There may be different thesis out there, but the fact of prices going down definitely doesn't help the case that the major AI labs are are overvalued and need to vastly increase their revenue to stay afloat amidst their contractual obligations.
I honestly don't think there are too many people arguing "AI is actually too expensive to take my job". "AI being too much of a commodity to make meaningful profits" could very well be the case though.
I have seen the former argued many times, though mostly not on HN. It is a cope meme I’ve seen mostly on Twitter and Bluesky, but some on HN and the Elixir forum as well.
They are literally hoping Anthropic & OpenAI go bankrupt (totally possible), and everyone goes back to using their meat brains for work (never happening).
In this context it just doesn’t matter if these companies are profitable after amortizing training cost, because deployment of open weight models that already exist will continue to disrupt economies for at least the next decade if all research ended tomorrow.