Why the $300 billion Oracle-OpenAI deal could be fueling an AI bubble

By Christine Ji

Big Tech's AI capital expenditures are soaring, which raises the risk companies are overbuilding infrastructure before it can be monetized. That's a trend one expert calls a 'risky gamble.'

OpenAI is teaming up with Oracle in a staggering $300 billion deal to power its AI ambitions.

Is Oracle Corp.'s reported $300 billion infrastructure deal with OpenAI a sign of robust artificial-intelligence spending for years to come - or an indication that AI players have gotten ahead of their skis?

Investors certainly seemed pleased with the latest momentum coming from Oracle, which forecast huge revenue growth in the coming years - on the heels of what the Wall Street Journal reported was a landmark new deal with the maker of ChatGPT. That all sent Oracle's stock (ORCL) up 26% last week, and Wall Street bid up other plays on the AI trade as well.

But Benjamin Lee, a professor of engineering and computer science at the University of Pennsylvania, called the development a "risky gamble" for the two companies, suggesting that it points to a broader overinvestment in data-center capacity.

In today's supply-constrained landscape, overbuilding AI infrastructure might seem like an unnecessary concern. More graphics processing units and data centers could alleviate current bottlenecks in the AI training and inference process. But there could be potential pressures on future data-center demand, according to Lee, who has previously consulted for Meta Platforms Inc. (META) on data-center design and Alphabet Inc. (GOOG) (GOOGL) on hardware supply chains. If demand for AI workloads diminishes, it'll take investors much longer than expected to realize returns.

The future of AI inferencing workloads is uncertain because demand depends on the pace of "technology uptake" among individuals and enterprises, Lee told MarketWatch. Without a clear killer AI app to drive demand, cloud-infrastructure spending could get ahead of technology adoption.

The lack of a killer app has already been weighing on certain parts of the market, as many large software names have seen their stock prices come under pressure this year as investors have started to question when AI-enabled revenues will materialize.

Additionally, the demand for massive data-center capacity could flatten as companies adopt more efficient models for specific tasks, a trend that would accelerate as compelling AI applications emerge, according to Lee.

"There may be diminishing returns from building ever larger models," Lee said.

Also read: AI is eating software, and Adobe is on the menu. Why the stock could be in trouble.

Few on Wall Street are talking about the risk of an AI infrastructure overbuild. On the contrary, the prevailing sentiment among investors is that the consequences of underbuilding are far more severe than that of overbuilding, according to a note from Mizuho desk-based analyst Jordan Klein last week.

Additionally, it's difficult for Big Tech players to do anything but overbuild. There's a sense of "FOMO," or the fear of missing out, among these companies as they scramble to grab market share in a rapidly growing field by steadily raising their capital expenditures.

For data-center operators, business is booming like never before as the industry faces an historic surge in demand. When asked if there was a risk of overbuilding data centers, Sean Farney, the vice president of JLL's Americas data-center strategy, answered: "Categorically, no."

"I've never been so confident in affirming a demand curve in my 30 years of dong this," he told MarketWatch.

While Farney is confident that data centers will be fully utilized in the future, that doesn't take away the possibility of an "equity pricing boom or bust."

Lee affirmed that sentiment: "I think data-center capacity is important and eventually we will soak it all up" once AI use cases become more apparent, he said.

"The data centers eventually will be useful," Lee added. "Whether or not they will be useful in the same time frame that investors are thinking about is a different question."

Representatives from Oracle and OpenAI did not immediately respond to requests for comment.

Also read: Google, Meta and Amazon are facing an expensive AI risk that nobody is paying attention to

-Christine Ji

This content was created by MarketWatch, which is operated by Dow Jones & Co. MarketWatch is published independently from Dow Jones Newswires and The Wall Street Journal.

(END) Dow Jones Newswires

09-18-25 1727ET