diff --git a/tools/run_tests/python_utils/upload_test_results.py b/tools/run_tests/python_utils/upload_test_results.py
index 09dcd57ad4c..63fa38b6789 100644
--- a/tools/run_tests/python_utils/upload_test_results.py
+++ b/tools/run_tests/python_utils/upload_test_results.py
@@ -163,6 +163,7 @@ def upload_interop_results_to_bq(resultset, bq_table, args):
         expiration_ms=_EXPIRATION_MS)
 
     for shortname, results in six.iteritems(resultset):
+        bq_rows = []
         for result in results:
             test_results = {}
             _get_build_metadata(test_results)
@@ -175,11 +176,15 @@ def upload_interop_results_to_bq(resultset, bq_table, args):
             test_results['test_case'] = shortname.split(':')[3]
             test_results['timestamp'] = time.strftime('%Y-%m-%d %H:%M:%S')
             row = big_query_utils.make_row(str(uuid.uuid4()), test_results)
-            # TODO(jtattermusch): rows are inserted one by one, very inefficient
+            bq_rows.append(row)
+
+        # BigQuery sometimes fails with large uploads, so batch 1,000 rows at a time.
+        for i in range((len(bq_rows) / 1000) + 1):
             max_retries = 3
             for attempt in range(max_retries):
-                if big_query_utils.insert_rows(bq, _PROJECT_ID, _DATASET_ID,
-                                               bq_table, [row]):
+                if big_query_utils.insert_rows(
+                        bq, _PROJECT_ID, _DATASET_ID, bq_table,
+                        bq_rows[i * 1000:(i + 1) * 1000]):
                     break
                 else:
                     if attempt < max_retries - 1: