To me, this is striking:
“I had previously assumed a 10-year depreciation curve, which I now recognize as quite unrealistic based upon the speed with which AI datacenter technology is advancing,” Kupperman wrote. “Based on my conversations over the past month, the physical data centers last for three to ten years, at most.”
In his previous analysis, Kupperman assumed it would take the tech industry $160 billion of revenue to break even on data center spending in 2025 alone. And that’s assuming an incredibly generous 25 percent gross margin — not to mention the fact that the industry’s actual AI revenue is closer to $20 billion annually, as the investment manager noted in his previous blog.
Meaning that currently his guess is that at current revenue, it’ll take 8 years to break even… So maybe they’ll get two years of profit at max, or be underwater by five years and $100 billion. Current revenue is going to change, but they’ll have to triple it in short order or be at risk… And that seems somewhat unrealistic with the lack of success in the AI products.
From https://hbr.org/2025/08/beware-the-ai-experimentation-trap apparently 95% of AI projects piloted by businesses have produced no measurable return.
AI revenue is going to go up. People and companies will pay for subscriptions. Military companies will buy tons of Ai. People just dont understand how society will completely transform, with robots, genetics, drones. But tech companies do. Thats why they invest now.
There wont be any AGI. These models cant reach AGI. But it will absolutely create huge revenue anyway.
I don’t know any people willing to pay a reasonable subscription. User uptake will be nowhere near 8x is a few years.
I have an entire team of colleagues who pay a subscription for it, and a girlfriend who pays for it also. She has Microsoft copilot at work and its useless, so she pays for chat gpt and it helps her deliver way more work much more quickly.
But I understand everyones milage may vary. I just see amazing value myself.
By the way, since we are on this topic - yesterday, I had the very bad luck of my Linux machine freezing exactly as it was doing a kernel build. Consequence was lots of corrupted files so I couldnt get into the desktop environment anymore.
With the help of chat gpt on mobile, I managed to fix a number of problems, like corrupted system libraries and gpg keys, and reinstall all the files from damaged packages, all without doing a full reinstall of the entire system. I could give it very strange error messages and it would help every step of the way.
I think its just incredible but sure, not everyone gets the same value from it.
You sound like an ad. There’s no way you “corrupted system libraries” from a hard shutdown unless you were doing something funky.
Just very happy. You would be too if it saved your entire system. You think im lying about what happened for a Lemmy comment? What would be the benefit of that?
The entire system froze during an update while it was working on the kernel. Had to reboot. When it came up, lots of library files were empty under /usr/lib, including pulseaudio libs and other important ones. No graphical user interface would start. Gdm didnt start.
Its the truth.
Oh I agree, it will go up. But having lived through the dotcom boom, I’m a bit more skeptical especially when I start seeing “OpenAI unveiled a “Central Park–size complex” in the Texas scrublands and disclosed plans to construct roughly a dozen more of them, a build-out that will cost $1 trillion.” when their revenue just cross $10 billion earlier this year. It’s not a linear growth, sure, but can you cite anything that indicates AI is describing a logarithmic growth rate?
What I’m reading in forecasts indicate a growth of the entire market of somewhere between 2.5 trillion and 4.8 trillion. So building a single data center (That’s just OpenAI, Microsoft, Orcale, etc are also building more) which is planned to cost somewhere over a half to a quarter of the total market grown estimate in 10 years is tech & investors who haven’t learned from history.
Remember that the dotcom’s were convinced that laying fiber across the US was a good bet at $100 billon (around $190 billion today), but it took another decade to actually use that fiber and start recouping the cost.
So sure, there’s growth and it’s likely to continue as products improve. However as an avid tester of narrow ai’s in various paid methods, I can say they are not justifying a doubling or tripling in subscription cost yet. They aren’t yet worth a human in many ways.
(Cite things and argue with me please! I’m happy to be wrong.)
nah, the person you responded too, does not understand that this is a really big hype bubble. The LLM predictive text stuff is really neat, but it cant be applied for anything critical due to always needing a human to double check the LLM output. The investment they are making is delusional megalomaniac billionaires who know they have to bet on the next big thing, but don’t understand what it is either.
This is the digital equivalent of throwing shit (GPUs) at the wall and hoping some will stick. It is possible the infrastructure they are building will be able to be used for something that doesn’t exist yet, but that’s really more thoughts and prayers than actual science.
Not to mention if they keep peddling LLMs as a holy grail they will keep tainting their AI stuff to slow adopters who wont give them another chance, or when LLMs will start causing more deaths per year than cars or smoking.
It’s a pipe dream and a massive gamble. It is late stage capitalism at it’s finest.
You think I dont understand when I just have a different opinion. I read what you think. I dont agree. That doesnt mean I dont understand.
You think this is a hype bubble. Sure, you can think that. Its not a fact. Your perception is your own. Some analysts agree with you, some doesnt. Either way, nobody can actually say if this is a bubble that will pop one day, or if it wont pop and revenues will come.
And you dont have to go into threads talking about me. You clearly dont even understand that its possible to have two different opinions about this and there is no right or wrong answer yet.
Sorry to offend you.
This is a classic internet argument paradox. Yes you can argue and have different opinions about subjective matters.
This is objectively a bubble because experts in bubbles say so. Just based on numbers.
You can still hold out hope and belief, but those are subjective and immaterial to this argument.
Here’s a handy schematic. https://lemmy.ca/post/53225771
I remember when “IT” or “Ginger” was the thing that Silicon.Valley tech titans were gaga over. Steve Jobs declared that IT would completely change how we build cities.
The idea has some traction in one-wheels, but mostly micromobilty is about ebikes and scooters. And it’s making inroads on changing our cities, but slowly.
The tech companies, that is to say, are not reliable predictors of what will be the next big thing.
Well, we will be around to see if this turns into a big crash but I dont think so. Im more thinking Trump will crash the US economy with his tariffs. Thats a real danger.
“I clearly hit a nerve in the industry, when judging by the number of individuals who reached out to chat,” he wrote in an followup blog post. “In total, I’ve spoken with over two-dozen rather senior people in the datacenter universe, and there was an interesting and overriding theme to our conversations; no one understands how the financial math is supposed to work. They are as baffled as I am, and they do this for a living.”
You can tell it’s clickbait garbage because there’s absolutely no data presented. It’s all ‘based on my conversations.’
It’s also disinformation. Most data center construction and electricity usage is not for AI. And with the models slimming, and equipment becoming more efficient, the usage for AI is actually trending downward. Which is why all of the anti-AI fomenting articles refuse to use data from this year.
