This analysis was first published in SvD Näringsliv, in Swedish, on December 20th, 2025. This piece was translated from Swedish by Claude. Some phrasing may differ from a human translation.

Hundreds of billions have been made by those who bet on Oracle as a winner in the AI race. Now the opposite is growing — investors who see a chance for profit if the tech company can no longer service its loans. Some are warning of a possible bomb.

A long table in dark wood runs through the conference room. On one side sits Michael Burry, played by Christian Bale, and on the other, representatives from investment bank Goldman Sachs.

The scene from the film The Big Short has become something of an icon. Burry convinces the bank to sell a financial product that pays out if a large number of mortgages cannot be repaid. The banker explains that what Burry is asking for would be seen as a very bad investment by most in the outside world.

Burry replies quickly — and it later turns out to be entirely correct: “Everyone else is wrong.”

The product Burry buys is called a CDS — a “credit default swap.” You can think of it as a form of insurance product, in that it pays out when a loan cannot be repaid.

But unlike insurance, you do not need to own the underlying asset to buy a CDS. You can profit from other people’s misfortune regardless.

And that is precisely what has started happening in the bubbling world of AI.

Larry Ellison has become increasingly prominent and well-known in recent years. As the founder of the previously rather sleepy database company Oracle, he had long been a billionaire — with luxury yachts and the classic trappings that come with that life. The company was stable, profitable, and a little uninteresting to most. Until around 2021, that is. That is where the next great upward journey began — and into the inner circle of the tech world.

Ellison’s Oracle is a partner to TikTok on data storage, and is set to become a part-owner of its American operations. But it is with the AI wave that Oracle has truly come into its own. All these data centres needed to expand capacity have to be built by someone — and Oracle has stepped forward. Data centres are something they have built before given their cloud business, but the scale is now something else entirely.

This week it emerged that Oracle’s lease payments for data centres yet to be built amount to 248 billion dollars — around 2,300 billion kronor. These are to be paid between now and 2028. The figure Oracle reported the previous quarter was around 100 billion dollars. The increase is somewhat astronomical, and analysis firm CreditSights described the situation as a “bomb.”

The question the market is now starting to ask is what happens if Oracle cannot make these payments. And shortly afterwards, people start wondering whether there is money to be made if Oracle fails to deliver. That is where the CDSs come in.

Since September, trading volumes in these products have increased by 90 percent. For CDSs linked specifically to Oracle, volumes have tripled during the year and are now trading at their highest price in over 15 years.

This does not necessarily mean the market believes Oracle will go bankrupt. But it clearly shows that many are concerned the company may have taken on too many large, expensive projects at once — and that investors want some form of protection or insurance should things start to go wrong.

Many AI companies are building data centres at the moment, so the situation is not unique. Microsoft, Amazon, and Meta all have major projects under way in the area. The problem here is the scale.

Oracle’s commitments are larger than all three of the other tech giants’ combined. The company itself estimates that nearly 75 percent of Oracle’s revenues in the coming months will be spent on data centre expansion.

The concern also sits one level above Oracle — with its customers. OpenAI has made a deal worth 300 billion dollars with them, for example. The money is to go towards data capacity over five years, starting in 2027. For Oracle to be able to deliver on this, they need to build.

But there is another problem: at present, OpenAI does not have all 300 billion dollars to pay with. That must be resolved before the payments are due.

Given the situation, it is easy to understand why there are people thinking like the investor Michael Burry from The Big Short. Do all these AI investments actually add up? And if they do not — how can I make money from the fact that everyone around me is wrong?

The last time the price of a credit default swap for Oracle was this high was 2009.

And most people probably remember what happened in the financial world back then.