This comes up high on Google, so I'd like to add some contextual information about the original question (emphasis mine):
Why does Node.js' fs.readFile() return a buffer instead of string?
Because files aren't always text
Even if you as the programmer know it: Node has no idea what's in the file you're trying to read. It could be a text file, but it could just as well be a ZIP archive or a JPG image — Node doesn't know.
Because reading text files is tricky
Even if Node knew it were to read a text file, it still would have no idea which character encoding is used (i.e. how the bytes in the file map to human-readable characters), because the character encoding itself is not stored in the file.
There are ways to guess the character encoding of text files with more or less confidence (that's what text editors do when opening a file), but you usually don't want your code to rely on guesses without your explicit instruction.
Buffers to the rescue!
So, because it does not and can not know all these details, Node just reads the file byte for byte, without assuming anything about its contents.
And that's what the returned buffer is: an unopinionated container for the raw bytes in the file. How these bytes should be interpreted is up to you as the developer.