@@ -1,5 +1,6 @@
 import os
 import time
+import zipfile
 from platform import platform
 from typing import TYPE_CHECKING, Callable, Optional
 
@@ -23,6 +24,7 @@
 from cycode.cli.files_collector.sca.sca_file_collector import add_sca_dependencies_tree_documents_if_needed
 from cycode.cli.files_collector.zip_documents import zip_documents
 from cycode.cli.models import CliError, Document, LocalScanResult
+from cycode.cli.utils.host_info import is_64bit
 from cycode.cli.utils.path_utils import get_absolute_path, get_path_by_os
 from cycode.cli.utils.progress_bar import ScanProgressBarSection
 from cycode.cli.utils.scan_batch import run_parallel_batched_scan
@@ -145,6 +147,7 @@ def _get_scan_documents_thread_func(
     is_git_diff: bool,
     is_commit_range: bool,
     scan_parameters: dict,
+    prezipped: Optional['InMemoryZip'] = None,
 ) -> Callable[[list[Document]], tuple[str, CliError, LocalScanResult]]:
     cycode_client = ctx.obj['client']
     scan_type = ctx.obj['scan_type']
@@ -164,9 +167,14 @@ def _scan_batch_thread_func(batch: list[Document]) -> tuple[str, CliError, Local
 
         should_use_sync_flow = _should_use_sync_flow(command_scan_type, scan_type, sync_option)
 
+        # the single ZIP flow already built the archive to check that it fits; don't build it twice
+        zipped_documents = prezipped
+
         try:
-            logger.debug('Preparing local files, %s', {'batch_files_count': len(batch)})
-            zipped_documents = zip_documents(scan_type, batch)
+            if zipped_documents is None:
+                logger.debug('Preparing local files, %s', {'batch_files_count': len(batch)})
+                zipped_documents = zip_documents(scan_type, batch)
+
             zip_file_size = zipped_documents.size
             scan_result = _perform_scan(
                 cycode_client,
@@ -189,6 +197,9 @@ def _scan_batch_thread_func(batch: list[Document]) -> tuple[str, CliError, Local
         except Exception as e:
             error = handle_scan_exception(ctx, e, return_exception=True)
             error_message = str(e)
+        finally:
+            if zipped_documents is not None:
+                zipped_documents.cleanup()
 
         if local_scan_result:
             detections_count = local_scan_result.detections_count
@@ -225,34 +236,77 @@ def _scan_batch_thread_func(batch: list[Document]) -> tuple[str, CliError, Local
     return _scan_batch_thread_func
 
 
+def _log_selected_upload_mode(mode: str, reason: str, documents_count: int) -> None:
+    logger.debug(
+        'Selected upload mode, %s',
+        {
+            'mode': mode,
+            'reason': reason,
+            'documents_count': documents_count,
+            'max_files_count': consts.ZIP_MAX_FILES_COUNT,
+            'zip64_enabled': is_64bit(),
+        },
+    )
+
+
+def _exceeds_non_zip64_files_count(documents_to_scan: list[Document]) -> bool:
+    """Whether a single ZIP can't hold all the documents because ZIP64 is unavailable.
+
+    Without ZIP64 (32-bit interpreter) the archive is capped at 65,535 entries.
+    """
+    return not is_64bit() and len(documents_to_scan) > consts.ZIP_MAX_FILES_COUNT
+
+
 def _run_presigned_upload_scan(
-    scan_batch_thread_func: Callable,
-    scan_type: str,
+    ctx: typer.Context,
+    is_git_diff: bool,
+    is_commit_range: bool,
+    scan_parameters: dict,
     documents_to_scan: list[Document],
     progress_bar: 'BaseProgressBar',
     printer: 'ConsolePrinter',
 ) -> tuple:
-    try:
-        # Try to zip all documents as a single batch; ZipTooLargeError raised if it exceeds the scan type's limit
-        zip_documents(scan_type, documents_to_scan)
-        # It fits: skip batching and upload everything as one ZIP
+    scan_type = ctx.obj['scan_type']
+    documents_count = len(documents_to_scan)
+
+    def run_batched() -> tuple:
         return run_parallel_batched_scan(
-            scan_batch_thread_func,
+            _get_scan_documents_thread_func(ctx, is_git_diff, is_commit_range, scan_parameters),
             scan_type,
             documents_to_scan,
             progress_bar=progress_bar,
-            skip_batching=True,
         )
-    except custom_exceptions.ZipTooLargeError:
+
+    if _exceeds_non_zip64_files_count(documents_to_scan):
+        # Don't waste time zipping documents we already know won't fit into a single ZIP
+        _log_selected_upload_mode('batched', 'files_count_exceeds_non_zip64_limit', documents_count)
+        return run_batched()
+
+    zipped_documents = None
+    try:
+        # Try to zip all documents as a single batch; ZipTooLargeError raised if it exceeds the scan type's limit
+        zipped_documents = zip_documents(scan_type, documents_to_scan)
+    except (custom_exceptions.ZipTooLargeError, zipfile.LargeZipFile):
+        # LargeZipFile is a safety net: the files count pre-check above should have caught it already
+        _log_selected_upload_mode('batched', 'zip_too_large', documents_count)
+        if zipped_documents is not None:
+            zipped_documents.cleanup()
+
         printer.print_warning(
             'The scan is too large to upload as a single file. This may result in corrupted scan results.'
         )
-        return run_parallel_batched_scan(
-            scan_batch_thread_func,
-            scan_type,
-            documents_to_scan,
-            progress_bar=progress_bar,
-        )
+        return run_batched()
+
+    # It fits: skip batching and upload everything as one ZIP. The archive we just built is the one
+    # that gets uploaded, so the scan doesn't pay for compressing every document twice
+    _log_selected_upload_mode('single_zip', 'fits_single_zip', documents_count)
+    return run_parallel_batched_scan(
+        _get_scan_documents_thread_func(ctx, is_git_diff, is_commit_range, scan_parameters, zipped_documents),
+        scan_type,
+        documents_to_scan,
+        progress_bar=progress_bar,
+        skip_batching=True,
+    )
 
