Oracle founder Larry Ellison is, by any definition, a very wealthy man.

Owning multiple rare sports cars, luxury homes, and even a Hawaiian island (Lānaʻi, purchased in 2012), Ellison’s personal fortune is said to be north of $300 billion.

For a time on Wednesday, he became the world’s richest man, overtaking Elon Musk as his net worth grew to $393bn.

The reason? Oracle’s soaring share price, with the company’s value shooting up more than 40 percent thanks to some better-than-anticipated financial results, driven by the growth of its public cloud platform. Ellison, who owns 41 percent of the firm, was the chief beneficiary. Things levelled off slightly, with Oracle stock closing just 36 percent up, meaning Tesla billionaire Musk is once again the world’s wealthiest individual.

Oracle and the OpenAI cash

Nonetheless, things are looking rosy for Ellison and Oracle, which has been a slightly unexpected beneficiary of the AI race. Once known primarily for its database software and derided by many for its meager cloud offering, the firm has successfully evolved into a key provider of AI infrastructure to the hyperscalers, with whom it also competes for enterprise and AI customers.

This was highlighted on its earnings call this week, with CEO Safra Catz referencing four multi-billion-dollar cloud contracts signed with three customers in Q1 of 2025 alone. The company’s recent deals include "significant cloud contracts with the who's who of AI," Catz said.

One of these is OpenAI. A conveniently-timed report from the Wall Street Journal, which dropped late on Wednesday, revealed the ChatGPT-maker had agreed to pay Oracle an eye-watering $300bn over five years, starting in 2027, for compute infrastructure. No doubt news of the deal, which has yet to be confirmed by either company but which would be one of the biggest cloud contracts in history, boosted Oracle’s value even further.

The fortunes of OpenAI and Oracle are already closely entwined thanks to Stargate, OpenAI’s sprawling data center project, which aims to build massive hyperscale campuses across the US and around the world. Oracle is a founding partner, and is working with OpenAI on the first Stargate data center, in Abilene, Texas.

In July, OpenAI announced it was sourcing 4.5GW of compute power from Oracle at a cost of $30bn a year. The deal reported by the WSJ dwarfs this, requiring a $60bn annual spend if the investment is evenly spread across the contract (details of how the deal is structured have yet to be disclosed).

Big numbers, big problems?

While improbably enormous numbers have become commonplace in the data center industry since the AI boom, it is stretching credibility to its very limits to think that this deal will come to fruition in the way it has been described.

Consider first OpenAI, a company that loses vast amounts of money every month and which, in June this year, said it had $10bn booked in annual recurring revenue. Even allowing for its impressive and rapid growth (it apparently made $1bn in July 2025 alone, suggesting next year’s revenue will be higher), the idea that it will be in a position to pay out $60bn a year *just to Oracle* in the next 18 months appears fanciful. For reference, Microsoft’s entire predicted capex for 2025 is $80bn, and its annual revenue is slightly higher than $10bn. Even if the deal involves lower initial payments, Sam Altman’s company is going to be in the hole for a lot more cash. According to a report in The Information last week, it already expects to burn through $115bn by 2029.

But it is not just OpenAI that could struggle to fulfill its side of the bargain. Oracle will need a lot of power and a lot of chips quickly if it is going to make good on its promises, and both are highly-prized and difficult to source commodities. When it comes to power, the industry has grown used to talking in gigawatts, but large projects usually promise to deliver this over long horizons. Will Oracle really be able to find the power it needs to meet OpenAI's data center desires in the next five years? And even if it does, will it be able to get its hands on enough GPUs to put this power to use? That’s before one considers Oracle’s other customers, the rest of the “who’s who of AI” Catz mentioned, that will also be demanding lots of power and lots of chips.

It feels like an infeasible tall order, and talking about such wildly high numbers really does no one, apart from Oracle shareholders, any favors. The data center industry is already being buffeted by accusations that AI is a bubble waiting to pop, and if its biggest names continue to hype up the sums of money involved (it should be said OpenAI and Oracle are far from the only offenders in this respect), then those accusations are unlikely to go away. It also seems somewhat distasteful to be hailing Oracle’s huge profits, and Ellison’s vast wealth, at a time when the company is laying off hundreds of staff around the world, as exclusively revealed by DCD last month.

Beyond that, talk of $300bn seems somewhat unnecessary. If nothing else, Altman has proved in recent years that he is a master salesman, and if he can continue to convince investors to back OpenAI’s potential, then lots of money is likely to flow Oracle’s way. It would just be nice if the sums discussed publicly bore some resemblance to reality.