diff --git a/mlir/include/mlir/Dialect/LLVMIR/LLVMOps.td b/mlir/include/mlir/Dialect/LLVMIR/LLVMOps.td
--- a/mlir/include/mlir/Dialect/LLVMIR/LLVMOps.td
+++ b/mlir/include/mlir/Dialect/LLVMIR/LLVMOps.td
@@ -764,6 +764,74 @@
 def LLVM_experimental_vector_reduce_v2_fadd : LLVM_VectorReductionV2<"fadd">;
 def LLVM_experimental_vector_reduce_v2_fmul : LLVM_VectorReductionV2<"fmul">;
 
+//
+// LLVM Matrix operations.
+//
+
+// TODO(ntv, zinenko): Provide more idiomatic tblgen support.
+// OTOH there are 4 intrinsics total atm so..
+// TODO(ntv): Use this once LLVM integrate landed.
+// def LLVM_MatrixMultiplyOp
+//     : LLVM_OneResultOp<"intr.matrix.multiply">,
+//       Arguments<(
+//         ins LLVM_Type:$A, LLVM_Type:$B, I32Attr:$M, I32Attr:$N)> {
+//   string llvmBuilder =
+//     "$res = MatrixBuilder::CreateMatrixMultiply("
+//     "$A, $B, $M.getZExtValue(), $N.getZExtValue()));";
+//   let builders = [
+//     OpBuilder<"Builder *builder, OperationState &result, Value A, Value B, "
+//     "IntegerAttr M, IntegerAttr N", [{
+//       result.addAttribute("M", M);
+//       result.addAttribute("N", N);
+//       result.addTypes(VectorType::get(
+//         A.getType().cast<VectorType>()->getElementType(),
+//         M.getZExtValue() * N.getZExtValue()));}]>,
+//   ];
+//   let assemblyFormat =
+//     "$A `,` $B attr-dict `:` type($A) `,` type($B) `->` type($res)";
+// }
+
+def LLVM_MatrixMultiplyOp
+    : LLVM_OneResultOp<"intr.matrix.multiply">,
+      Arguments<(
+        ins LLVM_Type:$A, LLVM_Type:$B, I32Attr:$M, I32Attr:$N, I32Attr:$K)> {
+  string llvmBuilder = [{
+    $res = builder.CreateCall(
+      llvm::Intrinsic::getDeclaration(
+        builder.GetInsertBlock()->getModule(),
+        llvm::Intrinsic::matrix_multiply,
+        {
+          llvm::VectorType::get(
+            $A->getType()->getScalarType(),
+            $M.getZExtValue() * $N.getZExtValue()),
+          $A->getType(),
+          $B->getType()
+        }
+      ),
+      {
+        $A,
+        $B,
+        llvm::ConstantInt::get(builder.getInt32Ty(), $M.getZExtValue()),
+        // LLVM call is M, K, N .. go figure
+        llvm::ConstantInt::get(builder.getInt32Ty(), $K.getZExtValue()),
+        llvm::ConstantInt::get(builder.getInt32Ty(), $N.getZExtValue())
+      }
+    );
+  }];
+  let builders = [
+    OpBuilder<"Builder *builder, OperationState &result, LLVMType resType, "
+              "Value A, Value B, IntegerAttr M, IntegerAttr N, IntegerAttr K",
+    [{
+      result.addAttribute("M", M);
+      result.addAttribute("N", N);
+      result.addAttribute("K", K);
+      result.addTypes(resType);}]>,
+  ];
+
+  let assemblyFormat =
+    "$A `,` $B attr-dict `:` type($A) `,` type($B) `->` type($res)";
+}
+
 //
 // Atomic operations.
 //
diff --git a/mlir/include/mlir/Target/LLVMIR/ModuleTranslation.h b/mlir/include/mlir/Target/LLVMIR/ModuleTranslation.h
--- a/mlir/include/mlir/Target/LLVMIR/ModuleTranslation.h
+++ b/mlir/include/mlir/Target/LLVMIR/ModuleTranslation.h
@@ -23,6 +23,8 @@
 #include "llvm/IR/Function.h"
 #include "llvm/IR/IRBuilder.h"
 #include "llvm/IR/Value.h"
+// TODO(ntv): enable
+// #include "llvm/IR/MatrixBuilder.h"
 
 namespace mlir {
 class Attribute;
diff --git a/mlir/test/Target/llvmir-intrinsics.mlir b/mlir/test/Target/llvmir-intrinsics.mlir
--- a/mlir/test/Target/llvmir-intrinsics.mlir
+++ b/mlir/test/Target/llvmir-intrinsics.mlir
@@ -130,6 +130,18 @@
   llvm.return
 }
 
+// CHECK-LABEL: @matrix_intrinsics
+//                                       4x16                       16x3
+llvm.func @matrix_intrinsics(%A: !llvm<"<64 x float>">, %B: !llvm<"<48 x float>">)
+//           4x3
+  -> !llvm<"<12 x float>">
+{
+  // CHECK: call <12 x float> @llvm.matrix.multiply.v12f32.v64f32.v48f32(<64 x float> %0, <48 x float> %1, i32 4, i32 16, i32 3)
+  %C = llvm.intr.matrix.multiply %A, %B { M = 4: i32, N = 3: i32, K = 16: i32 } :
+    !llvm<"<64 x float>">, !llvm<"<48 x float>"> -> !llvm<"<12 x float>">
+  llvm.return %C: !llvm<"<12 x float>">
+}
+
 // Check that intrinsics are declared with appropriate types.
 // CHECK-DAG: declare float @llvm.fma.f32(float, float, float)
 // CHECK-DAG: declare <8 x float> @llvm.fma.v8f32(<8 x float>, <8 x float>, <8 x float>) #0
@@ -153,3 +165,4 @@
 // CHECK-DAG: declare float @llvm.cos.f32(float)
 // CHECK-DAG: declare <8 x float> @llvm.cos.v8f32(<8 x float>) #0
 // CHECK-DAG: declare float @llvm.copysign.f32(float, float)
+// CHECK-DAG: declare <12 x float> @llvm.matrix.multiply.v12f32.v64f32.v48f32(<64 x float>, <48 x float>, i32 immarg, i32 immarg, i32 immarg)