diff --git a/tensorflow/compiler/xla/service/collective_ops_utils.h b/tensorflow/compiler/xla/service/collective_ops_utils.h
index d7f63cb2663..1552397d4bc 100644
--- a/tensorflow/compiler/xla/service/collective_ops_utils.h
+++ b/tensorflow/compiler/xla/service/collective_ops_utils.h
@@ -208,11 +208,6 @@ template <typename I, typename O,
               std::enable_if_t<std::is_base_of<ParticipantData, I>::value>>
 class Rendezvous {
  public:
-  struct ParticipantImplOutput {
-    bool is_primary;
-    O custom_output;
-  };
-
   virtual ~Rendezvous() {}
   explicit Rendezvous(const RendezvousKey& k) : key_(k) {}
 
@@ -246,8 +241,7 @@ class Rendezvous {
 
  protected:
   // Returns domain-specific output O and whether this replica is primary.
-  virtual StatusOr<ParticipantImplOutput> RunCollectiveOp(
-      const I& participant) = 0;
+  virtual StatusOr<O> RunCollectiveOp(const I& participant) = 0;
 
   // Initialize the rendezvous by the first ("primary") thread which reaches the
   // barrier. Returns whether this thread is primary.
@@ -300,8 +294,8 @@ class Rendezvous {
           participant.device_ordinal, participant.stream, key_.ToString());
     });
 
-    TF_ASSIGN_OR_RETURN(ParticipantImplOutput p, RunCollectiveOp(participant));
-    return std::make_pair(p.custom_output, returned_blocking_counter_);
+    TF_ASSIGN_OR_RETURN(O output, RunCollectiveOp(participant));
+    return std::make_pair(std::move(output), returned_blocking_counter_);
   }
 
   const RendezvousKey key_;
diff --git a/tensorflow/compiler/xla/service/cpu/cpu_runtime.cc b/tensorflow/compiler/xla/service/cpu/cpu_runtime.cc
index d4d78f5ac12..437dbd5cbb3 100644
--- a/tensorflow/compiler/xla/service/cpu/cpu_runtime.cc
+++ b/tensorflow/compiler/xla/service/cpu/cpu_runtime.cc
@@ -323,7 +323,7 @@ class CpuAllToAllRendezvous
       : xla::Rendezvous<AllToAllParticipantData, std::nullptr_t>(k) {}
 
  protected:
-  xla::StatusOr<ParticipantImplOutput> RunCollectiveOp(
+  xla::StatusOr<std::nullptr_t> RunCollectiveOp(
       const AllToAllParticipantData& /*participant*/) override {
     bool is_primary = InitializationBarrier();
 
@@ -373,7 +373,7 @@ class CpuAllToAllRendezvous
         }
       }
     }
-    return ParticipantImplOutput{is_primary, nullptr};
+    return nullptr;
   }
 };
 
@@ -384,7 +384,7 @@ class CpuCollectivePermuteRendezvous
       : xla::Rendezvous<CollectivePermuteParticipantData, std::nullptr_t>(k) {}
 
  protected:
-  xla::StatusOr<ParticipantImplOutput> RunCollectiveOp(
+  xla::StatusOr<std::nullptr_t> RunCollectiveOp(
       const CollectivePermuteParticipantData& /*participant*/) override {
     bool primary = InitializationBarrier();
 
@@ -415,7 +415,7 @@ class CpuCollectivePermuteRendezvous
         std::memset(p.destination_data.opaque(), 0, p.byte_size);
       }
     }
-    return ParticipantImplOutput{primary, /*custom_output=*/nullptr};
+    return nullptr;
   }
 };
 
@@ -426,7 +426,7 @@ class CpuAllReduceRendezvous
       : xla::Rendezvous<xla::AllReduceParticipantData, std::nullptr_t>(k) {}
 
  protected:
-  xla::StatusOr<ParticipantImplOutput> RunCollectiveOp(
+  xla::StatusOr<std::nullptr_t> RunCollectiveOp(
       const xla::AllReduceParticipantData& participant) override {
     xla::PrimitiveType datatype = participant.buffers.front().primitive_type;
     bool primary = InitializationBarrier();
@@ -465,7 +465,7 @@ class CpuAllReduceRendezvous
           LOG(FATAL) << "Unexpected datatype;";
       }
     }
-    return ParticipantImplOutput{primary, /*custom_output=*/nullptr};
+    return nullptr;
   }
 
  private:
diff --git a/tensorflow/compiler/xla/service/gpu/nccl_utils.cc b/tensorflow/compiler/xla/service/gpu/nccl_utils.cc
index 81f54ba08b6..6b6a9351384 100644
--- a/tensorflow/compiler/xla/service/gpu/nccl_utils.cc
+++ b/tensorflow/compiler/xla/service/gpu/nccl_utils.cc
@@ -223,7 +223,7 @@ class NcclCliqueRendezvous
         local_participants_(local_participants),
         callback_(callback) {}
 
-  StatusOr<ParticipantImplOutput> RunCollectiveOp(
+  StatusOr<LockedNcclClique> RunCollectiveOp(
       const NcclCliqueParticipantData&) override {
     tensorflow::mutex_lock lock(mu_);
     bool primary = !initialized_;
@@ -238,7 +238,7 @@ class NcclCliqueRendezvous
     if (primary) {
       lock_ = std::make_shared<absl::MutexLock>(clique->mu());
     }
-    return ParticipantImplOutput{primary, LockedNcclClique{clique, lock_}};
+    return LockedNcclClique{clique, lock_};
   }
 
  private: