cross-posted from: https://slrpnk.net/post/38083497
https://www.axios.com/2026/04/26/ai-cost-human-workers Uber’s chief technology officer already blew through his full 2026 AI budget due to token costs, according to The Information.
Lol. Lmao even
And you don’t think this is at least some of those CEOs’ point?? Now they can rehire people for even less than they paid them before
Meta was spending ~300 MILLION dollars per MONTH on tokens around a month ago.
I knkw they are using Claude, so the price estomation of 5$ per 1M tokens is kind of true, more or less.
Tell me how is this sustainable in the long term?
In a 30-day period, total employee usage on the dashboard exceeded 60 trillion tokens, and the highest-ranked individual user averaged 281 billion tokens. Using the least expensive version of Claude Opus 4.6, which costs $5 for every million tokens, that one user alone could have cost Meta more than $1.4 million.
https://fortune.com/2026/04/09/meta-killed-employee-ai-token-dashboard/
Using the least expensive model of Claude
Meta makes LLMs. They’ve made multiple, some of them are even tailored for coding.
Why are they using Anthropic’s AI instead of their own?
I can only imagine there’s some sort of ouroboros slop machine money laundering cycle going on or something
Because their LLMs suck, and they still want their devs to be productive, while watching their usage to train the internal models.
They give you a choice, and no one is using the internal models.
~300 MILLION dollars per MONTH Tell me how is this sustainable in the long term?
Every single rich SOB on earth is pooling their hedged bets together. Since 2022 Meta has been issuing bonds directly to investors. They announced another 30 Billion in bonds at the end of 2025.
People need to realize that the AI bubble will not pop by accident. It will pop as a direct result of the Epstien-Shareholder class cashing out and leaving retail investors hodling the bag.
I don’t think it’s a problem for them at all. They lose significantly less than the workers they ‘replace’ with AI. Relatively speaking, they come out on top by a huge margin. And that seems to be the main point.
Of course, there’s also the question of how well that AI will perform compared to workers it supposedly replaced. So far it doesn’t seem to be going well. Perhaps it doesn’t need to. Who knows.
https://mstdn.social/@hkrn/116589985138352696
A decent article about enterprise depending on AI subscriptions and a discussion on hackernews.
I don’t really get what the punchline is. It’s not sustainable seems to be it, but it doesn’t seem likely to be destroying OpenAI or Anthropic etc, which is what I want to happen.
That’s why it’s titled “It’s a ticking time bomb for enterprise” not necessarily for AI companies.
Hmm. Doesn’t seem like it’ll be good news for the average person afterwards then. But perhaps when the bomb goes off computer (ram) prices can go down(?).
If the bomb goes off AI companys’ enterprise customers will die like flies. That’s not good for business. If they scaled up too fast, AI companies can’t get a return on their investments and die with their clients.
Depends though, because companies like Google and Meta are still ad platform companies first when looking at profits.
Not sure what the most likely, favourable outcome is supposed to look like atp.
They’ll be fine. Mass layoffs depress wages.
At least with human labour companies have control. They withhold raises and bat down wages and the people tend to have to take it.
With AI you get suckered in and all of a sudden have a subscription or something similar. Then next year the costs go up. You’ve already tailored your workflow to use AI, you can’t just not pay the higher bill. Next year, costs go up again. What are you going to do? Not pay your bill and lose your entire workforce all at once? Now the company is being held hostage by the AI companies.
For anyone who has worked for a large company that was around for the IBM mainframe days, it’s hilarious watching them try to remove the mainframe from their tech stack, while simultaneously becoming dependent on AI.
Any service you depend on in the modern world, you need to be asking yourself ‘what will I do when this enshitifies itself’. And it is a when.
Short termism on all sides is destroying our society.
Tbf, open source software is significantly more resistant to enchittification
Tbh, this is just SaaS in general
Workers can and should do this too.
