diff --git a/paddle/fluid/operators/sum_op_npu.cc b/paddle/fluid/operators/sum_op_npu.cc
new file mode 100755
index 0000000000000000000000000000000000000000..0e811d51110c06f960ad10bd023835f62af19232
--- /dev/null
+++ b/paddle/fluid/operators/sum_op_npu.cc
@@ -0,0 +1,67 @@
+/* Copyright (c) 2021 PaddlePaddle Authors. All Rights Reserved.
+
+Licensed under the Apache License, Version 2.0 (the "License");
+you may not use this file except in compliance with the License.
+You may obtain a copy of the License at
+
+    http://www.apache.org/licenses/LICENSE-2.0
+
+Unless required by applicable law or agreed to in writing, software
+distributed under the License is distributed on an "AS IS" BASIS,
+WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+See the License for the specific language governing permissions and
+limitations under the License. */
+
+#include <memory>
+#include <string>
+#include <vector>
+
+#include "paddle/fluid/operators/sum_op.h"
+#include "paddle/fluid/operators/npu_op_runner.h"
+
+namespace paddle {
+namespace operators {
+
+using Tensor = framework::Tensor;
+
+template <typename DeviceContext, typename T>
+class SumNPUKernel : public framework::OpKernel<T> {
+ public:
+  void Compute(const framework::ExecutionContext& ctx) const override {
+
+    auto x = ctx.MultiInput<Tensor>("X");
+    auto* out = ctx.Output<Tensor>("Out");
+    out->mutable_data<T>(ctx.GetPlace());
+
+    auto place = ctx.GetPlace();
+
+    int n = static_cast<int>(x.size());
+    PADDLE_ENFORCE_EQ(n > 1, true,
+                      platform::errors::InvalidArgument(
+                          "The size of Input(x) list must larger or equal 2"));
+
+    auto stream =
+              ctx.template device_context<paddle::platform::NPUDeviceContext>()
+                      .stream();
+
+    auto runner = NpuOpRunner("Add", {*x[0], *x[1]}, {*out}, {});
+
+    runner.Run(stream);
+    for (int i = 2; i < n; i++) {
+        runner = NpuOpRunner("Add", {*out, *x[i]}, {*out}, {});
+        runner.Run(stream);
+    }
+  }
+};
+
+}  // namespace operators
+}  // namespace paddle
+
+namespace ops = paddle::operators;
+
+REGISTER_OP_NPU_KERNEL(
+    sum,
+    ops::SumNPUKernel<paddle::platform::NPUDeviceContext, float>,
+    ops::SumNPUKernel<paddle::platform::NPUDeviceContext,
+                       paddle::platform::float16>);
+
diff --git a/python/paddle/fluid/tests/unittests/npu/test_sum_op_npu.py b/python/paddle/fluid/tests/unittests/npu/test_sum_op_npu.py
new file mode 100755
index 0000000000000000000000000000000000000000..6cf840ac1d9346cd708f72de5a8a6d4c71b3ecff
--- /dev/null
+++ b/python/paddle/fluid/tests/unittests/npu/test_sum_op_npu.py
@@ -0,0 +1,85 @@
+#  Copyright (c) 2021 PaddlePaddle Authors. All Rights Reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#    http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+from __future__ import print_function
+
+import numpy as np
+import unittest
+import sys
+sys.path.append("..")
+from op_test import OpTest
+import paddle
+import paddle.fluid as fluid
+import paddle.fluid.core as core
+
+paddle.enable_static()
+SEED = 2021
+
+
+@unittest.skipIf(not paddle.is_compiled_with_npu(),
+                 "core is not compiled with NPU")
+class TestSum1(OpTest):
+    def setUp(self):
+        self.set_npu()
+        self.init_dtype()
+        self.op_type = "sum"
+        self.place = paddle.NPUPlace(0)
+
+        x0 = np.random.random((3, 40)).astype(self.dtype)
+        x1 = np.random.random((3, 40)).astype(self.dtype)
+        x2 = np.random.random((3, 40)).astype(self.dtype)
+        self.inputs = {'X': [("x0", x0), ("x1", x1), ("x2", x2)]}
+        y = x0 + x1 + x2
+        self.outputs = {'Out': y}
+
+        self.attrs = {'use_mkldnn': False}
+    
+    def init_dtype(self):
+        self.dtype = np.float32
+    
+    def set_npu(self):
+        self.__class__.use_npu = True
+
+    def test_check_output(self):
+        self.check_output_with_place(self.place, check_dygraph=False)
+
+
+class TestSum2(OpTest):
+    def setUp(self):
+        self.set_npu()
+        self.init_dtype()
+        self.op_type = "sum"
+        self.place = paddle.NPUPlace(0)
+
+        x0 = np.random.random((3, 3)).astype(self.dtype)
+        x1 = np.random.random((3, 3)).astype(self.dtype)
+        x2 = np.random.random((3, 3)).astype(self.dtype)
+        x3 = np.random.random((3, 3)).astype(self.dtype)
+        self.inputs = {'X': [("x0", x0), ("x1", x1), ("x2", x2), ("x3", x3)]}
+        y = x0 + x1 + x2 + x3
+        self.outputs = {'Out': y}
+
+        self.attrs = {'use_mkldnn': False}
+    
+    def init_dtype(self):
+        self.dtype = np.float16
+    
+    def set_npu(self):
+        self.__class__.use_npu = True
+
+    def test_check_output(self):
+        self.check_output_with_place(self.place, check_dygraph=False)
+
+if __name__ == '__main__':
+    unittest.main()