Did Nvidia Just Prove There Is No AI Bubble?

Take a wild guess.

Photo by Ryan Braxton on Unsplash

“There is a lot of talk about an AI bubble,” Nvidia CEO Jensen Huang said during Nvidia’s recent Q3 earnings call. “We see something different.” It’s easy to see why Huang might think that. After all, this earnings call was basically a litmus test for the AI industry, which they aced. Their revenue for the quarter jumped 62% from last year to $57 billion, higher than Wall Street’s predictions. I’m sure from atop Huang’s mountain of gold, everything seems fine. Indeed, it seems like everyone has agreed that fears of the AI bubble catastrophically bursting are unfounded. But I am calling bullshit, because this is no mountain of gold — it is a tower of cards.

What everyone seems to be missing is that demand for trowels or spades doesn’t mean tulip mania or the gold rush bubble isn’t about to pop. Likewise, demand for the AI chips Nvidia sells doesn’t mean there isn’t an AI bubble that is about to ruinously collapse.

Case closed. But everyone also seems to be missing the damning context of this earnings report. In a similar way to pulling one thread and having your jumper fall apart, when you follow the money here, the horrifying extent of the AI bubble is laid bare.

Strap in — this is a long one!

The Fragile Cascade Pt. 1 — Data Centres

90% of Nvidia’s Q3 revenue came from AI data centre chip purchases. This shouldn’t be surprising as the price of data centre compute time is at an all-time high due to skyrocketing demand from AI. Data centre operators, such as Microsoft, Amazon, Oracle and CoreWave, are scrambling to increase supply to meet this demand and are buying millions of Nvidia chips to do so.

With demand so dramatically outstripping supply, you’d think these data centres would be insanely profitable. In fact, arguably, they should be at their peak profitability right now. But, no. They are losing money hand over fist.

Praetorian Capital CIO Harris Kupperman recently revealed that the AI data centres being built today will incur $40 billion in annual depreciation while generating somewhere between $15 billion and $20 billion in revenue. So, before we even take into account operational costs like maintenance and energy, AI data centres are losing more money than they make in revenue!

This is supported by a recent report, which found that the AI industry will need to generate $2 trillion in annual revenue by 2030 to break even on the data centres they plan to build by 2030. Even the rosiest, most blindly optimistic predictions estimate the AI industry falling $800 billion a year short of that, meaning these data centres are unprofitable on a scale never seen before.

But, believe it or not, this problem could actually be even worse! Michael Burry, of “The Big Short” fame, has accused data centre operators of artificially boosting earnings by underreporting depreciation of the chips. If true, then these predictions are far, far too optimistic.

Let’s also not forget that as more data centres are built, supply will begin to meet demand, reducing prices and making data centres even less profitable.

Now, is 90% of your revenue coming from an unsustainable industry that is miles away from profitability a good thing? No, not at all! That alone should utterly terrify investors!

The Fragile Cascade Pt. 2 — AI Operators

Okay, but what about the AI operators like OpenAI? Data centres being so unprofitable isn’t a problem if they can cover these losses with their profits. That way, Nvidia’s customers don’t disappear into bankruptcy.

That could work, if not for the fact that AI operators are even less profitable than data centre operators.

I wrote about this topic before (read more here), but OpenAI is set to post an annual loss of between $14 billion and $27 billion by the end of this year. And it won’t get any better. OpenAI is on track to post annual losses in the hundreds of billions of dollars by 2030, even based on their own wildly optimistic revenue projections. That’s right, the monolith of the AI world is actually getting further away from profitability. In fact, all AI operators are.

Ultimately, their costs are increasing exponentially, yet their revenue growth is stalling. Both are caused by the efficient compute frontier.

The efficient compute frontier describes how AI training experiences diminishing returns, requiring exponentially more compute power and data, and in turn exponentially more investment, to keep improving at a linear rate. This is why data centre and chip demand are skyrocketing, AI companies have hit this hard limit, and they need significantly more compute power to make AI models with only marginally better performance. Take OpenAI; its 2025 operating costs are set to be 266% higher than 2024. But its average annual operating costs from 2026 to 2029 are set to be 1,236% higher than their 2025 operating costs!

This is also why the latest AI models are not much better than their predecessors, despite being far larger and more expensive to develop. This, in turn, has completely kneecapped AI revenue growth. Take OpenAI again; in 2023, they increased their revenue by 169% from 2022, and in 2024, they increased their revenue by 250% from 2023, and in 2025, they are set to increase revenue by only 56% from 2024.

