The AI Industry Is Panicking

Is the AI bubble bursting?

Photo by Ksenia Pixelesse on Unsplash

Have you noticed that the AI bros are starting to panic? All of a sudden, they have dramatically U-turned on the notion that AI will take vast swathes of the job market despite that claim being critical to AI’s hype and the entire point of this ‘revolution’, they are desperate to gain US government backing, and they are pushing some of the most overpriced IPOs in history despite all of their companies being closer to a cash bonfire than a functional business. This is peak frenzy behaviour, the kind of stuff you see right before the bubble bursts. It is a thinly veiled, desperate attempt for them to find a stooge of a bagholder and cash out before the entire thing collapses. But why? Well, a lot has changed over the past few weeks, so let me walk you through everything.

The Debt & Bet

Here is a question: how much money do you think has been invested in the AI industry?

According to Reuters, investors and companies had poured around $2 trillion into AI by the end of 2025. But that isn’t quite the whole picture, as the AI industry has not just received cash investments but also racked up a colossal amount of debt. Much of this debt is obfuscated using weird company structures, which means Reuters probably hasn’t taken all of it into account.

J.P. Morgan alone has $1.5 trillion in AI-related debt. AI debt also makes up roughly 15 -20% of corporate bond indices(referring to debt a company issues, not a bank), meaning there is well over $1.5 trillion in AI-related bond debt in the open market, too. In fact, AI is now linked to more debt than banks!

It is difficult to pull together an accurate total figure, but we can be sure that multiple trillions of dollars have been invested in the AI industry with much of that being debt. This debt is also set to balloon, with just a handful of companies poised to rack up another $1 trillion in AI-tied debt over the next few years. So, in the next year or two, AI-related debt alone could easily reach $3 trillion.

This creates an implicit bet.

To provide some rough context, let’s say that the entire AI industry has $3 trillion of bond debt, at a typical market rate of 3% over a ten-year period. Paying the interest and capital each year will total $309 billion.

This means that for the AI industry to service its debt, it needs to generate hundreds of billions of dollars in profit each year.

Even giant monopolies like Google don’t make enough profit to service that much debt. AI can’t just be a novelty industry; it needs to replace human labour on a colossal scale to service this debt. Let’s optimistically assume AI one day reaches a 10% profitability margin, a cost parity with human labour, and the ability to complete most jobs (none of which are currently the case). Well, the average US salary is roughly $66,000, so at a 10% profit, the AI company will make on average $6,600 per year per job it replaces. To generate the $309 billion needed to service their debt, the AI industry will need to replace 46.8 million jobs, equivalent to around 27% of the current number of jobs in the US.

While this is all very rough maths, it highlights the implicit bet created by the debt the AI industry has racked up. To simply not default on this debt, the AI industry has to rapidly displace human labour at a staggering scale, even if we are extremely optimistic about AI’s economics. In other words, the AI industry has made a gargantuan bet that AI will create a jobpocalypse, and if it fails, their entire industry, and much of the financial systems that our modern society depends upon, are on the line.

This is not just my opinion, but that of “AI Godfather” Geoffrey Hinton.

There is just one problem with this bet: it obviously isn’t paying off.

No Jobpocalypse

You may have noticed that, despite a whirlwind of propaganda trying to convince you otherwise, people are not losing their jobs to AI — at least not at a meaningful scale.

Oxford Economics has found that companies “don’t appear to be replacing workers with AI on a significant scale” and instead suggest that they are actually using the AI layoff narrative to cover up their own shortcomings.

Why? Because AI doesn’t actually increase overall productivity. It might increase the productivity of a single task when quality control is disregarded, but this increase doesn’t scale to job, organisational or economy-wide productivity growth. Indeed, Oxford Economics said, “If jobs are being replaced, where’s the productivity surge?” which is apt, given that economy-wide productivity is relatively stagnant at the moment.

More recent studies have supported this theory, such as this one, which surveyed 6,000 CEOs, CFOs and other C-suite executives across various countries, nearly 90% of whom reported that AI has had no impact on employment or productivity over the past three years.

In fact, we are actually seeing the opposite effect.

Companies like Starbucks and Pizza Hut tried to deploy AIs to replace human workers, but performance issues forced them to pull those systems. . Sinch also conducted an AI production paradox study. It surveyed 2,500 AI decision-makers from across multiple countries and industries and found that AI customer communications agents have a 74% rollback or shutdown rate. Customer communications was considered one of the easiest jobs for AI to replace, yet it is failing catastrophically. Indeed, many firms that claimed to replace workers with AI have had to desperately rehire them because it did not work.

