Everyone who isn’t a tech bro weirdo can see that the AI industry exists in a bubble. The big question so far hasn’t been “if” there is a bubble, but how large it is, when it will pop, and who it will affect. Sadly, because no one has a crystal ball, we can never hope to figure out when shit will truly hit the fan. But we have also struggled to pin down the size of this bubble, as the reality of AI economics has become muddied by Big Tech propaganda. Likewise, these corporations have propagandised their finances to the point of unreality, making it nearly impossible to figure out who is actually connected to the bubble. However, recent analysis has cut through this cloud of doubt, and reality is far more terrifying than we thought possible. This isn’t a bubble — it is a financial nuclear bomb.
Let’s start with the truly gut-flipping findings of Julien Garran.
Garran is an analyst for MacroStrategy Partnership, an independent research firm that advises 220 institutional clients. He also previously led UBS’s commodities strategy team. Needless to say, he understands his subject material. But his recent analysis discovered that the AI bubble was 17 times the size of the dot-com bubble and four times larger than the 2008 housing bubble!
How did he arrive at that figure?
Well, he calculated it in a very similar way to how I did in a previous article (read here), but with far, far better economic analysis.
Like me, Garran pointed to the efficient compute frontier and the diminishing returns of AI, causing costs to skyrocket and progress to stagnate. This is why ChatGPT-3 cost $50 million, ChatGPT-4 cost $500 million, and ChatGPT-5 cost $5 billion, yet the improvement between these generations has been so slim it is almost unnoticeable. This wouldn’t be a problem if these AIs had reached a point of usefulness, but they haven’t. Like me, Garran points to the plethora of studies that prove AI deployments offer practically no benefit, aren’t profitable, and actually make companies less efficient. As such, the AI industry and the gargantuan investment behind it are a dead end, creating a bubble.
But, unlike me, Garran used the economic analysis pioneered by economist Knut Wicksell to establish the size of this investing discrepancy, the 2008 bubble, and the dot-com bubble, and that is how he learned that the AI bubble is currently four times the size of the 2008 bubble at its peak!
I feel I need to put that piece of information into context. The housing market crash of 2008 caused 27 million people to lose their jobs worldwide, ten million people in the US alone to lose their homes, and created a catastrophic spike in suicides in the US, with some studies suggesting nearly 5,000 suicides in 2009 alone were caused by the crash. It crippled the global economy for years and kick-started a wave of extreme and degrading right-wing politics across much of the Western world. It is arguably the most impactful global event of the past 50 years.
Yet, somehow, the AI bubble is already four times as big and still growing! Just take a moment to let that sink in.
And here’s the thing: the 2008 crash was so devastating because of debt. It is how the bubble grew, and it is what tied the bubble to the rest of the market. Debt inherently ties other financial organisations into the bubble, but debt is also sold off as an investment for pension funds, major banks (including national banks), governments, and major corporations. Then, there is the fact that the loan market is highly interconnected and needs a reliable stream of corporations and people taking out debt to stay afloat. If a major player in this market suddenly stops taking out debt, the entire loan and debt marketplace can crash. Sadly, debt also plays a significant role in expanding our economy, so this crashes, too.
So, when the 2008 bubble burst, it first hit the financial institutions holding the debt that grew it, such as the Lehman Brothers bank. Afterwards, major financial institutions, governments and corporations that had bought much of this debt were dealt a huge blow. Critical financial institutions such as national banks had to step in to try and solve this problem, causing them to suffer greatly too. Then, the global debt/loan market suffered greatly, sending borrowing costs skyrocketing and stalling economies worldwide.
The main question is, is the AI bubble as connected to the outside world?
It seems AI companies have been growing through equity financing (selling shares of the company) rather than debt financing (borrowing money). Furthermore, investment seems to be incestuous to the tech industry; for example, OpenAI’s major investors are Microsoft and Nvidia. As such, many believed that the AI bubble is actually relatively isolated from the rest of the economy and therefore might not have as significant an impact on the wider economy when it bursts.
But sadly, we now know that isn’t true, because, as seen in 2008, the AI bubble is connected to the rest of the economy through debt.
