Originally published in Svenska Dagbladet by Björn Jeffery, September 12, 2025
Grand promises about the future just made Oracle founder Larry Ellison the world’s richest man. But the deal that sent the stock soaring is far from certain.
A strange thing happened with cloud company Oracle this week. The figures in the quarterly report were worse than analyst expectations. But the share price did not fall. Instead it surged over 30 percent and made major shareholder Larry Ellison the world’s richest man.
The reason was Oracle’s forecast for revenues from the ongoing AI boom. One example was the contract OpenAI signed with the cloud giant to purchase data services from Oracle for 300 billion dollars over five years. There is just one problem. It may sound straightforward, but it could prove difficult to resolve even for those with billionaire owners.
OpenAI does not have 300 billion dollars to spend. And Oracle does not have the capacity to sell 300 billion dollars’ worth of services either.
The forecast is ambitious, to say the least. Revenues from the company’s cloud services will come in just above 10 billion dollars for 2025. In five years, that figure is supposed to be 144 billion dollars. In total, Oracle announced that customer contracts — for which neither delivery nor payment has yet occurred — grew from 138 billion to 455 billion dollars. The OpenAI deal is included in that figure.
The measure Oracle uses is called “performance obligations” — a form of customer promise within accounting that indicates something will be purchased in the future. For the revenues to materialise, two things are required: the customer must be able to afford to buy, and the company in question must have enough to sell at that point in time.
In the case of Oracle and OpenAI, there is reason to question whether this will happen. The energy alone required to run these data centres is estimated at 4.5 gigawatts, equivalent to roughly four million households. As a comparison, that is slightly less than all the reactors at Sweden’s nuclear plants Forsmark and Oskarshamn combined.
Oracle will also need to invest in substantially more chips from companies like Nvidia. The strong demand — competing with the world’s other major cloud and data centre companies — could drive prices up considerably. And chip deliveries have so far been difficult to guarantee due to high demand.
OpenAI, for its part, faces a major financing challenge. The current plan, according to The Information, is that the company will lose 115 billion dollars — over a thousand billion kronor — between now and 2029. A positive cash flow is not expected until 2030.
Intense negotiations are currently under way with major shareholder Microsoft about how OpenAI’s corporate and ownership structure might look going forward. The aim is partly to prepare the company for a potential stock market listing, which would help with financing.
These are not small challenges on either side. The Oracle-OpenAI deal runs for five years, starting in 2027. Until then, all of this must be resolved.
It is normal for projects of this magnitude to take a long time to plan. We are talking about an investment of around 2,800 billion kronor, after all. That preparations are required is of course reasonable. But despite Oracle not having booked any of these revenues here and now, the share price surged over 30 percent. The market appears to believe the contracts will be honoured and will come to fruition — and is pricing Oracle on the basis of that information.
This is not a property deal or anything similar, however. The predictability and visibility around the AI market even 1.5 years from now — when the contract is set to begin — is murky at best. A number of questions are piling up. What hardware will be needed in AI data centres at that point? Will there be sufficient energy supply? Will Oracle manage to build all of this in time?
And perhaps the most important question of all: does OpenAI — the customer in question — even have the money to pay for the party?
The stock market is currently showing no restraint in its enthusiasm for what AI development can do for these major tech companies. Shares are being traded right now based on high hopes for the future.
But when evaluating a deal of this size, a reasonable starting point is to ask whether the seller is able to deliver the service and whether the buyer can afford to purchase it.
It is far from obvious that the answer is “yes” to both questions.
