The AI bubble is no longer the elephant in the room. It is a demented knife-wielding clown. You can't ignore it, and if you try to, it will be at your peril. The big question has not been if the bubble bursts and the clown attacks, but when. You see, this bubble is now so enormous and so interconnected with our economy and financial systems (read more here) that when it bursts, it will damage everything. So, predicting when this clown will go on its murderous rampage is critical. The trouble is, it is also impossible to make these kinds of predictions. However, over the past week, there have been significant signs that this bubble is already beginning to pop. We may not have to predict anything, because it looks like the collapse has already started.

Meta

Zuckerberg is playing catch-up. Facebook as a platform is at best stagnant and at worst a failing social media site propped up by boomers and AI bots. But, being the capitalist he is, Zuckerberg wants growth! So, for the past few years, he has been jumping on whatever trend might help the company grow, from launching his own digital metaverses to trying to launch his own cryptocurrency to smart glasses and now AI. But to even think about playing with the big boys, Zuckerberg has to sink a ton of money into AI development, and that is the problem.

During the company's last earnings call, Meta announced that its annual AI spend will be $70 billion to $72 billion, up from its previous estimate of $66 billion to $72 billion. Now, that is only a roughly 3% increase (taking the average of both estimates) and still within their initial budget, so you'd think investors would be happy. They weren't.

The very next day, Meta's share price slid a massive 11%! When asked, multiple institutional investors pointed to the fact that Meta's investment in AI was ballooning, but they wouldn't show how it would ever generate a return. Brian Mulberry of Zacks Investment Management even said that "They have to start doing a better job of showing us when that comes back to the balance sheet."

Why does this matter? Well, it shows that investor sentiment on AI, at least outside of the circular financing, has pivoted. They no longer want to dump money into an AI race and value companies purely based on their AI capex. They want to know how AI will generate a return and are pulling out if that question can't be answered. This change of stance from investors is one of the many factors that popped previous bubbles, such as the dot-com bubble.

But what caused this switch? Well, just as with the dot-com bubble, the economy isn't doing so well right now, which means investors are growing more cautious and placing more emphasis on fundamentals than speculation.

Is this a problem? Oh yes. No generative AI company, or generative AI division, is profitable. They all run at gargantuan losses. These LLMs are extremely expensive to build, develop and run. For example, OpenAI loses a ton of money for every one of its top $200 per month users (read more here). There was even a leaked memo that suggested they would need to charge more, like $2,000 per month, to break even (read more here). And, for many reasons, as AI expenditure increases, these models become even less profitable (read more here). So, Meta and every other generative AI driver can't tell investors how this will ever generate a return, because it never will. That is why this investor stance switch and Meta's faltering share price are a major sign that the bubble is about to pop.

Despite this clear sign of investor dissatisfaction, Meta is pushing ahead with plans to increase its AI capex to $600 billion by 2028. This is supposed to enable them to compete with the likes of OpenAI. However, considering Meta only has about $44.5 billion in liquid assets on hand, $62 billion in profits each year and no circular financing buddies, they will have to saddle up hundreds of billions of dollars of debt to reach this goal. This announcement sent Meta stock plummeting another 10%, demonstrating that investors are beginning to pull back from AI hype and speculation.

Circular Fall Off

Meta proves that much of the current AI racket comes from the circular financing that has been going on between OpenAI, Corewave, Nvidia, Microsoft, AMD and Oracle. Basically, these companies have been passing around the same bundle of cash to each other in a circlejerk, buying smaller and smaller pieces of each other each time. No functional growth has happened, but on paper, these companies have been able to substantially increase their value (read more here).

However, even this circular bullshit has reached its limits.

Last week, Nvidia lost a gargantuan $350 billion in market value, equivalent to a 10% drop. Microsoft, AMD and Oracle experienced similar slides. At the same time, Meta underwent its 10% slide, and Palantir dropped by 9%. In other words, companies inside the AI circular financing bubble dropped as much as AI companies outside this circle, and at the same time.

I will explain why I don't believe this is a momentary blip or statistical variation in a minute. But this is huge. It shows that this circular financing hasn't removed the circular AI players from the broader economy or market forces. In other words, this investor stance switch will likely apply to them as much as companies like Meta, which is a huge problem, as OpenAI and data centre operator Corewave are light years away from profitability (read more here for OpenAI and here for data centre profitability) and have no viable route towards it.

In other words, this shows the broader forces that could pop Meta's AI bubble will pop this circular bubble too.

Softbank

So, why don't I believe this is a momentary blip? Why am I reading so much into a 10% drop in stock prices?

Because of Softbank. You could consider them the shadow player in this AI circular financing crap. Microsoft invested a huge, undisclosed amount into Softbank's Vision Fund. This fund has invested heavily in OpenAI, Oracle and Nvidia. Essentially, the lump sum of money being passed around to make this unproductive circular financing setup work came from SoftBank. So while they haven't directly participated in these deals, they have been silently holding onto their shares of these companies, watching this circlejerk prop up the value of their investment.

To give you an idea of just how much money Softbank is making from this, since 2022, which is roughly the same time Softbank started investing in Nvidia, Nvidia's share price has gone up by a factor of 14! In fact, SoftBank's profits have more than doubled this year to $16.6 billion, driven mainly by the increase in its OpenAI shares, which, in turn, have been pushed up by its deals with Nvidia.