But Jobpocalypse Soon?

Okay, so why is AI struggling to replace jobs? Is it because it isn’t being deployed correctly? And will the immense amount of investment being poured into AI solve this problem in the near future?

AI is struggling to replace jobs because it keeps making mistakes. For example, Carnegie Mellon University found that even the best “agentic” AIs fail basic tasks 70% of the time. Another study found that the best current AIs failed 97.5% of realistic real-world freelancing jobs. These studies identified AI errors, or “hallucinations”, as the primary cause of this astonishing failure rate. This is why AI doesn’t actually increase productivity or threaten jobs.

Take coding, for example. The University of Waterloo found that even the best generative AI coders only have a 75% accuracy rate when tasked with very basic coding tasks. Or what about research from Veracode, which found that 45% of AI-generated code contained security flaws? Or the study from Coderabbit, which found that AI-generated code has 70% more bugs than human-written code? A human coder might save a lot of time using AI to write their code, but they will have to spend more time debugging and correcting said code than the time the AI originally saved them. Indeed, METR found in their last conclusive AI coding study that these tools actually decrease coders’ overall productivity because they spend so much more time correcting the AI’s mistakes.

This is not a flaw that can be solved with a different kind of deployment. These systems require extensive human oversight in order to locate and mitigate hallucinations to function correctly. As such, AI inherently can’t replace human labour at any meaningful scale.

But will this always be the case? There is a boatload of money currently being poured into AI to give it more data and computing power to make these AIs more capable and accurate. So, surely something will solve or reduce this problem, right?

Well, no.

OpenAI’s own research found that increasing the computing power behind AI, or providing it with more data, can’t reduce AI “hallucinations” below its current level. In fact, they found no viable way to reduce AI hallucinations. This has been supported by a more recent paper by Vishal Sikka and his son, Varin Sikka, which found that AIs are mathematically incapable of reliability or carrying out computational and agentic tasks beyond a certain complexity. This is because AI isn’t actually intelligent; it is just statistics, and there is a limit to how accurate a statistical model can be.

In other words, AIs are not going to get much more accurate than they currently are, meaning they can’t boost productivity on a wide scale or replace jobs anytime soon. To do that, an entirely different technology is needed, which we simply don’t have.

The Costpocalypse

But, even if AI were able to replace jobs, it wouldn’t actually solve the problem at hand, because over the past few months, the AI costpocalypse has reared its ugly head.

For reasons we will get into in a second, AI companies have recently switched most of their enterprise contracts from a flat subscription to token-based pricing. This effectively means that companies will be charged relatively for the cost ofoperating these AIs, rather than a heavily discounted flat fee.

The fallout of this slight change has been monumental.

The cost of GitHub’s AI has increased five to nine times; Uber blew through its entire 2026 AI budget in just a few months; Microsoft has cancelled its Claude Code licence because it cost too much; and one company racked up a half a billion dollar bill with Anthropic in a single month.

In fact, leaders at Uber have publicly stated that AI is now more expensive than human workers. But they also said that AI usage doesn’t correlate to delivering new features, meaning that, unlike human workers, AI has no real measurable return on investment (ROI).

This costpocalypse is turning many AI bulls into bears, as the cost of AI not only dwarfs that of human workers but also forces them to pay attention to AI’s complete uselessness.

This all heavily suggests that corporate use of AI is set to dramatically drop. When you consider that all the major AI labs and their backers were relying on widespread corporate use to generate them the vast sums of revenue they need to pay their colossal debts, that isn’t good news.

So, even if AI could be improved to the point where it can actually complete tasks effectively, it won’t replace human workers, because it costs more than they do. This is why every AI CEO has suddenly U-turned on their crass, degrading and aggressive jobpocalypse narrative, because you can’t go around saying your AI is going to take people’s jobs if the AI costs significantly more than the workers — it makes you look like a total idiot. But this begs the question: where are these companies going to get those hundreds of billions of dollars to pay their debt?

The Cash Drought

Okay, but first, why did they switch to token-based pricing? Surely they knew such a cost increase wouldn’t do their already bad PR any favours.

Well, truth be told, they had to, because they are running out of cash at an unprecedented rate.

AI companies are chronically unprofitable. Despite OpenAI raising an insane amount of investment over the past few months, some analysts are predicting it will declare bankruptcy in 2027 (read more here); that is how much cash it is bleeding. In fact, OpenAI’s losses are only increasing as the company grows, with its operating margin for 2026 Q1 being a staggering -122%, meaning it lost $1.22 for every dollar of revenue it made. Anthropic has tried to meekly suggest that its books are better, but as Ed Zitron has pointed out, this is more of an accounting trick and market circumstance than actual profit. There is also the context here that both Anthropic and OpenAI are in this position despite them running on heavily discounted compute supplied by their backers like AWS and Microsoft (as per Ed Zitron). This means that all major AI suppliers are losing money hand over fist, even though their operating costs have been significantly artificially lowered.

