2 ms·

> analysts have estimated that it will take $2 trillion a year in revenue to pay for the infrastructure that has already been built I doubt any credible analys

by jsnell 2mo ago

> analysts have estimated that it will take $2 trillion a year in revenue to pay for the infrastructure that has already been built I doubt any credible analyst has claimed that. What's the total AI capex that's already been spent? About $1T? It's a pretty absurd idea that those DCs need to make $2T/year for 5-7 years -> $10T-14T over their lifetime to break even. (Yes, this is nitpicking in the sense that there are probably analysts talking about how projected and sustained capex at >1T/year will require 2T/year in revenue, so patching the article won't be a biggie. But this article is cosplaying as financial analysis and leads off with such an obviously incorrect argument. What does that say about the credibility of the rest of the article?) > Diseconomies of Scale. Another very basic mistake here. The author starts talking about efficiency in the context of past technologies. That's lower unit costs as scale increases. But for AI, they seem to switch from talking about unit costs to total costs. Or at least I can't explain what they say about models getting more expensive over time in any other way, because that is not true about unit costs. We've never seen economies of scale as large as for AI. For a given quality level, the cost has been dropping at >10x per year, not increasing.

Schiendelman 2mo ago
I appreciate someone else here sees this math is way off. With what seems to be their current revenue, the frontier companies are likely better off than Uber was pre-IPO. The "circular financing" we've seen probably lowers risk relative to external financing, since both parties get aligned incentives. It's too bad Nvidia keeps getting hammered, they're selling the shovels in a gold rush that'll last for years, and I don't think that's priced in.
B56b 2mo ago
He might be referring to this? https://www.wheresyoured.at/big-tech-2tr/ https://www.wheresyoured.at/big-tech-2tr/ That figure is $2T in the next 4 years though. $2T/year would be quite silly.
tedmiston 2mo ago
The post is just full of half-baked conjectures masquerading as facts... combined with "things they read" by unspecified authors and sources... the author seems to prefer engaging in AI doomerism as opposing to actually understanding. > Diseconomies of Scale. Every new technology I can think of thrived, in part, due to economies of scale, where the larger the industry grew, the more efficient it got. AI is going in the opposite direction, where every new AI model consumes more resources than its predecessors. This may turn out to be the fatal flaw – the bigger the industry gets, the more its operating costs increase. [Agreed] Newer gens of models are more power-efficient per task, not less. This is a low quality post full of basic errors.
- hakfoo 2mo ago
  It's possible that the demands on the model are growing faster than it gets efficient though. A lot of the improvement in output quality/metrics come from more churning and passes. You could argue that a RTX5090 needs less power than a 1996 3Dfx Voodoo 1 to render GLQuake at 320x200/60fps, but it's rarely being used for that.
tim333 2mo ago
I agree a lot of the dooming is from people who don't really do the maths. Total capex so far is probably about $1T. Recent annual earnings are $132bn at Google, $101bn at msft, 60bn at Meta, $120bn nvda adding up to $413bn after tax. They could pretty much write down the whole capex to zero against two years earnings from those. It's not that out of whack. Not mentioning that we are kind at the point in the Wait but Why cartoon between that's cute and wft https://waitbutwhy.com/wp-content/uploads/2015/01/Intelligence2.png https://waitbutwhy.com/wp-content/uploads/2015/01/Intelligen... From the article from ten years ago https://waitbutwhy.com/2015/01/artificial-intelligence-revolution-1.html https://waitbutwhy.com/2015/01/artificial-intelligence-revol... It's interesting that it shows human level at 2025 and we've just seen a wtf story on HN https://news.ycombinator.com/item?id=49048681 https://news.ycombinator.com/item?id=49048681

by jsnell 2mo ago

Schiendelman 2mo ago

I appreciate someone else here sees this math is way off. With what seems to be their current revenue, the frontier companies are likely better off than Uber was pre-IPO. The "circular financing" we've seen probably lowers risk relative to external financing, since both parties get aligned incentives. It's too bad Nvidia keeps getting hammered, they're selling the shovels in a gold rush that'll last for years, and I don't think that's priced in.

B56b 2mo ago

He might be referring to this? https://www.wheresyoured.at/big-tech-2tr/ https://www.wheresyoured.at/big-tech-2tr/ That figure is $2T in the next 4 years though. $2T/year would be quite silly.

tedmiston 2mo ago

The post is just full of half-baked conjectures masquerading as facts... combined with "things they read" by unspecified authors and sources... the author seems to prefer engaging in AI doomerism as opposing to actually understanding. > Diseconomies of Scale. Every new technology I can think of thrived, in part, due to economies of scale, where the larger the industry grew, the more efficient it got. AI is going in the opposite direction, where every new AI model consumes more resources than its predecessors. This may turn out to be the fatal flaw – the bigger the industry gets, the more its operating costs increase. [Agreed] Newer gens of models are more power-efficient per task, not less. This is a low quality post full of basic errors.

hakfoo 2mo ago
It's possible that the demands on the model are growing faster than it gets efficient though. A lot of the improvement in output quality/metrics come from more churning and passes. You could argue that a RTX5090 needs less power than a 1996 3Dfx Voodoo 1 to render GLQuake at 320x200/60fps, but it's rarely being used for that.

tim333 2mo ago

I agree a lot of the dooming is from people who don't really do the maths. Total capex so far is probably about $1T. Recent annual earnings are $132bn at Google, $101bn at msft, 60bn at Meta, $120bn nvda adding up to $413bn after tax. They could pretty much write down the whole capex to zero against two years earnings from those. It's not that out of whack. Not mentioning that we are kind at the point in the Wait but Why cartoon between that's cute and wft https://waitbutwhy.com/wp-content/uploads/2015/01/Intelligence2.png https://waitbutwhy.com/wp-content/uploads/2015/01/Intelligen... From the article from ten years ago https://waitbutwhy.com/2015/01/artificial-intelligence-revolution-1.html https://waitbutwhy.com/2015/01/artificial-intelligence-revol... It's interesting that it shows human level at 2025 and we've just seen a wtf story on HN https://news.ycombinator.com/item?id=49048681 https://news.ycombinator.com/item?id=49048681