diff --git a/packages/markitdown/src/markitdown/converters/_xlsx_converter.py b/packages/markitdown/src/markitdown/converters/_xlsx_converter.py
index 9f794a3b77..706645517f 100644
--- a/packages/markitdown/src/markitdown/converters/_xlsx_converter.py
+++ b/packages/markitdown/src/markitdown/converters/_xlsx_converter.py
@@ -150,7 +150,14 @@ def convert(
 
             for s in sheets:
                 md_content += f"## {s}\n"
-                html_content = sheets[s].to_html(index=False)
+                html_content = sheets[s].to_html(
+                    index=False,
+                    # pandas defaults large/small floats to 6-significant-digit
+                    # scientific notation, silently dropping digits the user
+                    # entered (123456789.123 -> 1.234568e+08). repr gives the
+                    # shortest string that round-trips the exact stored value.
+                    float_format=lambda value: repr(float(value)),
+                )
                 md_content += (
                     self._html_converter.convert_string(
                         html_content, **kwargs
@@ -241,7 +248,14 @@ def convert(
         md_content = ""
         for s in sheets:
             md_content += f"## {s}\n"
-            html_content = sheets[s].to_html(index=False)
+            html_content = sheets[s].to_html(
+                index=False,
+                # pandas defaults large/small floats to 6-significant-digit
+                # scientific notation, silently dropping digits the user
+                # entered (123456789.123 -> 1.234568e+08). repr gives the
+                # shortest string that round-trips the exact stored value.
+                float_format=lambda value: repr(float(value)),
+            )
             md_content += (
                 self._html_converter.convert_string(
                     html_content, **kwargs
diff --git a/packages/markitdown/tests/test_files/float_precision.txt b/packages/markitdown/tests/test_files/float_precision.txt
new file mode 100644
index 0000000000..8c5e46edf9
--- /dev/null
+++ b/packages/markitdown/tests/test_files/float_precision.txt
@@ -0,0 +1,3 @@
+Synthetic BIFF8 workbook generated with xlwt 1.3.0.
+One sheet named Precision; A1 = value, A2 = 123456789.123, A3 = 0.1, A4 = 1e-10.
+No real user data. The committed fixture avoids adding an XLS writer to test or runtime dependencies.
diff --git a/packages/markitdown/tests/test_files/float_precision.xls b/packages/markitdown/tests/test_files/float_precision.xls
new file mode 100644
index 0000000000..58e98b3fa6
Binary files /dev/null and b/packages/markitdown/tests/test_files/float_precision.xls differ
diff --git a/packages/markitdown/tests/test_xlsx_float_precision.py b/packages/markitdown/tests/test_xlsx_float_precision.py
new file mode 100644
index 0000000000..7ad8dcfc44
--- /dev/null
+++ b/packages/markitdown/tests/test_xlsx_float_precision.py
@@ -0,0 +1,55 @@
+"""Spreadsheet float values must survive conversion without pandas' 6-digit scientific notation."""
+import io
+from pathlib import Path
+
+import numpy as np
+import pandas as pd
+from openpyxl import Workbook
+
+from markitdown import MarkItDown, StreamInfo
+
+
+def test_xlsx_float_precision_preserved():
+    wb = Workbook()
+    ws = wb.active
+    ws["A1"] = "money"
+    ws["A2"] = 123456789.123
+    ws["A3"] = 0.1
+    ws["A4"] = 1e-10
+    buf = io.BytesIO()
+    wb.save(buf)
+    buf.seek(0)
+    result = MarkItDown().convert_stream(buf, stream_info=StreamInfo(extension=".xlsx"))
+    assert "123456789.123" in result.markdown
+    assert "1.234568e+08" not in result.markdown
+    assert "| 0.1 |" in result.markdown
+    assert "| 1e-10 |" in result.markdown
+
+
+def test_xls_float_precision_preserved():
+    fixture = Path(__file__).parent / "test_files" / "float_precision.xls"
+    with fixture.open("rb") as stream:
+        result = MarkItDown().convert_stream(
+            stream, stream_info=StreamInfo(extension=".xls")
+        )
+    cells = [
+        line.strip("| ")
+        for line in result.markdown.splitlines()
+        if line.startswith("|")
+    ]
+    assert "123456789.123" in cells
+    assert "0.1" in cells
+    assert "1e-10" in cells
+
+
+def test_float_format_handles_numpy_scalars():
+    # pandas hands float_format numpy scalars, and with numpy >= 2 a bare
+    # repr(value) would render "np.float64(0.1)" in the table. The float()
+    # cast inside the converter's lambda is what prevents that.
+    df = pd.DataFrame({"money": [np.float64(0.1)]})
+    buf = io.BytesIO()
+    df.to_excel(buf, index=False)
+    buf.seek(0)
+    result = MarkItDown().convert_stream(buf, stream_info=StreamInfo(extension=".xlsx"))
+    assert "np.float64" not in result.markdown
+    assert "| 0.1 |" in result.markdown