To ward off bankruptcy, these AI companies need a constant and ever-growing flow of investor cash. Initially, this was provided by selling equity to Big Tech backers and venture capital firms, but these sources of money have begun to dry up. So, they switched to the debt market through corporate bonds, bank loans and private credit. But now these sources are drying up too. Risky AI debt has saturated the corporate bond market so heavily that borrowing costs have skyrocketed to unusable levels.

The flow of actual cash investment into AI companies (not investment in the form of compute vouchers) is slowing down dramatically. If they don’t do something to resolve this issue, their ravenous cash burn will eat them alive from the inside out and bankrupt them.

This is why the industry has turned to token-based pricing, because it can help reduce their losses. OpenAI had to admit a while ago that its top-of-the-line $200 per month subscription was losing a huge amount of money, with some inside the company estimating that if this subscription were sold at cost, it would be more than $2,000 per month. Now that token-based pricing is here, we can see that their claims are relatively accurate. The flat-fee subscriptions that 99.999% of AI users had been paying were aggressively subsidised in an obvious attempt to hook new users. Turning to token-based pricing doesn’t mean AI companies aren’t losing money, but it enables them to control and manage their losses.

That will not only enable them to reduce the cash burn and delay their pending bankruptcy, but it might also make their books look favourable enough to go public. SpaceX will have already gone public by the time this article is published, but OpenAI and Anthropic are poised to follow suit soon after. This will unlock a new pool of money from retail investors and institutional investors like pension funds. But for an IPO to be successful, a company can’t be bleeding cash and on the verge of going bust. Token-based pricing might polish their books just enough for their IPOs to succeed.

The Failed Blitzscale

If this all seems like it was destined to fail, you’d be right. The AI industry has effectively tried to Blitzscale. This is where you use large amounts of investors’ cash to offer a rapidly scalable service at a heavily subsidised rate to soak up as much market share as possible, and once your competition has been driven out of the market (also known as having a monopoly), you can jack up your prices and make a killing. Airbnb and Uber are famous examples of this method.

The AI industry obviously thought that they could out-compete the human job market and replace an incredible number of jobs by selling AI at an astronomical loss. Once this market was disrupted and AI systems became integral to businesses, they could jack up the price and make their bag.

The only problem is the AI is so damn expensive that it has bled their wallets dry before they have enough of a monopoly to be able to squeeze the market and jack up the prices! Not to mention thatAI isn’t good enough to replace human workers, so they were never going to be able to secure enough market share anyway.

The Panic

This is why the AI industry is panicking. Their plan for world domination has utterly failed, and unless they can find a new direction to bring in unprecedented, obscene amounts of revenue, the entire industry will implode under the colossal weight of its own debt.

SpaceX’s, Anthropic’s and OpenAI’s IPOs are a test of this. If a single one goes badly, this bubble will burst, because it is a clear sign that investors aren’t willing to keep feeding this cash-hungry monster. That revelation will stop or slow down the flow of cash into these AI giants, which will kill them.

But even if these IPOs go well, they aren’t out of the woods yet. The amount of cash these IPOs will generate is only enough to keep these companies going for another year or two, maybe three if they are lucky. In that time, they have to somehow find a use for AI that is profitable and will generate hundreds of billions of dollars of profit a year just to prevent their debts from eating them alive. The chances of them doing that are functionally zero. So even if these IPOs go well, the companies are still doomed; they have just delayed the inevitable, and AI bros now know it.

The myth of AI is rapidly dying, and the grifting leaders of the industry and their moronic backers have only just realised they have drunk their own Kool-Aid. Only now do they understand that AI is too inaccurate to increase overall productivity and too expensive to replace human workers, and that the vast amount of debt they have racked up on the premise that AI will replace the job market is now threatening to crash down on their heads hard. Expect more panicking, more U-turns, more pleas for government backing, more desperate new directions, and more delusional outbursts from these people, because the AI bros are now desperate to keep this gravy train going at any cost. But sadly, the question isn’t whether this entire thing will collapse, but when.

---

Thanks for reading! Everything expressed in this article is my opinion, and should not be taken as financial advice or accusations. Don’t forget to check out my YouTube channel for more from me, or Subscribe. Oh, and don’t forget to hit the share button below to get the word out!