Without having access to the source code we can only speculate but I believe even in those days YouTube already outgrew vertical scaling and thus had to be built as a horizontally-scalable system. That is the hard part.
Adding extra nodes to an existing horizontally-scalable system (that has already been operating and has its bugs ironed out) is much easier.
Adding extra nodes to an existing horizontally-scalable system (that has already been operating and has its bugs ironed out) is much easier.