The alternative prediction is that this is in fact sustainable and AI companies will in fact have revenue to keep the bubble inflated for a lot longer, just in the worst way - by extracting the value of human-created reliability and trust from the market:

CEOs have also bought into AI almost to a person, and are using it to replace workers, results be damned. AI can’t do the things they believe it can, but to them, if they can fake satisfying a need with AI for $5, that is preferable to actually satisfying a need with a real employee for $10.

The CEO is happy because his company saved $5 and he’s met his stock option incentive target, the AI companies are happy to pocket that $5 instead of the employee getting $10. Maybe they even raise the customer’s price to $12 as AI rent-seeking starts rising, and both companies get $6 each. Win-win, life will go on, just worse for everyone else.

PC costs certainly aren’t helping, but there’s an entire cross-section of income and age demographics whose only computing device is and has always only been their phones.

I was curious so I looked it up. This site suggests 1 in 7 households in the US “either lack a computer at home or rely solely on a smartphone for internet access”, heavily weighted to lower-income states like Mississippi, West Virginia, Arkansas and Louisiana: https://www.benton.org/blog/computer-ownership-and-digital-divide