While equity investment in AI has ballooned, it hasn’t been enough to cover the industry’s capital expense and operational losses. Naturally, AI “hyperscalers” like Meta and OpenAI are being forced to turn to debt to keep themselves in the AI race. But they know that if they are publicly seen raising billions of dollars in debt rather than equity, then the entire bubble might collapse around them. Debt brings more scrutiny than equity, given that repayment is required, and, as I have covered before, no aspect of AI is profitable, and as such, it can’t service debt.
Their solution has been to raise debt in secret. Dario Perkins, managing director of global macro at TS Lombard, has found that many AI companies are increasingly using SPVs to raise significant amounts of debt financing off the books. This covers their tracks and obfuscates the debt, making it “look” like the company is running on equity finance instead. Due to this, it is incredibly difficult to get an accurate figure on how much of the AI industry’s expenditure and growth comes from debt — but we know it is a lot!
Goldman Sachs has found that at least $141 billion of the $500 billion in capital expenses the AI industry has spent so far this year came from debt directly tied to the main corporate body through corporate credit issuances. To give you an idea of how insane that is, the entire AI industry capital expenditure in 2024 was $127 billion. In other words, the AI industry has taken on significantly more debt so far this year than it ever spent in total last year. This also means that we know that about 30% of the AI industry’s annual expenditure for this year came from “on the books” debt. This alone is a shockingly high figure that should be a blazing red flag!
We don’t know how much of the capital expenditure comes from “off the books” debt raised through things like SPVs, but we still know it is frequent. We know that Meta is looking to raise $26 billion in debt through an SPV by the end of the year. This one deal equates to 5% of the total AI industry’s capital expenditure for this year. But that is just one deal, from one company, and financially speaking, every major AI company is in the same boat (it’s just that some are more secretive than others).
This means we don’t know how much of that $500 billion capital annual expenditure actually came from debt — all we know is that it will likely be far higher than 30%. It could easily reach 40%, if not 50%.
Therefore, this bubble is four times bigger than the 2008 housing bubble and is deeply connected to the wider economy. It isn’t isolated at all. Indeed, if the AI bubble fails to pop soon, then its finances will be more connected to the wider economy than the internal tech industry economy.
This is a perfect storm. When this pops, OpenAI, Nvidia, Microsoft, Tesla, and others could easily face total collapse. They flew too close to the sun, but as their charred bodies crash down, they will land on others and destroy them too. Major banks, pension funds, and the loan market could face complete collapse or near-fatal damage. Unlike in 2008, we are already teetering on a global recession before this bubble bursts, so combining that with the fallout of the AI bubble bursting will create destruction on a scale we have never seen before.
When the bubble burst in 2008, it caused a cost-of-living crisis that has only gotten worse. We have been squeezed beyond our breaking point for years now. The pain, trauma, brutalisation and disenfranchisement of this experience have directly led to the rise of neo-fascism and the solidification of a super-wealthy ruling oligarchal class across the entire Western world. The AI bubble bursting will cause another recession that is even more brutal than 2008. It will take the ills we face today, like neo-fascism, and turbocharge them.
Do not take financial advice from me. Go and do your own research. However, I can see this hurricane coming right at us, and I can’t help but feel we need to brace ourselves, not just financially, but emotionally. We must retain our humanity through this. Only then can we hold those truly responsible to account and ensure our pain isn’t wielded against us. We need to make sure that after the dust settles, our society, politics, democracy, economy and culture actually serve us, the people, and not the uber wealthy who created this mess in the first place.
Thanks for reading! 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!
Sources: Market Watch, Axios, Will Lockett, Will Lockett, FT, The BMJ, Investopedia, Fortune, Investopedia, Investopedia, Paul Kedrosky
"It will take the ills we face today, like neo-fascism, and turbocharge them."
I'm in the "worse is better" camp as far as the crash is concerned. This time around, it will be hard to shift the blame away from the right. Not only are the Repubs in total control of the US political system but Wall Street (seen as a Democratic constituency in 2008) is in their camp, as is the tech industry, and the crypto world. The only coherent policy response is "blame the billionaires".
I watched 40% of my 401K evaporate in 2008. Poof, just gone, about 12 years before I retired.
I remember bailing out Wall Street while screwing people over poorly designed mortgages.
If AI does go under, what happens to the data centers, Will?