Oracle chief executive and former Oracle Industries president Mike Sicilia said: “We’re really in a unique situation to deliver what we call applied AI.” He claimed the offering covers everything from infrastructure to analytics and applications.
Sicilia now shares the top job with Clay Magouyrk, previously in charge of the company’s cloud infrastructure racket. They have been handed the wheel just as fears of an AI investment bubble start to bubble over.
Oracle’s stock shot up more than 40 per cent in September after the cloud outfit revealed it had snagged $317 billion in future contract revenue during the quarter ending 31 August. That figure included a whopping $300 billion deal with OpenAI, which reportedly stretches over five years.
That mega-deal has raised eyebrows. OpenAI boss Sam Altman admitted the company will not turn a profit until 2029, and Moody’s pointed out the risks of Oracle hinging its balance sheet on such a flaky partner. The bean counters are not thrilled about it either.
Earlier this month, Oracle’s shares slid by as much as 7.1 per cent after word got out that margins on renting out Nvidia’s AI chips were painfully thin.
Sicilia and Magouyrk face the unenviable task of convincing investors, especially at Thursday’s investor day, that all this spending will eventually lead to something resembling profit.
Gartner analyst Balaji Abbabatulla told the Wall Street Journal that Oracle’s strategy relies on flogging a full stack of AI solutions to big corporates.
“They’re not going to be able to show clear returns if they don’t go for those large and multibillion-dollar deals,” he said.
The goal, he added, is to link Oracle’s AI infrastructure to its bread-and-butter business apps and databases in one fat package that makes switching to another vendor a logistical nightmare.
The next battleground is AI inference, which is running the models once they have been trained. This could prove more lucrative and sustainable than model training. Whether Oracle can pull off both is not clear.
Magouyrk thinks it can “Do inferencing right alongside their data with the best models.”
He said Oracle’s new AI Data Platform would allow customers to ramp up AI usage “as much as one thousand times” once they start using it.
SoundHound AI chief information officer Shawnna DelHierro said her firm uses Oracle’s cloud for training and inference, processing over a billion voice queries a month. She claimed the choice was down to Oracle offering “zero latency” and behaving like a “true partner.”
Still, most companies are yet to see real returns from their AI splurges, even as they keep pouring cash into it.
Oracle is also fighting the perception that it is burning money on low-margin data centres. The firm staked its reputation on being a cheaper cloud provider, which makes the economics of renting expensive Nvidia kit a bit harder to swallow.
Magouyrk disagreed, saying that “margins are the wrong way to look at the business. Scale changes everything and it ends up being a profitable business."
TD Cowen analyst Derrick Wood agreed, to a point. He said margins would be thin early on. “You have to go build the infrastructure before you can turn on all the revenue meters,” he said. “But as the consumption meters start going on, you start to recoup a lot more of your capital expense and start to see gross margins significantly improve.”
To fund all this, Oracle flogged $18 billion in investment-grade bonds at the end of September. That debt has investors twitchy, particularly around massive projects like its Stargate AI data centre plan with OpenAI.
Magouyrk admitted focusing on the debt alone was “a very scary picture” but said it looked better when you factored in expected revenue and cash flow.
He insisted Oracle is not putting all its chips on OpenAI.
“Pretty much all of the big model providers use our cloud in one form or another. You can’t get more than 100 per cent of the Pokémon,” he said.
