The AI Time Bomb

This will explode in our faces.

Photo by Logan Voss on Unsplash

A few weeks ago, OpenAI seemingly proposed that the US government back its loans, only to walk it back after a huge backlash. Why would they ask for this? Well, let’s be clear, this is really OpenAI asking for a bailout for when things inevitably fall apart. But that isn’t really the whole story. You see, this isn’t OpenAI trying to lock in “corporate socialism”. OpenAI, like all generative AI projects, is not, and will never be, profitable, and almost its entire value is based on hype-driven speculation (read more here). Such a bailout would not save them from collapse. No, this is the AI giant desperately trying to prevent the ticking time bomb at the core of the AI bubble from exploding. Sadly, though, the fuse is already lit.

How? Well, all we have to do is follow the money.

As always, none of this is financial advice.

Funding AI

The AI boom was initially funded by the mountains of cash tech giants like Microsoft, Google, Amazon, and Meta had amassed from not paying enough tax for the past two decades. This gave the likes of OpenAI their big leap forward and instigated the generative AI hype.

But, as I have covered before, AI experiences diminishing returns, meaning that AI development costs increase exponentially. As such, the AI boom threatened to drain these tech giants of all their liquid cash. So, the industry switched to equity financing, which involved selling off shares to raise enough cash to keep the lights on. However, this quickly devolved into the infamous circular financing deals between OpenAI, Nvidia, Oracle, CoreWeave and Microsoft, which, in the end, wasn’t even enough cash to cover the spiralling costs of AI.

So, over the past year, Big Tech has turned more and more towards debt financing, or taking out loans, to cover their mounting AI costs. To give you an idea of how rapidly they are taking on debt, Big Tech is predicted to spend $400 billion on AI in 2025 and has issued $200 billion in AI-related bonds (or debt) so far this year, with half of that happening since September.

Those of you with even a moderate understanding of finance might see this debt as a giant red flag. Keeping a wildly unprofitable venture going by rapidly taking on debt is just going to make the cash problems worse.

But wait — it gets so much worse.

AI costs are really spiralling out of control. AI expenditure is set to double next year. By 2028, Morgan Stanley predicts that the AI industry will spend $2.9 trillion, the vast majority of which will require debt financing. According to J.P. Morgan, the AI sector will need $1.5 trillion in bonds (or debt financing), to cover the costs of constructing planned data centres alone.

Given the amount of money being poured into AI, you would assume that the models would improve significantly and become profitable. But no, in fact, much of the science suggests this gigantic expenditure isn’t going to make the AIs much better at all (read more here). In fact, OpenAI itself has had to admit they can’t make their models much more accurate than they already are (read more here), which is a serious problem, as corporate AI adoption rate and revenue growth are beginning to stall (read more here and here) because the models aren’t good enough. So, OpenAI is pivoting to other markets, like social media apps and web browsers, to generate more revenue, but even if it dominated these markets, there simply wouldn’t be enough available revenue to cover its losses (read more here).

This ludicrous expenditure isn’t about long-term growth, developing better AI, or even making these business models remotely sustainable. It is about short-term speculative gain, and seemingly nothing more.

However, the real ticking time bomb is the way they are financing this speculative drive — the bonds.

Bonds, CDS & AI

To understand why, we need to understand what a bond is and how it relates to Credit Default Swaps (CDS). Stick with me; this gets juicy, I promise.

Bonds are a way for a corporation or a government to raise funds by directly selling their debt to investors. It is basically a corporation or government saying, “Give us your money now, and we will pay it back plus interest over time.”

But many investors want to mitigate the risks of their bonds. After all, if the government or corporation defaults on this debt, there is no way for the investors to get their money back.

This is where CDS comes in. These are effectively insurance policies for people who hold bonds, typically issued by investment banks or similar. So, if a bondholder is worried the government or corporation isn’t going to pay out and will fold on their obligations, they take out a CDS. This will cost them an annual premium, which is confusingly called a spread, based on how risky the debt is considered to be. Spreads are usually presented as a percentage cost of the bond value, so if a CDS has a spread of 1%, it will have an annual premium of 1% of the value of the bond covered. But in return, if the bond issuer defaults, the organisation providing the CDS will reimburse the full bond value.

Okay, so that is bonds and CDS. What does that have to do with the AI bubble?

Well, the entire future of AI depends on bonds.

The US AI sector is only set to rake in $41 billion in revenue this year. That is dwarfed by the more than $200 billion they have raised in equity finance and more than $200 billion in debt finance (which is almost entirely comprised of bonds) so far in 2025.

But don’t forget, in order to stay afloat, AI is transitioning from equity to debt financing. The first half of the year was when the majority of the more than $200 billion in equity finance was raised. But AI soaked up significantly more than half of the available global venture capital in the first half of 2025. So this source of cash wasn’t enough to satiate AI’s exponentially growing cash needs. Since then, bonds have been used to rapidly fill that gap, and as a result, half of the $200 billion in bonds the industry has raised this year have materialised since September.

