Link: https://jacobin.com/2026/10/artificial-intelligence-investors-finance-bubble

If you want to know whether artificial intelligence is a gigantic economic bubble, don’t just take pundits’ word for it. Major investors are now moving their money to try to protect themselves from the AI bubble popping.

For an industry that we’re told is central to our collective future, the artificial intelligence (AI) industry increasingly seems to be built on a combination of artificially inflated hopes and bad accounting.

The industry’s critics have been raising concerns about its finances for a while: the fact that most of its flagship firms are grossly unprofitable, that they’ve had to jack up their prices to try break even (leading their business customers to pump the brakes on using their products), or that many of the eye-popping investment announcements that are meant to show demand for AI is strong have actually been circular financing — meaning different AI firms cycling money to and from each other.

But none of it has seemed to slow the industry’s boom or the relentless hype machine behind it. Until now, anyway. It used to be that you mostly heard concerns about the industry’s underlying economics in newspapers, blogs, social media posts, and podcasts, and that those concerns went completely unreflected in the markets. But these concerns seem now to have finally bled over into investors’ thinking.

Imagine when that bubble pops and AI goes back down to “tool” level, and everyone goes back to work with a brand new dislike for work.

One of the "leaders" of the industry posted an $80 billion loss this year. Wall Street is not nervous enough about AI.

Not just wall street, some of the pro-slop people in my work are getting increasingly desperate and unhinged.

They formed an 'AI native team' to develop 'groundbreaking new applications' more rapidly than our dumb old software development teams are doing even with AI tools available.

The logic: Obviously software development skills just make the developers too opinionated and they stand in the way of AI.

The problem: After 6 months, every single vibed up slop project has failed to get through QA, and the developers can't get AI to address the issues QA is reporting (similar stuff as people have been reporting with the slop-clones of Office and Photoshop, UI elements that fail to do anything despite looking like they should, the software doing one thing when it was asked to do another).

The proposed solution emailed out: QA obviously is a problem getting in the way of the slop the customers clearly want to pay for, so we should consider skipping QA and working with the customers instead to fix it. I'm hoping this is the line that executives won't cross, most of them have been chewed out by customers enough over more mundane fuckups that they could see what a shitstorm it would be to skip the only thing protecting us from shit.

QA obviously is a problem

Oh lord...

Even its tool status only exists because these companies subsidize the cost. In the event any survive the pop, realistic costs would price it out of being a frequent tool for typical workers.

A lot of people are getting addicted, though. I have the sinking feeling they're trying to keep this going for as long as they can so there's enough addicts who will then sink their savings to get their fix once the bubble pops.

Why not a local model on owned hardware?

And if you know how few of the real problems QA normally detects before a product is being shipped to the customer you know the shit show that those products will cause.

Hand me the popcorn, that show has a lot of potential.

The problem: After 6 months, every single vibed up slop project has failed to get through QA, and the developers can’t get AI to address the issues QA is reporting (similar stuff as people have been reporting with the slop-clones of Office and Photoshop, UI elements that fail to do anything despite looking like they should, the software doing one thing when it was asked to do another).

• Is AI imploding within itself through poisoning and no new content?
• Are they putting in their own guardrails to take away the really good innovations because of greed?
• AI was never as good as they claimed, even the super advanced models?
• All of the above?

I'm more curious than anything. It seems to be getting worse for the free models.

working with the customers instead to fix it

Ow, my sides!

I don't think businesses want the overhead, but maybe some will. The capabilities will be much lower, though. The subsidized costs are high because of the hardware and power needed to generate the content it does. Companies willing to upgrade to comparable hardware and oversight would probably be just as willing to pay a higher subscription. Also, I would hope any company considering it would be able to see by then that the results aren't worth the costs.

Because its more expensive than openAI giving 90% discounts.

Only for the hyper scale centralized compute llms. Real machine learning tools predate and will persist past the bubble pop, but data center cloud based llms will go the way of the dodo one way or the other. Which is a disaster for the big companies that want to sell you compute as a utility, but really totally fine for most everyone else.

