345
Profits?!? Where we're going, we don't need profits!
(media.piefed.ca)
you are viewing a single comment's thread
view the rest of the comments
view the rest of the comments
this post was submitted on 30 Sep 2025
345 points (100.0% liked)
Political Memes
9751 readers
1900 users here now
Welcome to politcal memes!
These are our rules:
Be civil
Jokes are okay, but don’t intentionally harass or disturb any member of our community. Sexism, racism and bigotry are not allowed. Good faith argumentation only. No posts discouraging people to vote or shaming people for voting.
No misinformation
Don’t post any intentional misinformation. When asked by mods, provide sources for any claims you make.
Posts should be memes
Random pictures do not qualify as memes. Relevance to politics is required.
No bots, spam or self-promotion
Follow instance rules, ask for your bot to be allowed on this community.
No AI generated content.
Content posted must not be created by AI with the intent to mimic the style of existing images
founded 2 years ago
MODERATORS
.
Yeah but AI is burning so much that there literally won't be any VC money by 2027. Like, at all.
.
Burning money in this context does not mean that the money disappears, but that the thing you get for your money is worthless, but you pretend it isn't. Here, it's GPUs.
The problem is that how it should really work is that people make a product, they sell that product for money, and since the product they sold is worth as much as the money paid for it, that value is now twice in the economy, once in the shape of cash and once in the shape of the product. This is how economic growth should work.
Now you could say that if someone paid that much for that product, it should be worth that much. But if it's just you and your friend passing it back and forth, that would feel fraudulent, as it isn't worth as much to most people. All the AI money and growth, which is a significant portion of the US GDP growth, is money that's passed between like 7 companies, in a circular way.
And the main valuable product in the whole AI boom is not even AI, but GPUs, as in that's the only thing people actually pay for. GPUs have a lifetime of months to years at most, even if the demand for them does not crash.
.
The AI GPU is the only product that actually makes money, as in a positive value chain that is not subsidized by VC money and is able to exist without being subsidized by VC money?
The point is, almost all of the AI growth and all of the profits have been AI GPUs. If I'm missing the point everyone else seems to as well, and when all your customers keep missing the point of your products, your business is not long for this world.
My point is, which product, built by AI or built on AI is going to earn a trillion to pull OpenAI out of the hole and make AI keep existing after 2027?
Whatsapp messages are incredibly cheap, while AI queries are expensive to the point nobody wants to pay the actual price of it.
And where is that new industrial revolution? AI revenues are abysmal, costs are astronomical, the integration into our way of working is massively unprofitable, conversion rates are microscopic both in the consumer and enterprise spaces, while even paid users cost a lot of money to companies.
My point is, OpenAI has been around for a decade, the original LLM papers for 8 years, ChatGPT for three years. There is no evidence it's going to even break even, much less be a "new industrial revolution". Research shows using AI does not speed up, and in some cases actually slows down work.
And the problem is, the VC money that can fund the kind of loss leaders you are speaking of is going to run out in six quarters. I don't mean that OpenAI has a six quarter runway, I mean the whole of the US venture capital sector has that. There is literally no time left for anything that is only "going to be". This is it. All of it.
That industrial revolution should have come already, at this point this thing is a flop like the Metaverse, only difference it's a much bigger crater.
.
Okay, but since subscriptions for foundation models will have to go up at least ten to fiftyfold since it's not profitable, how much money have you made with it? Is your convenience worth a few grand per month? Are there enough people like that to pay for this? And it's not a hypothetical, the United States is running out of money to give OpenAI to fuck around in like 6 financial quarters.
Yeah that works until it doesn't and it breaks the computer. Still, how much money have people made with this?
It really doesn't matter what I do. I got a subscription for a year for one of the chatbots as a freebie with my credit card. I've used it a lot for internet search. It's useful. I would pay maybe 2-5 EUR a month for it if the company was decent and would provide first class support for my use cases, like a Linux app, which it doesn't. I got a local model running, which is useful, I might keep it.
But that's beside the point. An industrial revolution means that a company using this thing is significantly more efficient than one that doesn't. So where is that one company that this is true for? And please, spare me the CEOs bullshitting, I'm asking for numbers. Who is going to make a trillion with this next year? Because otherwise OpenAI and Anthropic is dead, and FAANG stocks are going to go down like 30%, taking the market with them.
And I'm still going to run my little local model, which does not mean that AI is going to be bigger business in a few years than selling microwaves. They are also useful, even ubiquitous, but they are not "an industrial revolution" and nobody is claiming "you are missing the microwave revolution" or that "every company is a microwave company".
Of course it won't end up with nothing. Me and you will still use LLMs on our desktops, and make stuff with it.
All I'm saying is that Jen from accounts receivable won't be using LLMs in Excel sheets, because on the one hand it's stupid and on the other hand it's not economical. And if that's true, then the current hype will result in an equivalently sized crater in the US economy.
And I'm saying this especially because the rate of adoption is unsustainable, we can't pay for so many people to use this tech for so little, it either has to do much, much more or be adopted much, much less. I only expect that gap to close, and the fear is that it will close with a lot of force.
Yeah but if Jen's work does not change, then where is the revolution? All previous revolutions moved Jen around, from the fields to the assembly line, then from the assembly line to the office, then from the office to a computer and arguably home.
How does Jen using Excel and using an LLM to calculate something in the books a "revolution"? And where is it happening then? Whose work has gotten "revolutionized"? What is actually going to happen if it hasn't already? Like precisely?
And yeah, cost per token is dropping, but since all progress in the past year at least has increased tokens per query, cost per query is actually going up. And we can't even be sure that the actual cost per token is going down, since OpenAI can set any price at this point, since it is losing money either way. We don't know the cost per token, only the price per token.
Again, how is this different to the invention of the microwave? What makes it "revolution"-scaled, other than the money sunk into it?
Okay, but when? Studies show that Jen's job has not changed yet at all so far. What's needed for it to change?
How specifically?
The combined total yearly revenue of the AI industry is by a generous estimate around five billion. That is less money that Mario Kart 8 made in revenue. And we didn't need to restart nuclear reactors to run Mario Kart 8.
And that revenue is mostly made by OpenAI, mostly by selling to Microsoft, which basically owns it, so one might ask, how does that even count.
.
It gave anyone with half a brain the impression it could do that.
Everyone with a full brain is stuck staring at the chaos while all the half brains make comments like yours.
Indeed I do. It's not capable of doing what you just said. I worked with every major model in a professional capacity and you are objectively wrong. You can certainly smash prompts and context at these things over and over until you get a hallucination that you can hammer into something functional if you have some expertise (borrowed or otherwise) but it cannot do what you claim. I've probably spent more time reviewing LLM generated code than most vibe coders have ever written.