One thing he got wrong is that computer power and network speed will scale exponentially to accomodate infinitely large blockchain size. Intel's Q6600 of 2007 is still not too much slower than today's processors.