So, with costs increasing exponentially and revenue growth rapidly slowing down, their losses are only set to grow.

However, AI promises to augment and automate huge portions of the economy. Surely the vast increase in AI investment driving Nvidia’s earnings growth will create the exponential increase in computation needed to make AIs capable of automating the economy, and when they do, a huge revenue pool will open up for these operators, propelling them to profitability? Sadly not. Let me explain.

We know why the recent AI revenue slowdown happened. Studies from MIT found that 95% of AI pilots failed to deliver meaningful returns, studies from METR proved that AI coding tools actually slow down developers, and studies from Carnegie Mellon University discovered that the best agentic AIs out there failed the simple tasks given to them 70% of the time. All of these studies point to AI hallucinations (where the AI gets it wrong) as the culprit. Essentially, the amount of human labour needed to identify and correct these AI hallucinations is greater than the human labour saved by deploying the AI. As such, AI isn’t even a widely viable option for augmentation, let alone automation.

But OpenAI recently admitted it had found a sort of limit to the efficient compute frontier. Their latest research paperdetailed how putting more data into AI and increasing the computational power behind these models isn’t enough to reduce current AI hallucination rates. It also detailed how they have no viable alternative way to reduce hallucination levels, either.

In short, AI can never be accurate or trusted enough to majorly disrupt the labour market through automation or augmentation, and this pool of revenue is never coming their way.

This is why AI companies like OpenAI are turning to other markets to try and fill this revenue gap, like browsers, generative social media websites, or porn generation. However, even if they dominate these markets, the annual revenue available is simply not enough (read more here).

No matter how you spin it, there is simply no viable path to profitability for AI operators, and, in reality, they are set to rack up greater and greater losses as time goes on.

The Fragile Cascade Pt. 3 — Funding Squeeze

Okay, so the datacentres buying Nvidia’s chips are wildly unprofitable. Not only that, but the AI operators who are these data centres’ primary customers, and whose ravenous demand is the cause for the data centres buying so many chips from Nvidia, are even more unprofitable. Furthermore, both are set to get more unprofitable, not less. Suddenly, Nvidia’s primary revenue stream isn’t looking great. It looks like a fragile tower of cards, ready to fall.

But again, let’s follow the money. Where is all this money, which is being pissed away into the wind, coming from?

The AI industry started with Big Tech splashing out the huge cash reserves they built up from not paying enough tax over the past two decades. Then, when that began to dry up, it switched to equity finance (selling shares to raise funds). This powered the AI circular funding situation we have now, where OpenAI, Nvidia, Oracle, Microsoft and a few others are passing around the same pile of cash to buy smaller and smaller parts of each other, which pushes up all their values in the process. But this scheme can only go so far and doesn’t inject new money into the system, so the AI industry has turned to debt financing, also known as loans, to meet its insatiable cash appetite.

Alphabet, Amazon, Meta, Microsoft, and Oracle have issued around $100 billion in bonds (a form of debt financing) so far this year, with the majority being issued since September. But that is the tip of the iceberg, as the AI industry as a whole has generated $1.2 trillion in AAA-rated bonds (debt as an investment) in just a few years — though whether that should be rated as AAA ‘investment grade’ debt is another question. This debt market is set to boom, with OpenAI planning on funding around 75% of its upcoming $1.5 trillion expenditure between now and 2030 with debt.

However, lenders are painfully aware of how unprofitable and unsustainable AI companies and data centre companies are.

As such, there has been a sharp spike in AI-related bond insurance demand. Bond insurance is insurance that a lender takes out to ensure that they will get their money back in the case of the borrower defaulting. The AI-tied bond insurance market from January to September was already high at $25 billion, but since September, it has ballooned to $100 billion. This means that all the lenders who bought bonds from Alphabet, Amazon, Meta, Microsoft, and Oracle bought bond insurance because they think they will run out of cash.

Let that sink in. Pretty much all these investors are worried that Big Tech will run out of money.

Bond insurance premiums increase dramatically as demand increases. This means lenders require larger returns to cover the cost. But loans are assumed to be riskier the larger the returns are, and the amount being covered by the insurance also goes up. Together, these factors push insurance premiums way up, creating a vicious circle. As such, this rapid spike in bond insurance will severely reduce how much cash the AI industry can raise through debt.

