Obviously it’s a question that needs to be addressed: where are the returns on Nvidia GPUs?
Here are some of my thoughts on what’s missing from the article – not to say I have an answer, but to look at the article/question from other perspectives
1/ what are the risks of LOSING revenues if falling behind in AI?
Case in point – likely that Google can’t keep its grip on the AI equivalent of iOS search engine bar (or whatever the gateway to AI functions and monetization) in the next decade.
2/ what are the risks of losing top talents’ interest if not doing AI? And the culture of being at the frontier?
People wants to be part of the next gen thing. To be precise, top talents want to be the center of the next gen thing. If they see peers doing AI, they will not forgive themselves not doing AI.
If the company is seen as not the frontier, that’s a big risk in the next decade of not getting the top talents naturally. I bet now Google needs to do EXTRA to get talents it wants vs in the early days they were drawn to Google.
3/ what are the “infrastructure”?
In the article, author mentioned railroads. So what exactly is the “railroad” now?
The GPUs?
Or the data centers?
Or the foundation models?
Where does the overinvestment risk lie? (I bet it’s the third; also DC to some extend)
4/ If there any “moral hazard” here?
By having investments in AI, VCs should be inclined to discourage similar investments from other people.
Few entrants would enhance their return – whether it’s foundation models or GPU resources (easier to get GPUs / falling prices).
It’s tricky.
Like inflation – you need to tell the public to spend less to lower inflation; if you tell the public that you expect inflation to be down too early, it would be harder to come down.
If you are telling people AI is a good investment, then it may make returns lower.