CSV Limits: Maximum Rows, Columns and Field Length

In short: the CSV format has no limits. RFC 4180, the document that describes CSV, sets no maximum file size, row count, column count or field length, so every number on this page belongs to a program that reads or writes CSV rather than to CSV itself. The first one most people meet is Excel, whose worksheet holds 1,048,576 rows and 16,384 columns. RowSlice’s tools read a file a few megabytes at a time and are limited by free disk space rather than by a row count, except where a row has to be recognised again later — Remove Duplicates, Compare, the Analyzer and the Validator. Those four ceilings are listed below, each read from the constant in the code that enforces it.

What RFC 4180 Defines, and What It Does Not

RFC 4180 is the closest thing CSV has to a specification. It is an Informational document rather than a standard, nothing enforces it, and it is what RowSlice’s parser follows. It is also the answer to the question people most often ask about CSV size, because it declines to answer it.

The size limits a CSV does meet belong to whatever opens, stores or carries it — a program’s memory, an email attachment cap, a FAT32 drive, an upload form — and the guide to CSV file size limits gives each one in megabytes or gigabytes, with its source.

Spreadsheet Limits

Excel has a second limit that is not printed on the grid: memory. Microsoft’s guide to Excel’s resource errors states that “in any 32-bit application, there’s a 2-GB limitation”, while “the 64-bit version allows all available physical RAM on the machine to be used”; a later change doubled 32-bit Excel’s allowance to 4 GB on 64-bit Windows. A CSV large enough to hit that is almost always over the row limit as well. Excel’s row limit explained covers what happens to the rows that do not load.

If a CSV may be over the row limit, count its rows first rather than opening it in Excel; past 1,048,575 data rows, split the CSV into parts of 1,000,000 rows, each starting with the header, and every part opens whole.

Excel Changes Values Before It Runs Out of Rows

The 15-digit limit applies only once Excel has decided a value is a number, which is why a CSV converted to a real .xlsx workbook with every value already written as text keeps every digit. Stop Excel turning long numbers into 4.5E+15 and why Excel removes leading zeros cover the two cases separately.

Google Sheets Counts Cells, Not Rows

Google’s limit is 20 million cells, counted as rows multiplied by columns across every tab of one spreadsheet, so how many rows fit depends on how wide the file is. The header row counts as a row: RowSlice’s CSV Analyzer reports a file’s data rows and columns, so add one for the header and multiply. The table below is that division; it assumes one tab and no other data in the spreadsheet.

Google also caps a spreadsheet created in or converted to Sheets at 100 MB, so a file with long values can fail before it reaches 20 million cells. Google’s help page gave both figures when we read it on 10 October 2026. The cell limit was 10 million until Google doubled it in a rollout that began on 10 September 2026, so older references still give 10 million. What to do when a CSV is too large for Google Sheets covers the options when a file does not fit.

Import Tools

Import limits are published per product and are stated in rows, in megabytes, or in both. These were read from each vendor’s own documentation on October 10, 2026.

How to split a CSV to fit an import limit covers more products, the daily allowances some of them add on top, and how to work out the right part size.

What RowSlice’s Own Tools Handle

RowSlice reads a file in chunks of a few megabytes in the browser and writes results to the browser’s storage on the visitor’s own disk, so memory use does not grow with file size and most tools have no row limit at all. Four of them — Remove Duplicates, Compare, the Analyzer and the Validator — have to recognise a row again after it has gone past, and each keeps a 64-bit fingerprint per distinct row or key in memory, which is where their ceilings come from. Every figure below is the constant that enforces it, in the file that holds it.

The largest run RowSlice has measured is a 1 GB file of 25 million rows split in Firefox with every row intact. Tools with no entry above — splitting by rows, filtering rows, merging files, writing JSON out of a CSV and converting the encoding — keep nothing from one row to the next except the counts they report, so they have no row ceiling of their own.

Encodings

A CSV does not record its own encoding, so every program that opens one is guessing. Which encoding a file is in decides whether accented characters survive, and whether RowSlice can page through the file row by row at all.

A byte order mark, or BOM, is a short signature at the very start of a file naming the Unicode encoding that follows. It is not part of the data, and a program that does not recognise one reads its bytes as characters and glues them to the first column name.

Excel wants a UTF-8 BOM on a file it is asked to open by double-click; JSON readers reject one, which is why RowSlice’s CSV to JSON converter never writes one and its JSON to CSV converter writes one by default. UTF-8, UTF-16 and Windows-1252 explained covers how to tell which encoding a file is in from what is on screen.

Delimiters

RFC 4180 names the comma and nothing else, but the character that separates fields varies by region and by the program that wrote the file. Excel splits a double-clicked CSV on the Windows list separator, which is a semicolon wherever the decimal separator is a comma — so the same file opens in columns on one machine and in a single column on the next. RowSlice detects comma, semicolon, tab and pipe automatically, and the UTF-8 converter changes a file’s delimiter as well as its encoding.

Definitions

The first two of those are why two programs report different counts for the same file. RowSlice’s CSV Analyzer reports all three numbers — data rows, CSV records and physical lines — for a file of any size, and how to count the rows in a CSV explains why they differ.