Post

This is going to be a worse capital misallocation than "all we can do is break down these ships for scrap"

But they also won’t be because there will be no reason to make new generations of ai compute

Bingo. It is much, much, much worse than the dot.com bubble. Huge circular financing of Capex that is being recognised right now in "earnings". So when the bubble pops earnings won't just stop growing, they will massively contract. It's the ultimate example of the Kalecki profits equation in action.

i love how “it’s just like the dotcom bubble” is seen as a defense. Just because we got cheap fiber or whatever doesn’t mean it was okay to waste all those resources in the bubble.

Love that the entire economy is turning into a furious self-suck session. Very normal.

Regular Data centers need about 10x more cap and op ex. Hyperscaler DC’s need 20x. I just don’t see a path to making money here. Outside of the initial machinations. It’s a super bad investment.

Maybe they will be profitable in year 2030! (2030*2029*2028*...*1)

Those use cases are teeny tiny compared to the amount of data centers being built to the point that they may as well not exist

Well, by 2030 they'll be very outdated, and they're not particularly well suited to other tasks, so they'll be an outdated card being used inefficiently. You can dismantle the datacenter and repurpose the cards, but you're selling them at a tiny fraction of what you paid.

AI inference/training boxes are so weird that it's not really economical to use them for anything else; Nvidia boxes run on their own ARM CPUs and a 4RU box will be pulling like 20kW. They're highly specialized for AI and not a whole lot else - they can't even really do graphics anymore.

Plus, AI art now looks all the same. AI writing is converging. In their best-case scenario, the more the tech gains a foothold, the less new human intelligent input will exist to borrow from. You can see it in the way they are now gathering up so many old pre-AI books for training. It's nonsense.

Fair. They aren't exactly boosters. But they have blinders on to the risks.

Nah, they just let a complete moron write an opinion piece that will prove fantastically incorrect.

video games will run at 900,000 frames/sec tho

better climate models. new materials science. modeling protein structures. finding new medicines. image processing. like diagnostic images, or telescope images of deep space. finding patterns in large data sets, could be in medical records, or discrimination in criminal sentencing. lots of stuff.

I dunno. With all those GPUs you could play some really kewl video games.