3 ms·

> An AI cloud can generate revenue of $10-12 billion dollars per gigawatt, annually. Citation needed.

by Apreche 9mo ago

Aurornis 9mo ago
This is coming from a group that does analysis on the semiconductor and cloud industries and provided very expensive access to their models and info. They are the citation.
- Apreche 9mo ago
  So I guess it’s not a bubble then since these companies are raking in the big revenues? Or maybe they are counting all those circular investments as revenues somehow?
  - airstrike 9mo ago
    Revenue isn't profit, and the presence or absence of either, separately or jointly, isn't sufficient to determine there is a bubble.
  - throw10920 9mo ago
    You can make a lot of money but still be in a bubble if speculation is significantly higher than (actual or potential) revenue. I think that's most people's assumption. It's not that AI is worthless, but that it's significantly less valuable than investors are betting on.
- dannersy 9mo ago
  I mean, if so then they are lying through their teeth.
  - Aurornis 9mo ago
    Based on what? I’m inclined to trust a well known industry analyst over an HN comment with no basis.
    - dannersy 9mo ago
      Fair enough, but it has been stated over and over that OpenAI's (as well as others) plan for profit is subscriptions. If their revenue predictions are based on that, then like others have said, it is mathematically impossible. Have you considered that the industry analysis might be a biased source since they are all in on a economic model that must grow at all costs or it collapses? Do you trust McKinsey consulting because they give industry analysis? Blind trust in these corporate entities is how we get Enron, WorldCom, and an opioid crisis. But hey, I'm just some asshole on the internet. Carry on.
zozbot234 9mo ago
If you do the math, that's $10-$12 per watt year. There's approx 24×365.25=8766 hours in a year, so assuming that the datacenters would be running 24×7, that boils down to $1.14 to $1.37 in revenue per kWh. That's not a bad deal if power really is a major part of the expense.
- jsnell 9mo ago
  As far as I can tell, power isn't actually a major part of the expense, it's dwarfed by the capex. Just the amortization on the GPU will be an order of magnitude higher than the cost of the power to run the GPU at 100%. (Assuming a 5 year depreciation period.)
mikelitoris 9mo ago
“Could”, sure... and I “could” fly if I strapped a jet engine to my ass
strange_quark 9mo ago
Even if that’s true, that seems like a putrid number, no? Assuming a single 1GW the data center runs 24/7 365, it’s consuming 8.76 TwH per year. Only being able to generate $10-$12B in revenue (not profit) per year while consuming as much electricity as the entire state of Hawaii (1.5M people) seems awful.

> An AI cloud can generate revenue of $10-12 billion dollars per gigawatt, annually. Citation needed.

by Apreche 9mo ago

> An AI cloud can generate revenue of $10-12 billion dollars per gigawatt, annually. Citation needed.

This is coming from a group that does analysis on the semiconductor and cloud industries and provided very expensive access to their models and info. They are the citation.

Apreche 9mo ago
So I guess it’s not a bubble then since these companies are raking in the big revenues? Or maybe they are counting all those circular investments as revenues somehow?
- airstrike 9mo ago
  Revenue isn't profit, and the presence or absence of either, separately or jointly, isn't sufficient to determine there is a bubble.
- throw10920 9mo ago
  You can make a lot of money but still be in a bubble if speculation is significantly higher than (actual or potential) revenue. I think that's most people's assumption. It's not that AI is worthless, but that it's significantly less valuable than investors are betting on.
dannersy 9mo ago
I mean, if so then they are lying through their teeth.
- Aurornis 9mo ago
  Based on what? I’m inclined to trust a well known industry analyst over an HN comment with no basis.
  - dannersy 9mo ago
    Fair enough, but it has been stated over and over that OpenAI's (as well as others) plan for profit is subscriptions. If their revenue predictions are based on that, then like others have said, it is mathematically impossible. Have you considered that the industry analysis might be a biased source since they are all in on a economic model that must grow at all costs or it collapses? Do you trust McKinsey consulting because they give industry analysis? Blind trust in these corporate entities is how we get Enron, WorldCom, and an opioid crisis. But hey, I'm just some asshole on the internet. Carry on.

zozbot234 9mo ago

If you do the math, that's $10-$12 per watt year. There's approx 24×365.25=8766 hours in a year, so assuming that the datacenters would be running 24×7, that boils down to $1.14 to $1.37 in revenue per kWh. That's not a bad deal if power really is a major part of the expense.

jsnell 9mo ago
As far as I can tell, power isn't actually a major part of the expense, it's dwarfed by the capex. Just the amortization on the GPU will be an order of magnitude higher than the cost of the power to run the GPU at 100%. (Assuming a 5 year depreciation period.)

mikelitoris 9mo ago

“Could”, sure... and I “could” fly if I strapped a jet engine to my ass

strange_quark 9mo ago

Even if that’s true, that seems like a putrid number, no? Assuming a single 1GW the data center runs 24/7 365, it’s consuming 8.76 TwH per year. Only being able to generate $10-$12B in revenue (not profit) per year while consuming as much electricity as the entire state of Hawaii (1.5M people) seems awful.