diff --git a/libc/src/__support/math/exp.h b/libc/src/__support/math/exp.h
index 5c43e753ea687..5c8ea089635ef 100644
--- a/libc/src/__support/math/exp.h
+++ b/libc/src/__support/math/exp.h
@@ -67,7 +67,7 @@ namespace {
 // Return expm1(dx) / x ~ 1 + dx / 2 + dx^2 / 6 + dx^3 / 24.
 // For |dx| < 2^-13 + 2^-30:
 //   | output - expm1(dx) / dx | < 2^-51.
-static constexpr double poly_approx_d(double dx) {
+static double poly_approx_d(double dx) {
   // dx^2
   double dx2 = dx * dx;
   // c0 = 1 + dx / 2
@@ -85,7 +85,7 @@ static constexpr double poly_approx_d(double dx) {
 // Return exp(dx) ~ 1 + dx + dx^2 / 2 + ... + dx^6 / 720
 // For |dx| < 2^-13 + 2^-30:
 //   | output - exp(dx) | < 2^-101
-static constexpr DoubleDouble poly_approx_dd(const DoubleDouble &dx) {
+static DoubleDouble poly_approx_dd(const DoubleDouble &dx) {
   // Taylor polynomial.
   constexpr DoubleDouble COEFFS[] = {
       {0, 0x1p0},                                      // 1
@@ -106,7 +106,7 @@ static constexpr DoubleDouble poly_approx_dd(const DoubleDouble &dx) {
 // Return exp(dx) ~ 1 + dx + dx^2 / 2 + ... + dx^7 / 5040
 // For |dx| < 2^-13 + 2^-30:
 //   | output - exp(dx) | < 2^-126.
-static constexpr Float128 poly_approx_f128(const Float128 &dx) {
+static Float128 poly_approx_f128(const Float128 &dx) {
   constexpr Float128 COEFFS_128[]{
       {Sign::POS, -127, 0x80000000'00000000'00000000'00000000_u128}, // 1.0
       {Sign::POS, -127, 0x80000000'00000000'00000000'00000000_u128}, // 1.0
@@ -127,7 +127,7 @@ static constexpr Float128 poly_approx_f128(const Float128 &dx) {
 // Compute exp(x) using 128-bit precision.
 // TODO(lntue): investigate triple-double precision implementation for this
 // step.
-static constexpr Float128 exp_f128(double x, double kd, int idx1, int idx2) {
+static Float128 exp_f128(double x, double kd, int idx1, int idx2) {
   // Recalculate dx:
 
   double t1 = fputil::multiply_add(kd, MLOG_2_EXP2_M12_HI, x); // exact
@@ -160,7 +160,7 @@ static constexpr Float128 exp_f128(double x, double kd, int idx1, int idx2) {
 }
 
 // Compute exp(x) with double-double precision.
-static constexpr DoubleDouble exp_double_double(double x, double kd,
+static DoubleDouble exp_double_double(double x, double kd,
                                                 const DoubleDouble &exp_mid) {
   // Recalculate dx:
   //   dx = x - k * 2^-12 * log(2)
@@ -184,7 +184,7 @@ static constexpr DoubleDouble exp_double_double(double x, double kd,
 
 // Check for exceptional cases when
 // |x| <= 2^-53 or x < log(2^-1075) or x >= 0x1.6232bdd7abcd3p+9
-static constexpr double set_exceptional(double x) {
+static double set_exceptional(double x) {
   using FPBits = typename fputil::FPBits<double>;
   FPBits xbits(x);
 
@@ -234,7 +234,7 @@ static constexpr double set_exceptional(double x) {
 
 namespace math {
 
-static constexpr double exp(double x) {
+static double exp(double x) {
   using FPBits = typename fputil::FPBits<double>;
   FPBits xbits(x);