diff --git a/integration_test/circt-synth/comb-lowering-lec.mlir b/integration_test/circt-synth/comb-lowering-lec.mlir
index 4891299e7d63..0d51431e578f 100644
--- a/integration_test/circt-synth/comb-lowering-lec.mlir
+++ b/integration_test/circt-synth/comb-lowering-lec.mlir
@@ -6,9 +6,9 @@
 // COMB_BIT_LOGICAL: c1 == c2
 hw.module @bit_logical(in %arg0: i32, in %arg1: i32, in %arg2: i32, in %arg3: i32,
                 in %cond: i1, out out0: i32, out out1: i32, out out2: i32, out out3: i32) {
-  %0 = comb.or %arg0, %arg1 : i32
-  %1 = comb.and %arg0, %arg1 : i32
-  %2 = comb.xor %arg0, %arg1 : i32
+  %0 = comb.or %arg0, %arg1, %arg2, %arg3 : i32
+  %1 = comb.and %arg0, %arg1, %arg2, %arg3 : i32
+  %2 = comb.xor %arg0, %arg1, %arg2, %arg3 : i32
   %3 = comb.mux %cond, %arg0, %arg1 : i32
 
   hw.output %0, %1, %2, %3 : i32, i32, i32, i32
diff --git a/lib/Conversion/CombToAIG/CombToAIG.cpp b/lib/Conversion/CombToAIG/CombToAIG.cpp
index 94ec4f85461e..19711a068ca2 100644
--- a/lib/Conversion/CombToAIG/CombToAIG.cpp
+++ b/lib/Conversion/CombToAIG/CombToAIG.cpp
@@ -91,6 +91,42 @@ struct CombXorOpConversion : OpConversionPattern<XorOp> {
   }
 };
 
+template <typename OpTy>
+struct CombLowerVariadicOp : OpConversionPattern<OpTy> {
+  using OpConversionPattern<OpTy>::OpConversionPattern;
+  using OpAdaptor = typename OpConversionPattern<OpTy>::OpAdaptor;
+  LogicalResult
+  matchAndRewrite(OpTy op, OpAdaptor adaptor,
+                  ConversionPatternRewriter &rewriter) const override {
+    auto result = lowerFullyAssociativeOp(op, op.getOperands(), rewriter);
+    rewriter.replaceOp(op, result);
+    return success();
+  }
+
+  static Value lowerFullyAssociativeOp(OpTy op, OperandRange operands,
+                                       ConversionPatternRewriter &rewriter) {
+    Value lhs, rhs;
+    switch (operands.size()) {
+    case 0:
+      assert(false && "cannot be called with empty operand range");
+      break;
+    case 1:
+      return operands[0];
+    case 2:
+      lhs = operands[0];
+      rhs = operands[1];
+      return rewriter.create<OpTy>(op.getLoc(), ValueRange{lhs, rhs}, true);
+    default:
+      auto firstHalf = operands.size() / 2;
+      lhs =
+          lowerFullyAssociativeOp(op, operands.take_front(firstHalf), rewriter);
+      rhs =
+          lowerFullyAssociativeOp(op, operands.drop_front(firstHalf), rewriter);
+      return rewriter.create<OpTy>(op.getLoc(), ValueRange{lhs, rhs}, true);
+    }
+  }
+};
+
 // Lower comb::MuxOp to AIG operations.
 struct CombMuxOpConversion : OpConversionPattern<MuxOp> {
   using OpConversionPattern<MuxOp>::OpConversionPattern;
@@ -150,7 +186,9 @@ static void populateCombToAIGConversionPatterns(RewritePatternSet &patterns) {
   patterns.add<
       // Bitwise Logical Ops
       CombAndOpConversion, CombOrOpConversion, CombXorOpConversion,
-      CombMuxOpConversion>(patterns.getContext());
+      CombMuxOpConversion,
+      // Variadic ops that must be lowered to binary operations
+      CombLowerVariadicOp<XorOp>>(patterns.getContext());
 }
 
 void ConvertCombToAIGPass::runOnOperation() {
diff --git a/test/Conversion/CombToAIG/comb-to-aig.mlir b/test/Conversion/CombToAIG/comb-to-aig.mlir
index 4628603cb1e7..d65c0c150b29 100644
--- a/test/Conversion/CombToAIG/comb-to-aig.mlir
+++ b/test/Conversion/CombToAIG/comb-to-aig.mlir
@@ -1,18 +1,27 @@
 // RUN: circt-opt %s --convert-comb-to-aig | FileCheck %s
 
 // CHECK-LABEL: @test
-hw.module @test(in %arg0: i32, in %arg1: i32, in %arg2: i32, in %arg3: i32, out out0: i32, out out1: i32, out out2: i32) {
+hw.module @test(in %arg0: i32, in %arg1: i32, in %arg2: i32, in %arg3: i32, out out0: i32, out out1: i32) {
   // CHECK-NEXT: %[[OR_TMP:.+]] = aig.and_inv not %arg0, not %arg1, not %arg2, not %arg3 : i32
   // CHECK-NEXT: %[[OR:.+]] = aig.and_inv not %0 : i32
   // CHECK-NEXT: %[[AND:.+]] = aig.and_inv %arg0, %arg1, %arg2, %arg3 : i32
-  // CHECK-NEXT: %[[XOR_NOT_AND:.+]] = aig.and_inv not %arg0, not %arg1 : i32
-  // CHECK-NEXT: %[[XOR_AND:.+]] = aig.and_inv %arg0, %arg1 : i32
-  // CHECK-NEXT: %[[XOR:.+]] = aig.and_inv not %[[XOR_NOT_AND]], not %[[XOR_AND]] : i32
-  // CHECK-NEXT: hw.output %[[OR]], %[[AND]], %[[XOR]] : i32, i32, i32
+  // CHECK-NEXT: hw.output %[[OR]], %[[AND]] : i32, i32
   %0 = comb.or %arg0, %arg1, %arg2, %arg3 : i32
   %1 = comb.and %arg0, %arg1, %arg2, %arg3 : i32
-  %2 = comb.xor %arg0, %arg1 : i32
-  hw.output %0, %1, %2 : i32, i32, i32
+  hw.output %0, %1 : i32, i32
+}
+
+// CHECK-LABEL: @xor
+hw.module @xor(in %arg0: i32, in %arg1: i32, in %arg2: i32, out out0: i32) {
+  // CHECK-NEXT: %[[RHS_NOT_AND:.+]] = aig.and_inv not %arg1, not %arg2 : i32
+  // CHECK-NEXT: %[[RHS_AND:.+]] = aig.and_inv %arg1, %arg2 : i32
+  // CHECK-NEXT: %[[RHS_XOR:.+]] = aig.and_inv not %[[RHS_NOT_AND]], not %[[RHS_AND]] : i32
+  // CHECK-NEXT: %[[NOT_AND:.+]] = aig.and_inv not %arg0, not %[[RHS_XOR]] : i32
+  // CHECK-NEXT: %[[AND:.+]] = aig.and_inv %arg0, %[[RHS_XOR]] : i32
+  // CHECK-NEXT: %[[RESULT:.+]] = aig.and_inv not %[[NOT_AND]], not %[[AND]] : i32
+  // CHECK-NEXT: hw.output %[[RESULT]]
+  %0 = comb.xor %arg0, %arg1, %arg2 : i32
+  hw.output %0 : i32
 }
 
 // CHECK-LABEL: @pass