diff --git a/poincare/src/multiplication.cpp b/poincare/src/multiplication.cpp
index 9db2b6c43..e3dfcf3f7 100644
--- a/poincare/src/multiplication.cpp
+++ b/poincare/src/multiplication.cpp
@@ -440,10 +440,9 @@ Expression Multiplication::shallowBeautify(ExpressionNode::ReductionContext redu
        * - Repeat those steps until no more simplification is possible.
        */
       Multiplication unitsAccu = Multiplication::Builder();
-      /* If exponents are not integers, FromBaseUnits will return the closest
-       * representation of units with base units and integer exponents.
-       * It cause no problem because once the best derived units are found,
-       * units is divided then multiplied by them. */
+      /* If exponents are not integers, FromBaseUnits will return a null
+       * vector, preventing any attempt at simplification. This protects us
+       * against undue "simplifications" such as _C^1.3 -> _C*_A^0.3*_s^0.3 */
       UnitNode::Vector<int> unitsExponents = UnitNode::Vector<int>::FromBaseUnits(units);
       size_t unitsSupportSize = unitsExponents.supportSize();
       UnitNode::Vector<int> bestRemainderExponents;
diff --git a/poincare/src/unit.cpp b/poincare/src/unit.cpp
index 9bb2370d0..901085695 100644
--- a/poincare/src/unit.cpp
+++ b/poincare/src/unit.cpp
@@ -128,6 +128,12 @@ UnitNode::Vector<int> UnitNode::Vector<int>::FromBaseUnits(const Expression base
       assert(exp.type() == ExpressionNode::Type::Rational);
       // Using the closest integer to the exponent.
       float exponentFloat = static_cast<const Rational &>(exp).node()->templatedApproximate<float>();
+      if (exponentFloat != std::round(exponentFloat)) {
+        /* If non-integer exponents are found, we round a null vector so that
+         * Multiplication::shallowBeautify will not attempt to find derived
+         * units. */
+        return vector;
+      }
       /* We limit to INT_MAX / 3 because an exponent might get bigger with
        * simplification. As a worst case scenario, (_s²_m²_kg/_A²)^n should be
        * simplified to (_s^5_S)^n. If 2*n is under INT_MAX, 5*n might not. */
diff --git a/poincare/test/simplification.cpp b/poincare/test/simplification.cpp
index b8bc06276..8ccc7718b 100644
--- a/poincare/test/simplification.cpp
+++ b/poincare/test/simplification.cpp
@@ -361,6 +361,13 @@ QUIZ_CASE(poincare_simplification_units) {
   assert_parsed_expression_simplify_to("1/(-3_°C)", Undefined::Name());
   assert_parsed_expression_simplify_to("-1×100×_°C→_K", Undefined::Name());
 
+  /* Rational exponents */
+  assert_parsed_expression_simplify_to("√(_m)", "1×_m^\u00121/2\u0013");
+  assert_parsed_expression_simplify_to("√(_N)", "1×_kg^\u00121/2\u0013×_m^\u00121/2\u0013×_s^\u0012-1\u0013");
+  assert_parsed_expression_simplify_to("√(_N)", "1.5527410012845×_lb^\u00121/2\u0013×_yd^\u00121/2\u0013×_s^\u0012-1\u0013", User, Radian, Imperial);
+  assert_parsed_expression_simplify_to("_C^0.3", "1×_A^\u00123/10\u0013×_s^\u00123/10\u0013");
+  assert_parsed_expression_simplify_to("_kat_kg^-2.8", "1×_mol×_kg^\u0012-14/5\u0013×_s^\u0012-1\u0013");
+
   /* Unit sum/subtract */
   assert_parsed_expression_simplify_to("_m+_m", "2×_m");
   assert_parsed_expression_simplify_to("_m-_m", "0×_m");
@@ -528,6 +535,7 @@ QUIZ_CASE(poincare_simplification_units) {
   assert_parsed_expression_simplify_to("-2×_A", "-2×_A");
   assert_parsed_expression_simplify_to("cos(1_s/1_s)", "cos(1)");
   assert_parsed_expression_simplify_to("1_m+π_m+√(2)_m-cos(15)_m", "6.3154941288217×_m");
+  assert_parsed_expression_simplify_to("√(16×_m^2)", "4×_m");
 }
 
 QUIZ_CASE(poincare_simplification_power) {