There's a really great point in this post from May: https://davekarpf.beehiiv.com/p/what-sort-of-ai-bubble-are-we-in (h/t Second Breakfast) that our estimation of the demand for AI is based almost entirely on a market that assumes the cost is effectively $0. Adjust the price to something that's actually profitable (or even just break-even), and a lot of that demand will evaporate. Which we're beginning to see now.