Union strong, babbyyy
Of course they are too dumb to expect this outcome, but it’s so fucking obvious this would happen. OpenAI is still burning billions each years, just wait and see how much it will cost when they need to make a profit. (They have already cut down on the free version)
I mean, it’s the same freaking market strategy that’s been used for over a decade in tech and nobody seems to realize they’re inside a slowly boiling water…
I spent an hour today on a webinar about how to optimize token usage with more than 500 other participants from the company I work for. Because billing is about to have its come-to-Jesus moment.
Cave man mode. Save much token. Very easy. Get straight to point. Missing semicolon line 182.
As in “Jesus, it’s costing how much??!!?”
Anyone else hate the structure and style of Axios reporting? Those bullet points like structure?
this is the part that really baffles me, though I guess I can’t expect much from the class that seems to be literally incapable of foresight.
Like say AI actually does get there. Some company is able to crack human consciousness and we get truly intelligent AI, good enough to replace your workforce with… that AI isn’t going to be free. It’s going to be a monthly fee for your business. and then when you’re fully reliant on AI, when skilled labor is simply too difficult to cultivate any further, it’s going to inevitably enshittify. These corpos will end up slaves to the AI company.
It’s like they fundamentally cannot see that they are giving up control. They’re so fucking stupid.
If it was good, the AI companies wouldn’t sell its use. They’d be using it to write all kinds of new cool software, dominate financial markets, etc.
Eh. It’s par for the course. 20 years ago, at the height of the frenzy of outsourcing things to China, I remember saying that this will just result in US companies creating their own competition. Anyone with a brain could see that Chinese companies weren’t going to be willing to serve as second-fiddle to their US masters. The idea that you could keep design and management, while sending production overseas, and that you can keep that arrangement stable long-term? Pure fantasy. Of course a country isn’t going to be content just doing the grunt work. They want the highly paying design and management jobs, not just the menial labor ones.
Well, it’s a good thing China doesn’t make cheap LLMs that you can run on local hardware…
This is what happens when mfkers don’t read any Marx.
American capitalism would be doing way better at profit maximizing if people in positions of power across the private sector had studied Marxian economics. Giving away your means of production as a software business would be an obvious no-no.
This is the shit that always blows my mind, to be honest.
You want to maximize profits? Train your workers, put them through school, invest in your workforce.
All the companies that just hire bodies and forget the most important part that sets a human apart from the rest of the animal kingdom astonish me. You get more value per person if you work towards increasing each person’s value instead of increasing the value of persons.
A lot of AI contracts and services were priced assuming literally everyone was going to adopt it so the MLMs all tried to undercut each other to get market share.
Turns out, not everyone wants it and they undercharged significantly so the snake oil salesmen are jacking up their prices to try and stay viable. Leaving a companies that early shrunk their workforce holding the bag paying more in expenses and lower productivity because they bought in to the bubble without even checking how factual claims were.
The cost has also shot up because a lot of the new frameworks are much more token heavy than the old ones.
So the original free plan might have made sense when people were only typing little questions into it, and using a handful of tokens, but is no longer cost-effective with things like modern agent pipelines constantly throwing tens of thousands of tokens at the service.
I tried running a little locally hosted agent thing on my computer the other day, and it was feeding a hundred thousand tokens at the model every few minutes, because it was keeping all the files in context. Sure, it hit the cache a lot, and so the effective cost would be less, but it’s still a lot more token usage than me poking the model with inane questions.
I’ve been kind of forced to use AI for my work, and since I had an unlimited access to whatever model I needed, I figured “why not try”. I was able to find my workflow that works, use it to explain the architecture of the code surrounding my bugfix/feature (I work in gamedev, on a project I’m only a contractor for, so I don’t know the codebase too well), make me a documentation, then draft a plan how to implement it. Implement it, then I just look through it for ideas, combine it with my domain knowlege to see if it missed any obvious things or solutions (which it usually does in a larger projects), and then build my own solution from scratch based on what I know, and what it suggested.
