Oracle’s Missteps in Cloud Computing Are Paying Dividends in AI

Oracle shares are up 34% since January, well outpacing the Nasdaq’s 14% rise and those of bigger competitors Microsoft, Amazon.com and Google.

It is a surprising revitalization for a company many in the tech industry had dismissed as a dinosaur of a bygone, precloud era. Oracle appears to be successfully making a case to investors that it has become a strong fourth-place player in a cloud market surging thanks to AI.

Its lateness to the game may have played to its advantage, as a number of its 162 data centers were built in recent years and are designed for the development of AI models, known as training.

In addition, Oracle isn’t developing its own large AI models that compete with potential clients.

The company is considered such a neutral and unthreatening player that it now has partnerships with Microsoft, Google and Amazon, all of which let Oracle’s databases run in their clouds. Microsoft is also running its Bing AI chatbot on Oracle’s servers.

Other big clients include xAI and Nvidia. Oracle co-founder and Chief Technology Officer Larry Ellison’s tight relationship with their respective chief executives, Elon Musk and Jensen Huang, helped seal those deals.

The company touted the benefits of a burgeoning AI business in its earnings Monday, posting cloud-infrastructure revenue that was up 45% from the same quarter last year. Its results beat expectations and, coupled with robust guidance, sent the stock up 9% in after-hours trading.

Ellison said on the call that building giant data centers is “something that Oracle has proven to be very good at. It’s the reason we’re doing so well in the AI training business.”

A recent report from Morgan Stanley estimated AI revenue will increase from 15% of Oracle’s total cloud revenue in its most recent fiscal year to more than half by 2027.

“Oracle is being looked upon more as an AI play as opposed to relational databases,” said Dan Morgan, a portfolio manager at Synovus Trust who owns Oracle shares. 

Late-mover advantage

The company rebuilt its cloud business in 2018, after a disappointing launch two years earlier. Now it is pouring billions of dollars into Ellison’s plan to build 100 more data centers.

Part of its pitch to clients is that it isn’t trying to upsell them on other cloud services or proprietary chips the way bigger providers often do. 

“They’re not testing out experimental things. They’re just building something that really works,” said Nick Frosst, the co-founder of AI startup Cohere. Oracle is one of the company’s investors and cloud providers. 

Ellison, who at 80 remains one of the most influential people in Silicon Valley, has played a key role in the company’s success in AI. Not long after xAI launched last year, he shared a late-night meal with Musk and Huang at the Japanese restaurant Nobu in Palo Alto, Calif., to hash out a deal for chips and cloud computing.

Nvidia selected Oracle to host its own cloud offering, which features Nvidia’s most advanced graphics processing units—the chips used to train AI models. Nvidia’s GPUs have been a scarce and highly sought commodity in the tech industry.

A boom that could fizzle

Analysts say the biggest risk to Oracle is that all the AI spending coming its way could prove transitory. 

xAI, one of Oracle’s biggest customers in the space, is building its own AI data center in Memphis to train the next generation of its Grok large language model, instead of using Oracle.

Many of Oracle’s AI customers have been using its cloud to train their models—a process that involves ingesting massive amounts of data to construct the brains of the software.

While it is an expensive process and good business for Oracle, it is also a one-time deal. Once the model is trained, it can operate on another cloud. Some AI companies might prefer to run their services with larger cloud companies that can sell it to their broader array of clients.

OpenAI, for example, plans to do some of its training work with Oracle but hosts its models exclusively on Microsoft’s cloud.

Ellison said Monday he believes demand for AI training will continue unabated.

“This business is just growing larger and larger and larger,” he said. “There’s no slowdown or shift coming.”

The company is also betting that budding relationships with AI startups and longstanding ones with companies that use its database software will help keep them on board.

As its bigger competitors keep building more AI data centers, there may be less excess demand coming Oracle’s way, however. With its more limited resources, the company may also struggle to maintain the edge on technology it got by coming to the cloud late. 

“Given the significant amount of [capital spending needed] to effectively compete in AI infrastructure longer term, I do wonder whether Oracle can sustain what I view as a transitory architecture lead,” said Stifel analyst Brad Reback.

Write to Tom Dotan at tom.dotan@wsj.com