diff --git a/src/util/mpz.cpp b/src/util/mpz.cpp
index 9441d48617..38beee43a5 100644
--- a/src/util/mpz.cpp
+++ b/src/util/mpz.cpp
@@ -34,14 +34,14 @@ static bool mul_overflows(int64_t a, int64_t b, int64_t & result) {
     return std::ckd_mul(&result, a, b);
 #elif defined(__GNUC__)
     return __builtin_mul_overflow(a, b, &result);
-#elif defined(_MSC_VER) && !defined(_M_ARM64)
-    // MSVC _mul128 intrinsic is unavailable when targeting ARM64.
+#elif defined(_MSC_VER) && !defined(_M_ARM64) && !defined(_M_ARM64EC)
+    // MSVC _mul128 is available on the non-ARM64 targets that still use this path.
     __int64 high;
     result = _mul128(a, b, &high);
     // Overflow if high bits are not the sign extension of result
     return high != (result < 0 ? -1 : 0);
 #elif defined(_MSC_VER)
-    // `-(v + 1) + 1` computes |v| without ever evaluating -INT64_MIN.
+    // `-(v + 1) + 1` computes |v| even for INT64_MIN, where direct negation would overflow.
     uint64_t x = a < 0 ? static_cast<uint64_t>(-(a + 1)) + 1 : static_cast<uint64_t>(a);
     uint64_t y = b < 0 ? static_cast<uint64_t>(-(b + 1)) + 1 : static_cast<uint64_t>(b);
     bool is_neg = (a < 0) != (b < 0);