + rate) ** nper
     fact = np.where(rate == 0, nper, (1 + rate * when) * (temp - 1) / rate)
     return -(fv + pmt * fact) / temp
 
diff --git a/numpy_financial/tests/test_financial.py b/numpy_financial/tests/test_financial.py
index 765e1b2..c245f52 100644
--- a/numpy_financial/tests/test_financial.py
+++ b/numpy_financial/tests/test_financial.py
@@ -750,6 +750,37 @@ def test_some_rates_zero(self):
         )
 
 
+class TestIntegerRates:
+    # (1 + rate) ** nper must not be evaluated with integer arithmetic,
+    # which wraps around on overflow. 2**64 wraps to 0 in int64.
+    @pytest.mark.parametrize(
+        ("func", "args", "desired"),
+        [
+            (npf.fv, (64, 0, -(2.0**-64)), 1.0),
+            (npf.pv, (64, -1, 0), 1 - 2.0**-64),
+            (npf.pmt, (64, 1), -1 / (1 - 2.0**-64)),
+        ],
+    )
+    def test_no_overflow(self, func, args, desired):
+        assert_allclose(func(1, *args), desired, rtol=1e-15)
+        assert_allclose(func(1, *args), func(1.0, *args), rtol=1e-15)
+
+    @pytest.mark.parametrize("dtype", [numpy.int32, numpy.int64])
+    def test_integer_arrays(self, dtype):
+        rate = numpy.array([1, 2], dtype=dtype)
+        nper = numpy.array([64, 41], dtype=dtype)
+        assert_allclose(npf.pv(rate, nper, -1, 0), [1.0, 0.5], rtol=1e-15)
+        assert_allclose(npf.pmt(rate, nper, 1), [-1.0, -2.0], rtol=1e-15)
+        assert_allclose(npf.fv(rate[:1], nper[:1], 0, -(2.0**-64)), [1.0])
+
+    @pytest.mark.parametrize(
+        ("func", "args"), [(npf.pv, (-1, 0, -1)), (npf.pmt, (-2, 1))]
+    )
+    def test_negative_integer_nper(self, func, args):
+        # Integer arrays reject negative integer powers.
+        assert_allclose(func(1, *args), func(1.0, *args), rtol=1e-15)
+
+
 class TestIrr:
     def test_npv_irr_congruence(self):
         # IRR is defined as the rate required for the present value of