I got confused by your explanation, I think partly because of conversion to decimal without specifying whether column A is 0 or 1. To match the row numbering system, I’d assume it’s 1, but mathematically it might be easier to analyze if it starts at 0.

Here’s my attempt to understand it better.

How many spreadsheet rows are 1 digit? 1-9, that’s 9 rows.

How many rows are 2 digit? 10-99, that’s 90 rows.

How many rows are 3 digit? 100-999, that’s 900 rows.

The pattern is: there are 9 × 10^(d-1) rows that are d digits.

Now for columns. How many columns are 1 letter? A-Z, that’s 26.

How many columns are 2 letters? AA-ZZ, that’s 26².

How many columns are 3 letters? AAA-ZZZ, that’s 26³.

The pattern is: there are 26^d columns that are d letters.

Now let’s look at running totals. How many rows are at most 2 digits? 9 + 90 = 99

How many rows are at most 3 digits? 99 + 900 = 999

The pattern is: 10^d - 1 rows are at most d digits.

How many columns are at most 2 letters? 26 + 26² = 26×(1 + 26) = 26×27

How many columns are at most 3 letters? 26×27 + 26³ = 26×(27 + 26²)

Alternatively: 26 + 26² + 26³

I can’t find a more compact way of writing it.

The pattern is: Σ26^k (k from 1 to d) columns are at most d letters

I might be wrong but I don’t think this matches up with what you said, which is 26×27^(d-2) + c (where c is some constant for converting between what we’re trying to measure, probably 1 or -1)