 
 def scan_documents(
@@ -277,18 +331,19 @@ def scan_documents(
         )
         return
 
-    scan_batch_thread_func = _get_scan_documents_thread_func(ctx, is_git_diff, is_commit_range, scan_parameters)
-
     # Presigned single-file upload is async-only; a --sync scan must stay on the batched inline path
     # so it never builds one oversized zip to POST synchronously.
     should_use_sync_flow = _should_use_sync_flow(ctx.info_name, scan_type, ctx.obj['sync'])
     if should_use_presigned_upload(scan_type) and not should_use_sync_flow:
         errors, local_scan_results = _run_presigned_upload_scan(
-            scan_batch_thread_func, scan_type, documents_to_scan, progress_bar, printer
+            ctx, is_git_diff, is_commit_range, scan_parameters, documents_to_scan, progress_bar, printer
         )
     else:
         errors, local_scan_results = run_parallel_batched_scan(
-            scan_batch_thread_func, scan_type, documents_to_scan, progress_bar=progress_bar
+            _get_scan_documents_thread_func(ctx, is_git_diff, is_commit_range, scan_parameters),
+            scan_type,
+            documents_to_scan,
+            progress_bar=progress_bar,
         )
 
     try_set_aggregation_report_url_if_needed(ctx, scan_parameters, ctx.obj['client'], scan_type)
 
@@ -214,6 +214,9 @@ def _scan_commit_range_documents(
 
     zip_file_size = from_commit_zipped_documents.size + to_commit_zipped_documents.size
 
+    from_commit_zipped_documents.cleanup()
+    to_commit_zipped_documents.cleanup()
+
     detections_count = relevant_detections_count = 0
     if local_scan_result:
         detections_count = local_scan_result.detections_count
 
@@ -227,6 +227,12 @@
 PRESIGNED_LINK_UPLOADED_ZIP_MAX_SIZE_LIMIT_IN_BYTES = 5 * 1024 * 1024 * 1024  # 5 GB (S3 presigned POST limit)
 PRESIGNED_UPLOAD_SCAN_TYPES = {SAST_SCAN_TYPE, SECRET_SCAN_TYPE}
 
+# the non-ZIP64 central directory stores the entry count in 16 bits; ZIP64 (64-bit interpreters) lifts it
+ZIP_MAX_FILES_COUNT = 65_535
+
+# the ZIP is built in memory up to this size, and spilled to a temp file beyond it
+ZIP_SPOOL_MAX_SIZE_IN_BYTES = 64 * 1024 * 1024
+
 DEFAULT_ZIP_MAX_SIZE_LIMIT_IN_BYTES = 20 * 1024 * 1024
 ZIP_MAX_SIZE_LIMIT_IN_BYTES = {
     SCA_SCAN_TYPE: 200 * 1024 * 1024,
 
@@ -1,3 +1,4 @@
+import zipfile
 from typing import Optional
 
 import typer
@@ -26,6 +27,13 @@ def handle_scan_exception(ctx: typer.Context, err: Exception, *, return_exceptio
             'Please try ignoring irrelevant paths using the `cycode ignore --by-path` command '
             'and execute the scan again',
         ),
+        zipfile.LargeZipFile: CliError(
+            soft_fail=True,
+            code='zip_too_large_error',
+            message='The path you attempted to scan contains too many files to pack into a single archive. '
+            'Scanning such paths requires a 64-bit Python interpreter. '
+            'Please try ignoring irrelevant paths using a .cycodeignore file and execute the scan again',
+        ),
         custom_exceptions.FileCollectionError: CliError(
             soft_fail=False,
             code='file_collection_error',
 