This is why OpenAI toyed with the idea of asking for US government-backed bond insurance, because it is one of the only things they can do to keep this debt-sourced cash flowing in and stave off bankruptcy — a topic I will be covering next week.

In short, the capital flowing into AI, the investments the entire industry not only uses to buy Nvidia chips but also needs to not implode, is being turned off as we speak.

The Fragile Cascade Pt. 4 — Crash

So, Nvidia’s earnings aren’t just a tower of cards but one where cards from the bottom are being actively removed.

Again, just a reminder that 90% of Nvidia’s earnings, and some $4 trillion worth of Nvidia’s value, are based on this flow of cash from equity and debt finance, through AI operators and AI data centres, and into Nvidia’s accounts. Let’s also not forget that Nvidia’s P/E ratio (its ratio of share value to revenue) is sky-high due to dubious proclamations of future growth, and when the opposite happens, this P/E ratio will become much more conservative.

This is the fragile cascade. The debt door is already closing on the AI industry, and when it fully shuts, AI operators and AI data centres will run out of cash and likely go bankrupt, Nvidia’s revenue will crash by around 90%, and Nvidia’s value will crash by more than 90%.

Suddenly, this earnings report isn’t looking so good, is it? It’s almost like relying on possibly the most unsustainable industry ever created for 90% of your revenue will make you incredibly vulnerable, no matter how high your current earnings are…

** Since writing, Nvidia’s shares have plunged in value to below the level they were at during the earnings call, suggesting investors have realised this vulnerability.**

Steady Future?

But Huang did predict that Nvidia will rake in half a trillion dollars from its AI chips next year. This seems to be what soothed investors the most, as it, at the very least, suggests that the AI bubble isn’t going to pop for a while.

But again, these predictions are based on circular financing and debt financing. With the circular financing, Nvidia is effectively buying its own chips via ‘investments’ in OpenAI and Oracle, which won’t actually add anything to Nvidia’s overall growth. AI operators and AI data centres are also going to have to raise hundreds of billions in debt next year to pay for their commitments to Nvidia, something this bond insurance spike suggests will be incredibly hard!

So, no, this revenue prediction doesn’t mean the AI bubble is far away at all. All it does is prove that Huang, like all the other AI CEOs, isn’t living in reality.

The False Equivalency

Okay, so why were Wall Street and all these investors so happy to see Nvidia beat its earnings targets? Surely they are asking the same questions as us? Surely they can see this dire situation?

Well, yes, they can see it — they just don’t care.

Investors know AI is a wildly unprofitable bubble that will eventually burst and deal untold damage to the economy. Hell, that is why all those lenders want bond insurance. They aren’t worried about AI companies racking up enough debt to sink the US economy just to keep the lights on. What they want to do is profit from the bubble, just as those selling spades during the gold rush made far, far more money than gold prospectors. They see Nvidia as their golden ticket. If Nvidia had missed its earnings, it would be a sign that selling spades is getting less profitable. The attention surrounding Nvidia was never about the existence, health, or risks of the AI bubble, but about the ability to parasitically profit from it, which is very, very different. We should not see their sighs of relief as a sign that things are okay, because they don’t remotely care about you.

Summary

That was an awfully long way of saying no, Nvidia hasn’t disproven the existence or risks of the AI bubble. Well done for making it all the way through!

Now you understand why seeing investors, financial pundits and news reporters hailing Nvidia’s earnings report as a sign the AI industry is doing well is so frustrating that it makes me want to get a hair transplant, just so I can rip it all out. This context of AI’s limitations, the impossibility of industry-wide profitability, the gravy train of debt that is keeping this entire thing afloat starting to close, and Nvidia’s highly dependent and wildly undiversified revenue stream are utterly essential. Yet no one seems to be talking about this context, let alone using it to understand what is going on with Nvidia.

This is a huge shame, because if they did, we would all see just how much of a hollow ruse this all is and how the AI industry is driving itself off a cliff at 150 mph, with the entire Western economy in the passenger seat. Mark my words, this earnings report isn’t the calm before the storm. It is the calm before the Category 5 hurricane.

---

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: BBC, CNBC, The Guardian, Reuters, The Street, Bloomberg, Fortune, Constellation Research, Fortune, WSJ, Toms Hardware, Nikkei Asia, OpenAI, Reuters, Lesswrong, Fortune, METR, MLQ, Will Lockett, Will Lockett, Will Lockett, Will Lockett, Will Lockett, Will Lockett