diff --git a/src/plugins/intel_gpu/src/graph/graph_optimizer/reorder_inputs.cpp b/src/plugins/intel_gpu/src/graph/graph_optimizer/reorder_inputs.cpp
index cbcdf14a9b3..5fcc1fb5748 100644
--- a/src/plugins/intel_gpu/src/graph/graph_optimizer/reorder_inputs.cpp
+++ b/src/plugins/intel_gpu/src/graph/graph_optimizer/reorder_inputs.cpp
@@ -872,14 +872,13 @@ void reorder_inputs::run(program& p, layout_optimizer& lo, reorder_factory& rf)
         auto& weights = fc_node.weights();
         auto& input = fc_node.input();
         auto input_layout = input.get_output_layout();
-
         // Change input data type of fully-connected node from i32 to f32
         if (input_layout.data_type == data_types::i32) {
             auto new_layout = input_layout;
             new_layout.data_type = data_types::f32;
             auto new_input = rf.get_reorder(input.id(),
-                                            static_cast<int32_t>(input.get_user_index(fc_node)),
-                                            static_cast<int32_t>(input.get_output_layouts().size()),
+                                            fc_node.get_port_from_deps(input.id()),
+                                            static_cast<int32_t>(input.get_users().size()),
                                             input_layout, new_layout);
             if (new_input.first) {
                p.add_intermediate(new_input.first, fc_node, 0);
diff --git a/src/plugins/intel_gpu/src/graph/include/layout_optimizer.h b/src/plugins/intel_gpu/src/graph/include/layout_optimizer.h
index 5ac9ca99d2e..5bc1a59226e 100644
--- a/src/plugins/intel_gpu/src/graph/include/layout_optimizer.h
+++ b/src/plugins/intel_gpu/src/graph/include/layout_optimizer.h
@@ -43,7 +43,7 @@ public:
     // (no need to add it to 'ouputs' etc.) for pair.first == nullptr, pair.second == true
     std::pair<std::shared_ptr<reorder>, bool> get_reorder(primitive_id src_id,
                                                           int32_t src_port,
-                                                          int32_t src_output_size,
+                                                          size_t src_num_of_users,
                                                           const layout& in_layout,
                                                           const layout& out_layout);
 
diff --git a/src/plugins/intel_gpu/src/graph/layout_optimizer.cpp b/src/plugins/intel_gpu/src/graph/layout_optimizer.cpp
index 89744006f33..f89d1656d9f 100644
--- a/src/plugins/intel_gpu/src/graph/layout_optimizer.cpp
+++ b/src/plugins/intel_gpu/src/graph/layout_optimizer.cpp
@@ -144,7 +144,7 @@ bool layout_optimizer::onednn_check_data_types_for_fc_gemm(data_types in_dt, dat
 
 std::pair<std::shared_ptr<reorder>, bool> reorder_factory::get_reorder(primitive_id src_id,
                                                                        int32_t src_port,
-                                                                       int32_t src_output_size,
+                                                                       size_t src_num_of_users,
                                                                        const layout& in_layout,
                                                                        const layout& out_layout) {
     if (in_layout == out_layout)
@@ -152,21 +152,17 @@ std::pair<std::shared_ptr<reorder>, bool> reorder_factory::get_reorder(primitive
 
     cache_key ckey{ src_id + "." + std::to_string(src_port), out_layout };
     auto itr = _cached_reorders.find(ckey);
-    if (itr != _cached_reorders.end())
+    // Finding existing reorder is used by with the key which was made with src_id, src_port.
+    // But that doens't cover the case where the node has single output but multiple users.
+    // Because in that case, port number to each user is same, and thereby reorder is created for only one user.
+    // For this case to be covered, the condition 'src_num_of_users <= 1' is needed.
+    if (itr != _cached_reorders.end() && src_num_of_users <= 1)
         return std::make_pair(itr->second, true);
 
     auto count = _cached_reorders.size();
     std::stringstream ss;
     ss << src_id << "_" << std::to_string(src_port) << "_reorder_" << count;
 
-    // src_port is used to name the reorder in the code just above,
-    // whereas used to index the output in the code below.
-    if (src_port >= src_output_size && src_output_size > 0) {
-        // Cover the case where the node has single output but multiple users,
-        // while port number to each user is same.
-        src_port = src_output_size - 1;
-    }
-
     auto reorder = std::make_shared<cldnn::reorder>(ss.str(), input_info{src_id, src_port}, out_layout);
     _cached_reorders[ckey] = reorder;