diff --git a/llvm/lib/Transforms/InstCombine/InstCombineVectorOps.cpp b/llvm/lib/Transforms/InstCombine/InstCombineVectorOps.cpp
--- a/llvm/lib/Transforms/InstCombine/InstCombineVectorOps.cpp
+++ b/llvm/lib/Transforms/InstCombine/InstCombineVectorOps.cpp
@@ -228,8 +228,9 @@
   // truncate a subset of scalar bits of an insert op.
   if (NumSrcElts.getKnownMinValue() < NumElts.getKnownMinValue()) {
     Value *Scalar;
+    Value *Vec;
     uint64_t InsIndexC;
-    if (!match(X, m_InsertElt(m_Value(), m_Value(Scalar),
+    if (!match(X, m_InsertElt(m_Value(Vec), m_Value(Scalar),
                               m_ConstantInt(InsIndexC))))
       return nullptr;
 
@@ -239,8 +240,19 @@
     // of elements 4-7 of the bitcasted vector.
     unsigned NarrowingRatio =
         NumElts.getKnownMinValue() / NumSrcElts.getKnownMinValue();
-    if (ExtIndexC / NarrowingRatio != InsIndexC)
+
+    if (ExtIndexC / NarrowingRatio != InsIndexC) {
+      // Remove insertelement, if we don't use the inserted element.
+      // extractelement (bitcast (insertelement (Vec, b)), a) ->
+      // extractelement (bitcast (Vec), a)
+      // FIXME: this should be removed to SimplifyDemandedVectorElts,
+      // once scale vectors are supported
+      if (X->hasOneUse() && Ext.getVectorOperand()->hasOneUse()) {
+        Value *NewBC = Builder.CreateBitCast(Vec, Ext.getVectorOperandType());
+        return ExtractElementInst::Create(NewBC, Ext.getIndexOperand());
+      }
       return nullptr;
+    }
 
     // We are extracting part of the original scalar. How that scalar is
     // inserted into the vector depends on the endian-ness. Example:
diff --git a/llvm/test/Transforms/InstCombine/vscale_extractelement-inseltpoison.ll b/llvm/test/Transforms/InstCombine/vscale_extractelement-inseltpoison.ll
--- a/llvm/test/Transforms/InstCombine/vscale_extractelement-inseltpoison.ll
+++ b/llvm/test/Transforms/InstCombine/vscale_extractelement-inseltpoison.ll
@@ -41,12 +41,10 @@
   ret i8 %r
 }
 
-; TODO: Instcombine could remove the insert.
 define i8 @extractelement_bitcast_wrong_insert(<vscale x 2 x i32> %a, i32 %x) {
 ; CHECK-LABEL: @extractelement_bitcast_wrong_insert(
-; CHECK-NEXT:    [[VEC:%.*]] = insertelement <vscale x 2 x i32> [[A:%.*]], i32 [[X:%.*]], i64 1
-; CHECK-NEXT:    [[VEC_CAST:%.*]] = bitcast <vscale x 2 x i32> [[VEC]] to <vscale x 8 x i8>
-; CHECK-NEXT:    [[R:%.*]] = extractelement <vscale x 8 x i8> [[VEC_CAST]], i64 2
+; CHECK-NEXT:    [[TMP1:%.*]] = bitcast <vscale x 2 x i32> [[A:%.*]] to <vscale x 8 x i8>
+; CHECK-NEXT:    [[R:%.*]] = extractelement <vscale x 8 x i8> [[TMP1]], i64 2
 ; CHECK-NEXT:    ret i8 [[R]]
 ;
   %vec = insertelement <vscale x 2 x i32> %a, i32 %x, i32 1 ; <- This insert could be removed.
diff --git a/llvm/test/Transforms/InstCombine/vscale_extractelement.ll b/llvm/test/Transforms/InstCombine/vscale_extractelement.ll
--- a/llvm/test/Transforms/InstCombine/vscale_extractelement.ll
+++ b/llvm/test/Transforms/InstCombine/vscale_extractelement.ll
@@ -41,12 +41,10 @@
   ret i8 %r
 }
 
-; TODO: Instcombine could remove the insert.
 define i8 @extractelement_bitcast_wrong_insert(<vscale x 2 x i32> %a, i32 %x) {
 ; CHECK-LABEL: @extractelement_bitcast_wrong_insert(
-; CHECK-NEXT:    [[VEC:%.*]] = insertelement <vscale x 2 x i32> [[A:%.*]], i32 [[X:%.*]], i64 1
-; CHECK-NEXT:    [[VEC_CAST:%.*]] = bitcast <vscale x 2 x i32> [[VEC]] to <vscale x 8 x i8>
-; CHECK-NEXT:    [[R:%.*]] = extractelement <vscale x 8 x i8> [[VEC_CAST]], i64 2
+; CHECK-NEXT:    [[TMP1:%.*]] = bitcast <vscale x 2 x i32> [[A:%.*]] to <vscale x 8 x i8>
+; CHECK-NEXT:    [[R:%.*]] = extractelement <vscale x 8 x i8> [[TMP1]], i64 2
 ; CHECK-NEXT:    ret i8 [[R]]
 ;
   %vec = insertelement <vscale x 2 x i32> %a, i32 %x, i32 1 ; <- This insert could be removed.