Because the numbers cubed are way larger than the numbers that Excel can represent precisely. Excel uses 64-bit IEEE 754 floats for all arithmetic. They can be used to accurately describe integers smaller than 2^53, or ~10^16. The first number cubed is ~-5*10^50, so you are very clearly outside the range where excel is precise.

You should not use Excel to do any serious mathematics, or really anything else where numerical precision is important.

Why is it that a program that people use for accounting doesn’t use arbitrary precision arithmetic?

Because for accounting the numbers easily fit in the supported range; even the US National Debt (https://www.usdebtclock.org/)...

No need to downvote this comment. It is a legitimate numerical computation question and one that has interesting answers as evidenced by this sub-thread. Not everyone knows about how computers do floating point arithmetic, handles large ints, or otherwise.

You need more precision. Try this in your browser console instead, using the new BigInt type in JavaScript:

    (-80538738812075974n)**3n + 80435758145817515n**3n + 12602123297335631n**3n

Welcome to numerical computing :D!

Btw:

julia> (-80538738812075974)^3 + 80435758145817515^3 + 12602123297335631^3

42

Or

  CL-USER> (+ (expt -80538738812075974 3)  (expt 80435758145817515 3)  (expt 12602123297335631 3))
  42

same as Google.

https://www.google.com/search?sxsrf=ACYBGNQRb1b08L-cU-JY3GHr...

Oddly enough, it works for Bing.

https://www.bing.com/search?q=%3D%28-80538738812075974%29%5E...

Excel is probably coercing them to floating point, instead of using infinite-precision integers. Wolfram-Alpha gets it right: https://www.wolframalpha.com/input/?i=%28-80538738812075974%...

Excel can't to math on arbitrarly large ints, so tuncates all the numbers and gives you the wrong result

Sadly `awk/gwak` fails as well returning

   198929873392615583518074640613244928

I take that back, needs the `M` flag

awk -M 'END{print (-80538738812075974)^3 + 80435758145817515^3 + 12602123297335631^3}' /dev/null

42

but `bc` of course works

echo '(-80538738812075974)^3 + 80435758145817515^3 + 12602123297335631^3' | bc

42

Integer overflow?

Works in Windows calculator.