Dr. Steve Hsu has looked into the question of whether or not AI spending is a 'bubble':https://www.youtube.com/watch?v=__p_8xF0oSQHis conclusion is that it has some bubble characteristics under traditional finance models (especially the kinds that would be used by bond investors), and that AI revenue is generally not aligned with the rate of returns traditionally demanded by investors.However, this is not a typical market, since the anticipated future revenue is still contingent upon an emerging technology, which leads to different overall investor psychology, and may sustain AI spending at higher rates than would be possible for other industries.He notes that hyperscaler firms (those which supply/build out data centers) are useful to watch, since an actual AI downturn might begin in the hyperscaler market rather than within AI firms themselves. AI firms generally don't do their own hyperscaling; it is purchased from other companies. So basically, under traditional metrics, AI is a bubble. However, AI is not a traditional product. That's more or less his conclusion. He says that this will all need to be at least partly resolved by 2028 due to the need for AI companies to issue bonds to raise capital. Thus far, bond-issued finance has not dried up for those firms.But bond investors are pickier about returns than more speculative investment finance firms, and they will need to see more results by 2028 or so.
