In the past you could do almost anything on a personal computer, it was generally about as fast as mainframe or high end workstation.
Training large AI models is currently impossible for most home users. They just do not have the processing power.
For all the examples mentioned in the parent article, PCs were significantly under-powered compared to workstations, much less main frames.
An explosion of hardware development between 2005 and 2020 has lead to an era where hardware outperformed software needs. Now software is catching up.
But there have always been use cases for high end hardware, and always will be.
video production, climate simulations, pdes, protein folding, etc etc