It's inevitable. The value proposition of the software isn't large enough to justify (by itself) the valuation. Not to mention it's trained on lots of public or scraped data.
It's a gold rush, and the only guy getting rich will be the one selling the shovel (nvidia + cloud in this case).
There's money now to _build_ these massive data centers. If there is long-term money to operate, we'll see.
It might mean we'll get cheaper cloud compute in the future, but I'm also positive on the market's ability to absorb the excess capacity quickly. It's Jervon's paradox.
I am less exuberant on market willing to pay the rates that are need to cover the loan servicing costs of these investments. Numbers look really suspicious because they are so significant compared to economy in general.
It's a gold rush, and the only guy getting rich will be the one selling the shovel (nvidia + cloud in this case).