nanotubes already failed, pretty much. we didn't find a way to make good contacts, so the CMOS/tube interface never worked as it should have worked.
Of all the crap that has been proposed in the last 10 years, gate all around is probably the most promising. End of the day, the industry will switch to whatever is the low-hanging fruit. We will be stuck with finfets until 3nm or so, then something else will come through, whatever that something else might be.