The thing is AI is expensive, and to be useful needs to be trustworthy. Just seems like that in the end it will need to be a pay service. Maybe quite expensive.
post
Who did not know this already?
The real issue is that nobody seems to care... I mean, the biggest IPO in history just took place, valuing SpaceX (which was in reality a tiny bit of SpaceX, entirely composed of Grok) at almost 2 TRILLION dollars after they showed a 5 BILLION loss in the last quarter
Reality is for suckers I guess
Don't you worry about burning venture capital, let me worry about blank!
https://www.youtube.com/watch?v=-6qIYgwebhM
Blank! Blank!! Your not looking at the big picture.
I expect that many companies will realize that an LLM is a tool that can help with certain tasks, but cannot fully replace most workers because there is a ton of context in people's jobs that cannot be condensed into language. Similarly, many companies are now realizing with token-based pricing that frontier LLMs are not cost effective in many applications. You shouldn't have a natural gas fired data center in Tennessee running a model to proofread your emails. You can have a local model do so MUCH more cheaply. This will leave the speculative data center companies holding the bag on a lot of hardware and capacity that is not actually needed.
When training ai, I was always confused that they never gave a sht about tokens, like that was never something you picked any model over another for or reviewed, it often says to ignore the length but make note of how long it took, tbh that could be because of reasoning text not shown to the user
That's interesting to hear. I wish efficiency had been considered from the start. It seems like there has been a ton of waste. They should only do the calculations needed to achieve the user objective. But they were prioritizing market share over efficiency. Maybe they thought they could afford another couple years of subsidizing wasteful use to build market share, but it hasn't turned out that way.
The only time character or word count was even looked at was when it was specified in the prompt. Like if someone asked for 900-1000 words you would check to see if it fit as one of your tasks which was odd since I literally used word counter sites making that process something that could be easily automated.
Losing BILLIONS a Year? That sounds like TRILLION Dollar Status!
-Wall Street!
"Actually, this is good news, as the amount of money lost shows how much money they have gotten. This means they are very good at getting money."
Those already dependant or addicted, will pay.
I disagree; they've been gorging on the cheap slop buffet for so long, when they realize it costs real money they're gonna revolt - ESPECIALLY WHEN THEY REALIZE IT'S STILL FULL OF HALLUCINATIONS lol.
Evidence:
I don't think there are enough deluded to represent a viable market for them. sure, there are some people who are woefully dependent on the shit, but they need a substantial chunk of the population to be enamored and it's not there. if anything what I'm seeing is that people are revolted by it more and more each day.
All aboard the hype train 🚂. Next stop hypeville. Final destination... Who gives a shit.
I always thank the LLM and ask if it wants a cup of tea. Is this my fault?
Yeah we didn't need the documents we already knew that but I suppose it's nice to have confirmation.
The thing about AI is that the bubble has very little to do with the capability of the technology and everything to do with the overvaluation of the company's developing it.
Clearly they're tokenmaxing. They, as a collective, deserve a raise and a promotion.

Hate on openai all you want but 4xing your revenue over a year is no small feat. Only 4 companies have gone from $1b to $10b in 3 years, Google, Uber, Cheniere and moderna
Not as bad as I thought honestly. Looks like they're making a profit on inference, it's just training the models is costing them a shit ton in R&D.
If we hit a plateau and training new models isn't worth it and they scale back there R&D the business could be profitable. Not enough to justify there absurd evaluation, but not a money pit that some people in this thread would have you believe.
It's not a small feat, but they also 4x'd their expenses, which made them lose significantly more. Long term as you mentioned, if they could entirely drop their R&D, which they'll never get to $0, but if they did, they'd still be almost -$2 billion in profit. Business modes can change to help accommodate that at that point theoretically though.
I just don't see them ever getting there. How many years can you lose $20 billion and stay solvent? They'll raise prices like everyone, but they may lose customers offsetting the gains made, or even if they get more, operating costs will go up too. With all of the DCs being built, I also don't see R&D going down anytime soon either.
How many years can you lose $20 billion and stay solvent?
As long as investors keep pumping money into it. Uber lost billions a year until relatively recently, and they didn't have nearly the same queue of investors ready to pour money into them at an insane markup. You underestimate the tolerance for silicon valley vcs to take in years of loss as long as the companies growing.
With all of the DCs being built, I also don't see R&D going down anytime soon either.
Wouldn't more data centers reduce there cost? More data centers means more capacity and more competition pushing the price down.
So a quick search shows that over the last 10 years, Uber was down a total of around 30 billion before turning profitable 3 years ago. This OpenAI report shows a 20 billion loss just this past year. They are surely different scales, but that's a lot more billions lost.
As far as more DCs costing more, well you have to buy for the structures, pay for all of the hardware, and then pay to run the hardware. The more DCs you have the more that's going to cost. I don't know how that affects capacity or if it's more for model training. However, I don't think there's a snowballs chance that prices are going down. Currently they're going up across the board, and seeing how much they are hemmoraging, they can't afford to go down. These first few years have entirely been about marketshare to build a client base and drive out competition. I can't think of any service I have ever used that ever dropped in price.
Maybe I'm wrong. We can only speculate at the moment and this wave is fairly unprecedented. I personally hope they all crash and burn. It's a blight on humanity and the environment. So I really hope I'm right and these companies go under leaving a bunch of bag holders that invested. I want it to hurt for everyone involved.
It is a small feat when you're already losing money.
Not really, tons of companies in silicon valley burn through billions in cash and never even reach $100M in revenue, much less 4x on billions in revenue, that's gotta be single digits in the amount of companies.
Snowflake in there prime when they had the largest IPO ever only got 50% yoy growth after $1b in revenue. 4x is insane
The problem is though that inference by itself is going to end up a low margin utility service that there will be loads of players offering, they'll never recoup their costs that way.
The only path to profitability I can imagine is to have a model that is vastly superior to what people can get elsewhere that they can somehow lock people into using and then charge them well over the cost of inference alone. None of which looks likely to happen.
Fundamentally models based on an LLM structure are never going to be remotely performant with current technology.
Part of the problem is that as you increase parameters, in some parts of the model the amount of digital processing time is increased close to or more likely above and beyond a magnitude. Absolute best case is somewhere around N^2 per sequence of layers where N is the number of neurons.
additionally floating point multiplication is involved which is like turning two numbers into their own matrix multiplication operation.
Agentic is really more like taking a crappy slot machine that does multiple rolls in one go to improve the odds of a more desirable outcome.
So to simplify the problem, this is like a bunch of stupid monkeys rubbing two rocks together hard enough to create enough friction to make it seem like they have discovered fire.
Taking that analogy a step further, so much effort is being put into rubbing rocks together faster that other important issues like food and shelter are being ignored.
Wasn't this obvious to everyone? And if someone thinks they won't start squishing every penny from ad sources in few years is idiot
How can you loss any money when you don't have any to start with?
It’s simple; you lose other people’s money
Ok, so who is making money off those losses?
According to this, Nvidia https://isaiprofitable.com/
1 comment