I bought a simple USB C, 8TB SSD last year (about 6 months ago) for around $300 with tax. I checked it a few months ago and it was up to $600, and I just checked yesterday and the same model is $900.
After decades of experiencing primarily an increase in performance of tech at decreasing prices (on average) it is so strange to see such a long period of older tech increasing in price so much.
It reminds me of artificial shortages during the pandemic as people perceived or expected a scarcity, and some hoarded masks, toilet paper and more. Many hoarders of scarce supplies choosing to sell these at markup online.
It also reminds me of any historic $winter_holiday_of_choice "must-have" gift which are in short supply to meet the demand of $winter_holiday_of_choice sales.
If this AI race is a bubble, the sooner it ends, the better. If it isn't a bubble, each specialization of "AI" will probably have only one leader, and several losers struggling to find ways to recover their large bets.
A more interesting problem:
I expect the code-base that runs and is fed datasets can have IP, but can the resulting trained data be IP? How can you patent the resulting "learned" data if you can't describe it in a way that is distinct and unique from all other AI gamblers on a patent application?
It is increasingly complicated if the training data is copyrighted and significant elements from that copyrighted data appear (and can be reconstructed) by prompting the "AI" "learned" dataset.
None of this is financial or legal advice. It is a list of worries I have about risks to tech.
DEF CON