AI Losses Are About To Spiral Out Of Control
When in a hole, just keep digging.
It seems like everyone, including the AI bros, now know that the AI industry is one of the biggest bubbles humanity has ever created. Yet, rather than actually do something about it, like regulate Big Tech, enforce anti-trust laws, crack down on round-trip investments, or have the stewards of our delicate financial systems tell these morons to f**king stop it, we collectively seem to be enabling these idiots and letting this entire circus self-implode. The trouble is, when something is this detached from reality, devoid of logic, and utterly irrational, like the AI bubble, all the typical forces that would cause said self-implosion achieve nothing. They can seemingly just choose to ignore something that would be a fatal blow to any other business/industry and carry on unscathed. It is going to take a mind-altering tsunami of reality to break this cult-like hysteria and pop the bubble. Well, that sucker punch of reality is just over the horizon.
As with all bubbles, this starts with the biblical cash bonfire at its heart.
The poster child for the cash black hole is OpenAI. Despite a reported $5.7 billion revenue in Q1 2026, they posted a non-GAAP loss of $6.9 billion. That means they lost $1.22 for every dollar of revenue they made. To make matters even worse, GAAP stands for ‘Generally Accepted Accounting Principles’, and non-GAAP accounts can exclude certain expenses, such as restructuring costs and certain kinds of depreciation, both of which OpenAI has in spades. So, their actual loss is likely noticeably higher. This is after racking up a $38.5 billion loss last year (read more here).
Oh, and don’t go thinking Anthropic is any better. They have tried to make it look like they are close to profitability, but this isn’t really the case. As Ed Zitron points out, this profit comes from an unsustainable market position and sleight-of-hand accounting that doesn’t reflect their reality (EBITDA profitability).
At this rate of burning cash, OpenAI and Anthropic are on a one-way trip to bankruptcy. In fact, OpenAI is on course to go bankrupt this year or early next year, even with the substantial private investment it receives (read more here).
So, what are they going to do to solve this?
Well, thanks to AI’s insanely poor economies of scale, they can’t really lower their costs. Plus, due to the circular-financing nature of the AI bubble, reducing their costs would also reduce the flow of incoming investment from their Big Tech investors (read more here), and they need that to pay for their losses.
Logically, the only other option is to increase the amount of incoming money. That is why Anthropic and OpenAI have shifted to token-based pricing rather than flat-fee subscriptions and are preparing for IPOs.
Let’s start with the switch to token pricing. We have known for a long time that AI subscriptions were heavily discounted. Famously, OpenAI was losing thousands of dollars per month on its $200 subscriptions. Not only that, but because this was a flat monthly fee, the revenue these customers generated wasn’t proportional to the cost to service them. So let’s say these customers tried to maximise their use of this AI; the losses incurred could spiral. In other words, this sales structure doesn’t give the vendor any real ability to control or predict losses.
That’s why, a few months ago, OpenAI and Anthropic switched to token-based pricing, which meant the price a customer paid was proportional to the cost of serving them. To be very clear, the current token prices are likely still discounted to the point of making losses. It’s just that OpenAI and Anthropic can actually decide how much money they are losing, which potentially allows them to reduce the amount.
However, this has had an enormous impact. It turns out, all the AI bros, the heaviest users of their products, had been using this service at a catastrophically low cost. So even this tiny correction, which is likely still losing them money, has made many AI services so expensive that they are functionally useless. It’s like renting someone a Ferrari for $10 a month — of course, they are going to love it. But as soon as you charge them the actual cost, they might suddenly not like it anymore.
For example, the cost of GitHub’s Copilot AI has increased by 10x to 50x for many users; 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-billion-dollar bill with Anthropic in a single month. In fact, leaders at Uber have said that AI is now more expensive than human workers. Yowch.
Token-based pricing has shone a light on a major flaw in the AI industry. It was only useful to people when sold at a catastrophic loss. As soon as these companies tried to even control how much money they were setting on fire, the costs became too much, destroying the very ‘usefulness’ of these tools.
I genuinely think the people at Anthropic and OpenAI knew they would receive a significant amount of backlash for the switch to token-based pricing. But they had to do it anyway.
These AI labs and the ecosystem they live in have funded themselves through private equity sales and debt. But these sources of funds are beginning to dry up. Venture capital firms are struggling to raise enough capital to invest, their Big Tech backers are nearing the limit of how much they can invest, and the bond market is so saturated with risky AI debt that borrowing costs are getting too high. These sources of cash simply can’t provide them with enough to survive anymore.
So, if selling AI services at a profit renders these tools useless because the cost is too high and their current source of funds is rapidly drying up, how can these AI companies bring in the cash they need to stave off bankruptcy for another year?
Simple. Go public. Let the retail investors who have been forcibly fed AI propaganda for years hand over their hard-earned cash to be burned. How about their pension funds, too, while we are at it? This is the only capital market the AI bubble has yet to fully tap.
There is just one problem. They have colossal, uncontrollable losses on their books, which makes an IPO impossible. Even after the insanity of the SpaceX IPO, they need to show some level of progress towards sustainability and profitability if they want the IPO to go even half as well as SpaceX’s. Let’s not forget, Altman and Amodei do not have the cult-like salesman skills of Elon Musk, so they need something more material to peddle their snake oil.
That is why they had to switch to token-based pricing: it makes it appear like they are making progress by controlling their losses, even if only for a short while. That alone is worth the backlash and worth releasing the biggest secret of this industry, that these models are exponentially more expensive than anyone thought, because it enables them to go public. And wouldn’t you know, that is exactly what they are doing.
However, there is one giant stick in the mud here.
OpenAI lost a ton of market share to Anthropic early this year, and the two have been in a battle for AI customers ever since. But this price jacking and widespread backlash are the perfect environment for both companies to grab huge slices of the market.
So, both Anthropic and OpenAI are considering making huge price cuts to gain more customers. They can do this because the timing of these cuts means they likely won’t be reflected in their IPO financials, as they will be for the previous quarter.
But do you see the problem here?
Yeah, cutting prices is just going to make their already bad money situation even worse. If you think these companies are experiencing horrific losses now, just wait until this strategy takes effect! Going forward with these price cuts is just speedrunning bankruptcy — especially when it isn’t that hard for users to switch to a different AI model. So, even if they can stomach the losses, when they eventually come to put the prices back up, the customers they once gained will just jump ship to whoever is cheaper at that time.
These AI companies are now effectively trapped in a giant trillion-dollar game of chicken. This is the setup for a rapid race to the bottom, followed by a devastating cash crunch. It has the potential to take all the oxygen out of the AI bubble and provide that heavy dose of reality that eventually pops it. That is, unless something utterly unprecedented and drastic happens soon.
That is the issue with bubbles like this. We know it is a bubble, we know it will pop, and we can see many different ways it could pop. But the markets can stay irrational far longer than you might think, so trying to predict when it will all come crumbling down is a fool’s game. All I can say is that the biggest problem the AI industry faces right now is about to get even worse.
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!