No science to it, more just a bit of idle musing on a few things rather than any properly thought through calculus tbh.
Think there's a difference between the established players and the startups themselves. In no particular order:
- most of the big beasts who are publicly quoted, whilst they've undoubtedly sunk shitloads into AI their business models aren't entirely predicated on them - they already have a lot of real value and revenue so for example if MS's Azure AI 'fails' it'll impact the margins of their Intelligent Cloud segment but it won't kill it because they won't lose those core Azure customers; similarly the Productivity and Business segment won't die on it's arse because Co-Pilot's not a big enough component of it's revenue stream yet
- their AI investments aren't financed wholly by debt but from reinvestment of revenues and most of those bigger companies are pretty well capitalised
- also I think we've already seen some market correction for most of them except maybe Alphabet, suggesting investors already think they've overegged it and aren't going at it quite so gung-ho, they'll probably be further corrections if/when the scale of how much they've cocked it up become apparent but not corrections that'll kill them
- though they may well have all over invested in compute infrastructure it'll be an over capacity issue rather than it all being complete redundant and I'd imagine that some smartarses will already have an answer to why their multi-gazillion sunk costs are actually a great platform for the next big thing they've dreamt up to use the capacity
- I think they'll take a bath on some of the external investments they've made particularly OpenAI but though the numbers are big again probably not enough to destabilise them
- I do think nvidia look a bit iffy because they appear to be in a variety of circular financing deals with the likes of OpenAI and Oracle and if that comes under more scrutiny, which it will if OpenAi ever gets to IPO then I'm not sure how that pans out
- I suspect some of the bigger companies will also be able to recover some of their loses by hoovering up bargains when the shitshow for the startups actually kicks in
think the start-ups themselves are a different kettle of fish and could be quite ugly but even then
- I think the bubble threat at the moment remains mostly confined to the VC arena and I'm not convinced that unlike the dot.com bubble enough of them are going to get to IPO in order to screw over the wider markets
- clearly the VCs will be looking to get their returns and offload their risk onto the rest of us but it might be that they'll actually have to take a big chunk of the pain themselves and though it'll be ugly for the people in those companies it is actually how the VC market is supposed to work in terms of risk/reward
- OpenAi now look unlikely to get to IPO this year and I think it's quite possible they've been rumbled, if we assume someone like Anthropic gets there first it'll determine to a large degree what it looks like for the rest of them but the longer it drags on the more data on the validity of this stuff will be out there and I suspect the more the markets will have cooled on these public offerings, ironically Trumps lunacy with the tariffs etc has probably helped by inadvertently slowing down some of it !
- I think I read somewhere that in 2025 over 50% of funding went into verticals rather than general AI which I would have thought had a better chance of if not living up to their valuations then at least offering some level of return, again I think that implies a little bit more discernment
As you can see from this ramble it's not a particularly well thought out view on my part and as I'm typing I think I'm probably describing something that will still be a bubble but maybe one that deflates rather than bursts. I think for all the overselling of AI it's still not the pump and dump madness of the dot.com bubble and you'd like to think markets wouldn't be that daft again...would they ? :-)
Tbf though, imo it needs to be deflated and slowed down significantly anyway because as with most technological innovation our understanding of how to govern it all is woefully behind the curve.