@@ -1,20 +1,48 @@
+import shutil
+import tempfile
 from collections import defaultdict
-from io import BytesIO
+from os import SEEK_END
 from pathlib import Path
-from sys import getsizeof
-from typing import Optional
+from typing import IO, Optional
 from zipfile import ZIP_DEFLATED, ZipFile
 
+from cycode.cli import consts
 from cycode.cli.user_settings.configuration_manager import ConfigurationManager
+from cycode.cli.utils.host_info import is_64bit
 from cycode.cli.utils.path_utils import concat_unique_id
+from cycode.logger import get_logger
+
+logger = get_logger('ZIP')
+
+_SPOOL_DIRECTORY_NAME = 'tmp'
+
+
+def _get_spool_directory(configuration_manager: ConfigurationManager) -> Optional[str]:
+    """Directory to spill big ZIPs into. None falls back to the system temp directory."""
+    try:
+        directory = Path(configuration_manager.global_config_file_manager.get_config_directory_path())
+        spool_directory = directory / _SPOOL_DIRECTORY_NAME
+        spool_directory.mkdir(parents=True, exist_ok=True)
+        return str(spool_directory)
+    except OSError as e:
+        logger.debug('Failed to create the spool directory; falling back to the system one', exc_info=e)
+        return None
 
 
 class InMemoryZip:
     def __init__(self) -> None:
         self.configuration_manager = ConfigurationManager()
 
-        self.in_memory_zip = BytesIO()
-        self.zip = ZipFile(self.in_memory_zip, mode='a', compression=ZIP_DEFLATED, allowZip64=False)
+        self._spool_max_size = consts.ZIP_SPOOL_MAX_SIZE_IN_BYTES
+        self._buffer = tempfile.SpooledTemporaryFile(  # noqa: SIM115  # closed by cleanup(), lives past close()
+            max_size=self._spool_max_size,
+            dir=_get_spool_directory(self.configuration_manager),
+        )
+
+        # ZIP64 lifts the 65,535 entries and 4 GiB caps of the original ZIP format.
+        # It requires 64-bit offsets, so we only enable it on a 64-bit interpreter.
+        self._allow_zip64 = is_64bit()
+        self.zip = ZipFile(self._buffer, mode='a', compression=ZIP_DEFLATED, allowZip64=self._allow_zip64)
 
         self._files_count = 0
         self._extension_statistics = defaultdict(int)
@@ -35,17 +63,54 @@ def append(self, filename: str, unique_id: Optional[str], content: str) -> None:
     def close(self) -> None:
         self.zip.close()
 
+    def cleanup(self) -> None:
+        """Release the buffer, deleting the spilled temp file if there is one."""
+        self._buffer.close()
+
+    def __enter__(self) -> 'InMemoryZip':  # noqa: PYI034  # typing.Self needs Python 3.11
+        return self
+
+    def __exit__(self, *_: object) -> None:
+        self.cleanup()
+
+    def stream(self) -> IO[bytes]:
+        """The whole archive as a file object, rewound. Doesn't copy it into memory.
+
+        Note: before Python 3.11 SpooledTemporaryFile isn't a real IOBase, so the returned object
+        has no seekable()/readable()/writable(). read/seek/tell work on every supported version.
+        """
+        self._buffer.seek(0)
+        return self._buffer
+
     def read(self) -> bytes:
-        self.in_memory_zip.seek(0)
-        return self.in_memory_zip.read()
+        self._buffer.seek(0)
+        return self._buffer.read()
 
     def write_on_disk(self, path: 'Path') -> None:
         with open(path, 'wb') as f:
-            f.write(self.read())
+            shutil.copyfileobj(self.stream(), f)
 
     @property
     def size(self) -> int:
-        return getsizeof(self.in_memory_zip)
+        position = self._buffer.tell()
+        try:
+            self._buffer.seek(0, SEEK_END)
+            return self._buffer.tell()
+        finally:
+            self._buffer.seek(position)
+
+    @property
+    def is_rolled_over(self) -> bool:
+        """Whether the archive outgrew the threshold and moved from memory to the disk.
+
+        SpooledTemporaryFile spills on the write that crosses max_size, and the archive only grows,
+        so the size says it without reaching into the private _rolled flag.
+        """
+        return self.size > self._spool_max_size
+
+    @property
+    def allow_zip64(self) -> bool:
+        return self._allow_zip64
 
     @property
     def files_count(self) -> int:
 
@@ -49,6 +49,11 @@ def _read_text_file(path: str) -> Optional[str]:
         return None
 
 
+def is_64bit() -> bool:
+    """Whether the running Python interpreter is 64-bit (not the OS)."""
+    return sys.maxsize > 2**32
+
+
 def get_hostname() -> Optional[str]:
     try:
         return socket.gethostname() or None