A piece that will haunt me if/when I am wrong
**POD! **On this week, Martin and I dig into Andy Burnham’s devolution agenda, talking about integrated transport ticketing, making the lines on the map more consistent, and much more. Then we speak to
YIMBY Pod Jamie Airnes from
Cornish Lithium about his attempt to turn Cornwall into a lithium superpower.
[Listen here, or wherever you get your pods!](http://yimbypod.com/)
If you’re a smart opinion-haver on the internet, there’s one position on AI you can hold where you can’t really go wrong.
AI, you tell everyone, isn’t snake-oil, like the loudest detractors claim. The people who cling on to phrases like “fancy autocomplete” and “stochastic parrots” are clearly missing the bigger picture at this point. According to OpenAI, a billion people use ChatGPT at least once a month. AI is clearly doing something useful for them.
However, having conceded this, you’re supposed to turn and caveat your view, pointing out that despite this, it can still be the case that the current AI moment is a bubble. The unprecedented capital investment, which has now reportedly topped $1 trillion since early 2023, will ultimately prove unsustainable.
There are plenty of weird-looking, circular-ish deals that would seem to evidence this too. For example, last September Nvidia pledged to invest up to $100bn in OpenAI – and in return OpenAI pledged to use the cash to buy huge numbers of Nvidia chips. And last November, Microsoft and Nvidia pledged to invest up to $15bn into Anthropic, in exchange for the company behind Claude committing to spend $30bn on Microsoft Azure cloud computing capacity that uses Nvidia chips.
There are other oddities too. In recent days, Nvidia has announced a new plan to work with institutional investors like pension funds to invest in chips as a new asset class.
So I can understand why seeing this, it might feel as though something strange is going on – and that it feels like at some point, the bubble is going to have to burst.
And as a narrative, this makes total sense. After all, this is the exact story of the original dot-com crash. The underlying ‘internet’ technology was far from worthless, but billions of dollars were wiped out nonetheless.
Anyway, this is all to say that I think the bubble hypothesis is a completely reasonable analysis of the current situation. It might even be correct. I am very open to the idea that this wildly hyped new technology might not work out in the short to medium term, as this is a reasonable default assumption, similar to predicting that Spain will do well in a World Cup, or that a Christopher Nolan film will receive positive reviews.
But here’s the thing that’s awkward to admit. The more I see of what AI is now capable of, and the more I experience these new capabilities for myself, the more I’m sympathetic to the Silicon Valley perspective that this is not a bubble – that this technology is really – rapidly – changing everything. And that, despite the ungodly sums of money spent fabricating chips and pouring concrete to build data centres, there might not be a bubble at all.
So this week, I thought I’d explain how I’ve ended up here – and make the bull case for the AI investment boom, and explain why the bubble explanation might not be true.
Keep reading with a 7-day free trial #
Subscribe to Odds and Ends of History to keep reading this post and get 7 days of free access to the full post archives.