diff --git a/addons/l10n_fr_fec/wizard/account_fr_fec.py b/addons/l10n_fr_fec/wizard/account_fr_fec.py
index 1e3b0943006..1da8fd008fd 100644
--- a/addons/l10n_fr_fec/wizard/account_fr_fec.py
+++ b/addons/l10n_fr_fec/wizard/account_fr_fec.py
@@ -302,6 +302,9 @@ class AccountFrFec(models.TransientModel):
             aj_name = f"COALESCE(aj.name->>'{lang}', aj.name->>'en_US')"
         else:
             aj_name = "aj.name"
+
+        query_limit = int(self.env['ir.config_parameter'].sudo().get_param('l10n_fr_fec.batch_size', 500000)) # To prevent memory errors when fetching the results
+
         sql_query = f'''
         SELECT
             REGEXP_REPLACE(replace(aj.code, '|', '/'), '[\\t\\r\\n]', ' ', 'g') AS JournalCode,
@@ -358,20 +361,44 @@ class AccountFrFec(models.TransientModel):
             am.date,
             am.name,
             aml.id
+        LIMIT %(limit)s
+        OFFSET %(offset)s
         '''
-        self._cr.execute(sql_query, {**params, 'date_to': self.date_to})
 
-        for row in self._cr.fetchall():
-            rows_to_write.append(list(row))
+        with io.BytesIO() as fecfile:
+            csv_writer = pycompat.csv_writer(fecfile, delimiter='|', lineterminator='')
+
+            # Write header and initial balances
+            for initial_row in rows_to_write:
+                initial_row = list(initial_row)
+                # We don't skip \n at then end of the file if there are only initial balances, for simplicity. An empty period export shouldn't happen IRL.
+                initial_row[-1] += u'\r\n'
+                csv_writer.writerow(initial_row)
+
+            # Write current period's data
+            query_offset = 0
+            has_more_results = True
+            while has_more_results:
+                self._cr.execute(sql_query, {**params, 'date_to': self.date_to, 'limit': query_limit + 1, 'offset': query_offset})
+                query_offset += query_limit
+                has_more_results = self._cr.rowcount > query_limit # we load one more result than the limit to check if there is more
+                query_results = self._cr.fetchall()
+                for i, row in enumerate(query_results[:query_limit]):
+                    if i < len(query_results) - 1:
+                        # The file is not allowed to end with an empty line, so we can't use lineterminator on the writer
+                        row = list(row)
+                        row[-1] += u'\r\n'
+                    csv_writer.writerow(row)
+
+            base64_result = base64.encodebytes(fecfile.getvalue())
 
-        fecvalue = self._csv_write_rows(rows_to_write)
         end_date = fields.Date.to_string(self.date_to).replace('-', '')
         suffix = ''
         if self.export_type == "nonofficial":
             suffix = '-NONOFFICIAL'
 
         self.write({
-            'fec_data': base64.encodebytes(fecvalue),
+            'fec_data': base64_result,
             # Filename = <siren>FECYYYYMMDD where YYYMMDD is the closing date
             'filename': '%sFEC%s%s.csv' % (company_legal_data, end_date, suffix),
         })
@@ -387,7 +414,7 @@ class AccountFrFec(models.TransientModel):
             'target': 'new',
         }
 
-    def _csv_write_rows(self, rows, lineterminator=u'\r\n'):
+    def _csv_write_rows(self, rows, lineterminator=u'\r\n'): #DEPRECATED; will disappear in master
         """
         Write FEC rows into a file
         It seems that Bercy's bureaucracy is not too happy about the