Pricing for AI services is modelled after “the drug-dealer’s algorithm” (the first go is free)
Specifically for enterprise usage, the model assumes that enterprises will find that using the service gives enough value to cover the subsidised cost of the service through the lock-in period.
The assumed depreciation of hardware used in AI services is wrong, and the bubble will burst soon. Providers will have to raise prices significantly to remain solvent.
This is the intention. David Rosenthal recently wrote a blog piece about it. It goes quite deep into the economics but this is what I understand from the piece:
Good piece.