diff --git a/src/plugins/intel_gpu/src/graph/network.cpp b/src/plugins/intel_gpu/src/graph/network.cpp
index c66ceb891a4..6a18faaa978 100644
--- a/src/plugins/intel_gpu/src/graph/network.cpp
+++ b/src/plugins/intel_gpu/src/graph/network.cpp
@@ -422,8 +422,13 @@ void network::set_arguments() {
                 // In that case some_op is static and we may want to set arguments once,
                 // but dynamic optimized out reshape means that output buffer of reshape is unavailable
                 // and attempt to set args will fail.
+
+                // (dynamic) -> static optimizable reshape -> static optimizable reshape -> some_op
+                // In that case, it is a limit about second reshape.
                 auto prim = dep.first->get_impl_params()->desc;
-                if (dep.first->can_be_optimized() && (dep.first->is_dynamic() || prim->type == read_value::type_id()))
+                if (dep.first->can_be_optimized() && (dep.first->is_dynamic() ||
+                                                      dep.first->output_memory_ptr() == nullptr ||
+                                                      prim->type == read_value::type_id()))
                     can_set_args = false;
             }
 
diff --git a/src/plugins/intel_gpu/src/graph/reshape.cpp b/src/plugins/intel_gpu/src/graph/reshape.cpp
index bf7c9657bd2..ccfba9e8cb0 100644
--- a/src/plugins/intel_gpu/src/graph/reshape.cpp
+++ b/src/plugins/intel_gpu/src/graph/reshape.cpp
@@ -207,6 +207,9 @@ void reshape_inst::update_output_memory() {
         return;
 
     build_deps();  // reshape need deps
+    if (node->get_program().get_config().get_property(ov::intel_gpu::allow_new_shape_infer) &&
+        input_memory_ptr() == nullptr)
+        return;
     OPENVINO_ASSERT(input_memory_ptr() != nullptr, "[GPU] Failed to reuse input in ", id(), " primitive: input memory was not allocated");
     _outputs = {_network.get_engine().reinterpret_buffer(input_memory(), _impl_params->get_output_layout())};
 }
diff --git a/src/plugins/intel_gpu/tests/unit/test_cases/reshape_gpu_test.cpp b/src/plugins/intel_gpu/tests/unit/test_cases/reshape_gpu_test.cpp
index eeef9ac9d17..fcf44a3c869 100644
--- a/src/plugins/intel_gpu/tests/unit/test_cases/reshape_gpu_test.cpp
+++ b/src/plugins/intel_gpu/tests/unit/test_cases/reshape_gpu_test.cpp
@@ -1043,6 +1043,56 @@ TEST(reshape_gpu_f32, basic_dynamic_shape_to_static_optimized_out) {
     }
 }
 
+TEST(reshape_gpu_f32, basic_dynamic_shape_to_static_optimized_out_static_optimized_out) {
+    auto& engine = get_test_engine();
+
+    auto input = engine.allocate_memory(layout{ov::PartialShape{1, 2, 10}, data_types::f32, format::bfyx});
+    topology topology;
+    topology.add(input_layout("input", layout{ov::PartialShape::dynamic(3), data_types::f32, format::bfyx}));
+    topology.add(reshape("reshape_1", input_info("input"), false, {1, 2, 10}, {1, 2, 10}));
+    topology.add(reduce("reduce_1", input_info("reshape_1"), reduce_mode::max, {1}, true));
+    topology.add(reshape("reshape", input_info("reshape_1"), false, {2, 10}, {2, 10}));
+    topology.add(reduce("reduce", input_info("reshape"), reduce_mode::max, {1}, true));
+
+    // clang-format off
+    std::vector<float> input_data = {
+        0.0, 1.f, 2.f, 3.f, 4.f, 5.f, 6.f, 7.f, 8.f, 9.f,
+        0.0, 1.f, 2.f, 3.f, 4.f, 5.f, 6.f, 7.f, 8.f, 9.f,
+    };
+    // clang-format on
+
+    set_values(input, input_data);
+
+    ExecutionConfig config = get_test_default_config(engine);
+    config.set_property(ov::intel_gpu::allow_new_shape_infer(true));
+    config.set_property(ov::intel_gpu::optimize_data(true));
+    network network(engine, topology, config);
+    network.set_input_data("input", input);
+    auto outputs = network.execute();
+
+    ASSERT_TRUE(network.get_primitive("reshape")->can_be_optimized());
+
+    ASSERT_EQ(outputs.size(), size_t(2));
+    ASSERT_EQ(outputs.begin()->first, "reduce");
+
+    auto output = outputs.at("reduce").get_memory();
+
+    ASSERT_EQ(output->get_layout().data_type, input->get_layout().data_type);
+    ASSERT_EQ(output->get_layout().format, format::bfyx);
+    ASSERT_TRUE(output->get_layout().is_static());
+    ov::PartialShape expected_shape = {2, 1};
+    ASSERT_EQ(output->get_layout().get_partial_shape(), expected_shape);
+
+    cldnn::mem_lock<float> output_ptr(output, get_test_stream());
+    std::vector<float> expected_res = {9.f, 9.f};
+    ASSERT_EQ(output_ptr.size(), expected_res.size());
+
+
+    for (size_t i = 0; i < expected_res.size(); i++) {
+        ASSERT_EQ(expected_res[i], output_ptr[i]);
+    }
+}
+
 TEST(reshape_gpu_f32, basic_runtime_dynamic_shape_activation_fusion) {
     auto& engine = get_test_engine();