Indeed, many predictions show that the bulk of AI investment required over the coming years to not just pay for the industry’s expenditure but also prevent major players like OpenAI from declaring bankruptcy will have to come from bonds.

In other words, the AI bubble is a giant Jenga tower precariously perched on top of a single brick, and that one brick is the bond market. That is the time bomb.

The Time Bomb

Here is the thing: as more and more investors want CDS to cover their bonds, the price of the CDS skyrockets. And since all these AI companies began issuing bonds in September, guess what has happened? The price of the corresponding CDS has gone through the roof!

The cost, or spread, of Oracle CDS has more than doubled since September to over 1%. Likewise, Microsoft (the major backer of OpenAI) has witnessed its CDS cost double. Google’s parent company, Alphabet, along with Amazon and Meta, have all experienced sudden and sharp increases in CDS costs of around 30%. But, by far the worst is CoreWeave, whose CDS spread is now well over 5%!

Okay, so what does this spike in CDS demand and cost mean?

Well, it shows widespread concern that these AI companies will fold. It shows that even the investors powering the AI bubble believe it will pop. That alone is a cataclysmic red flag and will make it far harder to raise funds through debt in the near future.

But it is also a bad omen of things to come, as this spike could drastically shorten the lifespan of these AI companies.

You see, as the CDS spread (annual cost) increases, bond buyers demand a higher interest to offset this cost. But CDS issuers see this higher interest rate as a riskier bond, so they raise the price of CDS. This is a vicious circle and means that the cost of borrowing for the bond issuer can skyrocket. We have already seen this. Take CoreWeave. Its CDS spread is well over 5%, which has sent its cost of borrowing through the roof. In fact, the interest on its bonds is nearly 10% and rising, and they are set to pay a billion dollars in interest on their debt this year alone!

But there is another factor increasing the cost of borrowing. In order to match the speculative hype that keeps their valuations high, these AI companies need to issue around 2.5 times as many bonds next year as they have this year. This is going to flood the bond market with too much supply, sending interest rates up to attract enough investors.

Combined, these two factors will likely make the cost of borrowing for AI companies shoot up. That can dramatically shorten their time until bankruptcy. They are using debt not just to invest in infrastructure but to soak up losses. But the cost of borrowing will increase their losses, so they will need to borrow more, which will increase their losses even more, and so on until the money just runs out. The debt effectively forces them to “speedrun” bankruptcy.

The Fuse

This is why OpenAI wanted the US government to back its future bonds. That would mean investors didn’t need CDS, as the US government would do it for free. This would keep borrowing costs down and attract investors, as it seems significantly less risky. Then, this would keep their cost of borrowing down and enable them to eke out their unprofitable existence for longer, enabling them to push their speculative value up higher.

We can see why OpenAI would want to do this, as this cost-of-borrowing issue is already severely impacting one of its major suppliers, CoreWeave. It was already a wildly unprofitable AI infrastructure business before it took on huge amounts of debt, posting a net loss of around 45% in 2024. But, as I said, it is projected to spend a billion dollars on interest this year, or 20% of its predicted annual revenue! Its debt is driving it even further into the red, and eventually it will collapse because borrowing will become too expensive or there won’t be enough investors to fill the gap.

If the AI bubble’s reliance on bonds is the time bomb, CoreWeave is the fuse. It is connected to every major player.

CoreWeave is one of Nvidia’s biggest customers, making up a huge portion of their revenue. Nvidia also holds $4 billion in CoreWeave stock. Not to mention OpenAI and Microsoft depend on CoreWeave’s data centres to train and operate their AIs.

When the shit inevitably hits the fan at CoreWeave, its collapse will serve as a trigger point. Firstly, it will have a profound impact on the operational and financial situations of the entire AI bubble. But it will also reframe the economics going on here, which will spook investors off. Both could be more than enough to burst the bubble and let the debt consume the entire industry from the inside out.

Summary

This is why I say the bubble has already burst. The mechanisms are in place, and the catastrophic ending is set in stone. It is just that the market hasn’t noticed yet. Will it notice before CoreWeave is destroyed by its debts? How long before CoreWeave runs out of cash — a few weeks, six months, or a year? I don’t know; I don’t have a crystal ball. I just know that this fuse is lit, and it’s only a matter of time before the bomb detonates. But I don’t know how long this fuse is, and equally well, something else could take this whole thing out before it explodes.

---

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: Bloomberg, Fidelity, Forbes, Economic Times, MooMoo, The Verge, Paul Kedrosky, Seeking Alpha, Market Watch, Deutsche Bank, Energy Wire, Futubull, UBS, Federal Reserve, Will Lockett, Investopedia, Reuters, Verdict, TMF, Fortune