However, Softbank's own shares are not doing so well. The company lost around 10% of its value, equivalent to $50 billion, at the same time as the other AI stocks. Again, investors and analysts pointed to investor sentiment towards AI souring as the cause.

Before I explain what Softbank has just done, I need to give some critical context. In 2019, Softbank was in a very similar position and sold its entire stake in Nvidia. But if they had held onto those shares for just two more years, they would have made a whopping $150 billion in profit! They were brutally burnt by failing to weather the storm and publicly decried the decision.

With how much money they are making from the circular financing, and given their past experience of selling out too soon, you'd think they would hold onto their shares at all costs, right? They can ride out a slight dip in share price, but they can't pass on this earth-shattering opportunity again.

Well, it turns out that SoftBank had sold all of its Nvidia stock for $5.83 billion in October, right at the peak of Nvidia's value, and weeks before the AI stock slide.

That is, to put it lightly, pretty fucking suspicious. Nvidia is one of the best-performing companies in this circular financing scheme, only being outdone by OpenAI. If Softbank expects the AI bubble to continue, or for these companies to execute on their promises — from how AI will revolutionise work to the data centres being built — selling this much stock in one go makes no sense at all.

But if Softbank knows something we don't and is aware that this bubble is going to pop imminently, then this makes total sense. They can't offload their investment in OpenAI; it isn't a public company, which means they are locked in. But Nvidia is a publicly traded company, so they can offload their stocks easily without crashing their value too much, particularly if they sell at the peak of their value when demand is sky high, as they did.

Softbank's complete sell-off of Nvidia shows that the puppet master who started this entire circular financing debacle, which is responsible for the vast majority of the AI bubble, is trying to exit it as fast as possible — or at the very least mitigate damage if the bubble bursts. If that isn't a sign that this bubble is about to go, I don't know what is.

Enterprise AI Demand Cool Off

But I think I know what spooked Softbank. Companies are realising AI is absolutely shite.

Geoffrey Hinton, a man often referred to as one of the "Godfathers of AI", recently stated that the only way generative AI companies can reach profitability is if they can replace human workers on a massive scale. He is entirely right, and I will probably write an entire article just on this statement soon.

But all the data suggests that businesses are actually turning their backs on AI rather than adopting it. RBC analysts have found a recent decline in US businesses paying for AI services, driven by a lack of productivity gains and pilot fatigue. They also found that AI adoption has fully stalled. In fact, other analyses have shown that AI adoption in large corporations has dropped this year.

This isn't that surprising, as multiple studies have shown that AI is incapable of augmenting a job, let alone replacing human workers (read more here), explicitly because of "hallucinations" (read more here). What's more, even OpenAI admitted in a recent research paper that you can't reduce, let alone prevent, hallucinations from happening (read more here).

I think this is what has spooked Softbank and many AI investors. The market data and research paint a clear picture: AI, in its current state, is not good enough to be a widely used productivity or job-automation tool. Yet, speculation that AI can automate jobs drove the immense investment in it from the likes of Softbank. Unfortunately, this OpenAI research paper, along with other AI research papers, has shown that the factors preventing AI from being a usable productivity tool or automation tool are fundamental to the system and can't be solved. In other words, AI can never live up to its speculation and, therefore, can never generate a return on its investment.

So, it isn't surprising Softbank ditched Nvidia at the exact same time the demand for enterprise AI started to wane. That was their signal that the fundamental falsehood propping up the entire AI bubble was crumbling, and it was time to get out.

AI Spending Slowdown

Last, but certainly not least, AI spending is expected to slow dramatically in the near future.

Forrester has predicted that a quarter of enterprises are delaying their AI spending plans until 2027, as they try to find ways AI can actually positively impact their bottom lines. Forrester found that only 15% of AI decision-makers could report increases in AI-related income. Note that this is income, not profit, meaning the number of AIs that actually improved their bottom line is even less, as AI is often far more expensive than getting human workers to do the job (read more here).

As such, Forrester summarised with "The disconnect between the inflated promises of AI vendors and the value created for enterprises will force market correction."

"Market correction" is a beautiful euphemism for a devastating bubble burst. The entire reason trillions of dollars have been poured into AI is to automate jobs. But if enterprises force a market correction, that entire justification falls away, and investors riding the AI bubble's coattails will sell up, and speculation will dry up.

Again, this is why I think Softbank sold Nvidia. Like Forrester, I think they can sense a market correction coming for the AI bubble as reality smashes its way through false promises.

Summary

This isn't the pop. These collective slides in AI share value, Softbank selling up, and the falling interest in AI enterprise are more like a precursor. We saw this with the dot-com bubble. Mark Cuban and Goldman Sachs witnessed the obvious signs of a bubble crashing — lack of profits, overinvestment, failed speculation, stock valuation faltering, and a change in investor stance — and sold near its peak, right before it burst, profiting billions of dollars in the process. History doesn't repeat itself, but it rhymes. Consequently, everything we have just discussed shows that the AI bubble has entered its final phase, rather than a burst. The bubble has reached its limit, and the big players are starting to exit, just as they did before. In a way, the AI bubble has already begun to pop and is now on the verge of bursting. We are at the point where the ticking time bomb has reached 0:00, and we are nervously waiting the inevitable explosion.

Thanks for reading! Don't forget to follow me on YouTube, Bluesky, and Instagram, or support me over at Substack.

(Originally published on PlanetEarthAndBeyond.co)

Sources: CNBC, The Telegraph, CIO, BI, The Register, This Is Money, CNBC, Futurism, Gartner