diff --git a/flang/include/flang/Optimizer/Builder/IntrinsicCall.h b/flang/include/flang/Optimizer/Builder/IntrinsicCall.h
index 1f7da10fdcc20..0e3c9aa22f994 100644
--- a/flang/include/flang/Optimizer/Builder/IntrinsicCall.h
+++ b/flang/include/flang/Optimizer/Builder/IntrinsicCall.h
@@ -274,6 +274,7 @@ struct IntrinsicLibrary {
                                       llvm::ArrayRef<fir::ExtendedValue>);
   template <Extremum, ExtremumBehavior>
   mlir::Value genExtremum(mlir::Type, llvm::ArrayRef<mlir::Value>);
+  void genFenceProxyAsync(llvm::ArrayRef<fir::ExtendedValue>);
   mlir::Value genFloor(mlir::Type, llvm::ArrayRef<mlir::Value>);
   mlir::Value genFraction(mlir::Type resultType,
                           mlir::ArrayRef<mlir::Value> args);
diff --git a/flang/lib/Optimizer/Builder/IntrinsicCall.cpp b/flang/lib/Optimizer/Builder/IntrinsicCall.cpp
index e5b70eed9926b..41c117800f231 100644
--- a/flang/lib/Optimizer/Builder/IntrinsicCall.cpp
+++ b/flang/lib/Optimizer/Builder/IntrinsicCall.cpp
@@ -502,6 +502,10 @@ static constexpr IntrinsicHandler handlers[]{
      &I::genExtendsTypeOf,
      {{{"a", asBox}, {"mold", asBox}}},
      /*isElemental=*/false},
+    {"fence_proxy_async",
+     &I::genFenceProxyAsync,
+     {},
+     /*isElemental=*/false},
     {"findloc",
      &I::genFindloc,
      {{{"array", asBox},
@@ -4362,6 +4366,17 @@ IntrinsicLibrary::genExtendsTypeOf(mlir::Type resultType,
                                      fir::getBase(args[1])));
 }
 
+// FENCE_PROXY_ASYNC (CUDA)
+void IntrinsicLibrary::genFenceProxyAsync(
+    llvm::ArrayRef<fir::ExtendedValue> args) {
+  assert(args.size() == 0);
+  auto kind = mlir::NVVM::ProxyKindAttr::get(
+      builder.getContext(), mlir::NVVM::ProxyKind::async_shared);
+  auto space = mlir::NVVM::SharedSpaceAttr::get(
+      builder.getContext(), mlir::NVVM::SharedSpace::shared_cta);
+  mlir::NVVM::FenceProxyOp::create(builder, loc, kind, space);
+}
+
 // FINDLOC
 fir::ExtendedValue
 IntrinsicLibrary::genFindloc(mlir::Type resultType,
diff --git a/flang/module/cudadevice.f90 b/flang/module/cudadevice.f90
index afb39ebdf0d07..106f3e20aeaee 100644
--- a/flang/module/cudadevice.f90
+++ b/flang/module/cudadevice.f90
@@ -2008,6 +2008,11 @@ attributes(device) function barrier_arrive_cnt(barrier, count) result(token)
     end function
   end interface
 
+  interface
+    attributes(device) subroutine fence_proxy_async()
+    end subroutine
+  end interface
+
   interface
     attributes(device) subroutine tma_bulk_commit_group()
     end subroutine
diff --git a/flang/test/Lower/CUDA/cuda-device-proc.cuf b/flang/test/Lower/CUDA/cuda-device-proc.cuf
index 487a6b2883f46..50c0938a09cab 100644
--- a/flang/test/Lower/CUDA/cuda-device-proc.cuf
+++ b/flang/test/Lower/CUDA/cuda-device-proc.cuf
@@ -420,6 +420,14 @@ end subroutine
 ! CHECK: %[[SHARED_PTR:.*]] = llvm.addrspacecast %[[LLVM_PTR]] : !llvm.ptr to !llvm.ptr<3>
 ! CHECK: nvvm.mbarrier.arrive.expect_tx %[[SHARED_PTR]], %{{.*}} : !llvm.ptr<3>, i32
 
+
+attributes(global) subroutine test_fence()
+  call fence_proxy_async()
+end subroutine
+
+! CHECK-LABEL: func.func @_QPtest_fence()
+! CHECK: nvvm.fence.proxy {kind = #nvvm.proxy_kind<async.shared>, space = #nvvm.shared_space<cta>}
+
 attributes(global) subroutine test_tma()
   call tma_bulk_commit_group()
   call tma_bulk_wait_group()