The part where it explains the architecture and data/systems flow is invaluable and it does make ot faster than I could have parsed through unknown code, while being verifiable enough to be trustworthy. It’s a good kickstarting process. Do I need it? Not really, but it would take me longer.
But. It eats tokens like hell. My average monthly token usage is around 800m tokens.
I’ve been told I’m not using the AI enough in my workflow, because I write my PR comments and don’t use AI for code reviews.
I’m seriously considering just leaving IT altogether. It’s just eroding my reverse engineering / codebase orientation skill, while I’m replacing it with something that costs the same amount as my sallary for doing it slightly faster/more easier, and the price will only get a lot worse. I don’t get it. How can’t they see it?? How can they look at “Oh, he’s costing us 2000$ a month in AI use [at current prices], but solves 3 instead of 2 tasks a week, while slowly loosing vital skills”, and say “that’s worth it! But he could use it more.”
I hate it. Fuck managment.
The whole AI scam has always been the techbro CEOs snatching the money from the non-techbro CEOs. Me or you as a private person was never the direct aim. . . . General artificial intelligence will not come. We don’t even have artificial intelligence yet. The terms artificial and intelligence contradict themselves. It’s an oxymoron. It’s still just machines that learn shit without body and soul…
IDK. I get really uneasy about claims that a computer or AI can never be intelligent or self-aware. Sure, it’s “just” circuits, but your brain is “just” cells passing information between each other. An individual cell is no more intelligent or self-aware than an individual transistor is. It’s deeply unscientific to believe there is some magic voodoo involved in biology that can’t be reproduced in a machine.
I have never argued that it is never possible, just like time travel or teleportation is technically possible given a radical breakthrough in our understanding of how the universe works. But as of right now, AGI/true intelligence/etc. is still purely science fiction and LLMs are not even remotely close (and are a complete deadend towards that goal).
Except it’s not comparable because we literally have general/true intelligence right now… it’s just not artificial. There is no radical breakthrough needed. We know it’s possible. We just don’t know how.
Teleportation is not possible as far as we know, and backwards time travel isn’t either. That’s not comparable to true intelligence, which already exists, just not artificially.
Exactly. We are beings of atoms and matter. If you want to believe in some immaterial soul, fine. But if you’re a materialist, then everything we are is atoms. We know atoms in one configuration can produce true intelligence. And there are likely many possible arrangements of atoms that can reproduce this effect. And since artificial minds are not subject to most of the constraints of biological minds, an artificial superhuman intelligence should be possible. Hell, even if biology was the only way to make it possible, you could always build an artificial biological brain and just make it a lot bigger than a human one. Even if human neurology really is the limit of what this universe allows for in terms of intelligence, we could best it by just making a bigger one.
Not going to fight about it. Sorry gotta go, to replace some transistors in the back of my head.
I think at some point we might get better and better at simulating intelligence, and maybe even building actual intelligence in machines, but we can agree that LLMs is closer to your phone suggesting the next word than it is to your brain. If we ever get general AI it can’t be based off LLMs so the current wave of speculation is little more than hype to drive profits up and distract from the problems that already arose.
They knew. They always knew. But you gotta bonus yourself amirite?
My boss is getting annoying with this shit.
I’m stuck on x issue
Have you tried putting x into ai?
Yes I’ve been doing that for two days.
Really? Huh. It should work.
K well no it doesn’t.
Just copy the error and put it into Claude
Yeah…I’ve been…doing that.
It should just give you the solution
Phase one: Fuck around 🤡
Phase two: Profit! 🤑
Phase three: Find out. 😱It’s management who fucked up, but let’s blame it on software because what we really care about is hating the software.
Did they seriously not see it coming? Lol.
Most management types can’t see beyond their next bonus.
That goes double for any publicly traded company.
Utterly delicious.