Win win?

I don't see how that's a win for anyone except the assholes peddling AI.

I agree with the general story but that example is not a great one, 90% of it is an accounting adjustment as a result of their share price going up. This would actually happen in a case where a company was being successful. What happens is the startup gets say $100m in cash in return for 1000 shares, valued at $100 each. Then later when the startup is doing well their share price is now worth $500 each, they still owe the same number of shares, but the value of the shares they owe is now $500m, so on paper that is a $400m loss. They don't actually pay or owe that in cash, they just owe x number of shares to the creditors, which they have to hand over at some point. So that is not actually a direct problem with the business model.

But otherwise in general yes it's a house of cards

I think technology will catch up and AI models will improve to the point where we have whatever ChatGPT is using at the time of this comment running locally on regular hardware.

That’s not saying the industry will die out of course, just that homegrown will be a real option.

But that’s not solving any real problems.

Corporations pretend to the public that their business is providing services when in reality their business is making profit. Corporations are just starting to realise AI research has pulled a similar switch on them, it pretended to be in the business of making profit but actually its business was something else, something about switching the means of production.

Might be an idea to start looking at vacancies in other companies. Just in case yours goes nuclear on the stupid.

CEOs and Boards will constantly fail on the model that demand, money and technology will scale indefinitely.

That's not the case with anthropic, though. They are simply burning money while not making a profit.

The proposed solution emailed out: QA obviously is a problem getting in the way of the slop the customers clearly want to pay for, so we should consider skipping QA and working with the customers instead to fix it.

Narrator: the people who agreed to this never interacted with a single customer.

Number of AI companies making a profit: 0

(No, NVIDIA doesn’t count)

And none of those are "tool" level for the average company, so it's not what's being discussed.

QA obviously is a problem getting in the way

Tbf, QA being seen as a problem slowing things down is nothing new in development. AI just gives it a new impulse.

The same thing is happening in my work. Just as we got the higher ups convinced that yes, having QA is very valuable, we get called the bottleneck again because dev is supposedly speeding up due to AI (according to dubious metrics).

The financial inbreeding didn’t trigger any alarms?

Well, so far, it's the "I've bet my entire career on AI" leader who has not yet gotten traction. He had a large benefit of the doubt but I feel like leadership is getting more and more skeptical about him and his band of merry sloppers.

As someone who went through the dot.com bust at a very unfortunate time in my career, I'm always looking for the escape hatch.

But I feel like the prevaling opinion in the leadership guy is increasingly dubious of this guy, so I'm hopeful sanity prevails.

I dread to find out what the new thing will be that will distract everyone from the actual problems. Brain-computer-interfaces?

I'd say it's never been as good as Musk and Altman especially make it sound.

If you drill into the nuance of their proof points that their companies bother to share, it's generally a bit more clear that things aren't exactly 'asterisk free' even then. Most recently I read how one of them proudly proclaimed they were 100% manual code free, but for the task described they still had 200 developers, which is significantly more developers than I have traditionally seen. To get things vaguely working, you need to be relentless in providing a super comprehensive set of requirements and criteria that are impeccable at fully capturing what is needed and could reasonably go wrong. Also a fanatical amount of detail to audit how the solution was reached and correct missteps, like when it decides to delete a failing test case instead of fixing the correlated problem. The 'purist' approach is, in my opinion, worse for the sake of just signaling how 'AI-friendly' you are.

I can at least buy that from a pragmatic perspective, the codegen can accelerate things if used under a more balanced approach, being able to be in the code and grasp in detail where it might have gone wrong and correct it before it gets lost in a mess of garbage. It will do things that a human might skip that they shouldn't, like hooking various unlikely errors and providing a nicer treatment. It is often useful to get a general idea of a good choice of library to do something and how it fits in your code, though you still have to beware that it may invent details that would be nice and convenient, but are wrong.

