diff --git a/src/plugins/intel_gpu/src/graph/primitive_inst.cpp b/src/plugins/intel_gpu/src/graph/primitive_inst.cpp
index c92c5854a8199e..1ad11f77adc7a0 100644
--- a/src/plugins/intel_gpu/src/graph/primitive_inst.cpp
+++ b/src/plugins/intel_gpu/src/graph/primitive_inst.cpp
@@ -392,7 +392,7 @@ void primitive_inst::update_shape() {
         if (i >= _deps.size())
             continue;
 
-        if (_deps[i].first->get_node().is_in_shape_of_subgraph()) {
+        if (_deps[i].first->get_node().is_in_shape_of_subgraph() && _deps[i].first->get_node().get_selected_impl()->is_cpu()) {
             bool can_skip = true;
             const auto& insts = _deps[i].first->dependant_shape_of_insts;
             for (auto& inst : insts) {
@@ -432,7 +432,7 @@ void primitive_inst::update_shape() {
             continue;
         }
 
-        if (!get_node().is_type<shape_of>() && !dep->get_node().is_in_shape_of_subgraph()) {
+        if (!get_node().is_type<shape_of>() && !dep->get_node().get_selected_impl()->is_cpu()) {
             has_runtime_deps = true;
 
             // Events may be not created for in-order queue, so take them for OOO queue only