The problem is that narrative is totally not what they need to justify the spend. They needed to take randos and prompt up things without full understanding. They needed AI codegen to be really valued on the same level as a full headcount. This trickles down to little folk like the team lead in my situation, where he thinks this is his only ticket to the 'big leagues' if he manages to drive a success story for 'AI native'. He doesn't understand any of it and his team doesn't understand any of it and so of course the AI can't bring what actually understanding the scenario would.

Even more rough for that guy: he imagines a world of slopping up code and then selling the result, without ever having unique skill or insight. Assuming that can work, then why would any company buy your slop instead of just slopping it up themselves?

I have found AI to be great as an extra pair of eyes, organizing, and finding obscure sources. So it's a great assistant, not creator. You have to know what you're doing.

It's like sending a fresh set of very smart, right out of school, employees to do the job. They have been trained to do everything you want them to do and more, but guiding them takes expertise. These recruits can't do it on their own, they need a leader.

Are the companies understanding this, or is greed blinding them?

The specific example you used, of Anthropic's $80 billion loss, does match the explanation, though. It should be closer to $8 billion before the accounting adjustment, which is still a loss, but the other $72 billion is just numbers changing in a spreadsheet, not actual cash going one way or another.

My conspiracy theory is that investors are letting the air out of the bag so that democrats will win and give everyone bail outs.

I look forward to LLM brained techbros having their paychecks be left behind

One early example I heard about slop code being bad was that if you have a process that is repeated, an LLM will not write a function that is broadly scoped so it can handle a significant number of related use cases, it will write different code for each instance, so when that process fails you have no idea which version is responsible. Seems like the kind of thing QA would flag.

Damn, I wish I could see the results of this inevitable shit storm.

https://isaiprofitable.com/

Looks like selling razor blades is profitable while the others are still trying to improve their brand of razor.

This is what happens when investment is done on possibility rather than reality.

I am now wondering how would it be nicer to have a local scheduled solution, an LLM as an office printer. Slow moving common queue that makes people get more conservative in their LLM usage, plan it's rent for themselves, send inputs to have outputs later that day. Some inconvinience to tone down hardware costs and establish a more mindful consumption.

It could be fun to run experiments with that as one of test groups, and others unlimited or limited in some way, to compare how people adapt and how it affects workflows.

Heya colleague!

If QA hadn’t been replaced with a chatbot, that is.

Well, they're right.

"If we stop testing right now, we'd have very few cases, if any," Trump asserted, "Without testing, or weak testing, we would be showing almost no cases."

I think with that metaphor it's selling steel, and what everyone's trying to desperately forget is that revenue is peaking and the steel they spent so much buying is, by design, disposable.

Moving QA to the customer base is straight Microsoft.

The red flags have been waving, alarms have been going off, everyone with two braincells to rub together can see this is a ponzi-like scheme of circular financing paired with extreme market concentration.

Wallstreet doesn't care. This level of incompetence considering almost everyone with retirement is tied to this is just too much to believe. Firing everyone in charge and preventing them from ever working in finance or even a leadership position for the rest of their lives should be the absolute minimum of their punishment for this complicity. If they spoke out early and took action, we can exempt them. Otherwise it is time to clean house.

Even that chart is unintentionally misleadingly optimistic for AI, since it's showing nearly a trillion dollars in revenue. Nearly all of that, however, is from hardware suppliers like Nvidia, Micron and AMD. That number should be revenue for the final AI services only and hardware suppliers should be a different category.

You can't claim digging for gold in a gold rush is a profitable profession by counting the money you pay for a shovel as income.

That isn't really compatible with this timeline because these companies are spending trillions of dollars building data centers, which serendipitously is extinguishing any hope of buying local AI hardware by cornering the market for GPUs, RAM, CPUs and both solid state and platter hard drives.

The implication of this all is that we will rent compute and run nothing locally. If hyperscalers succeed, that is the only way they make their money back. And if we will be able to afford local AI hardware, that also means they must fail.