Oracle eases Wall Street fears over AI spending

Oracle $ORCL keeps walking into earnings with roughly the same giant cloud-and-AI story and watching Wall Street rewrite the ending. 

In September, talk of more than $500 billion in booked cloud orders and RPO of $455 billion sent the stock up close to 30% after hours. In December, softer guidance and another $15 billion in planned spending turned that same story into a balance-sheet scare. On Tuesday, Oracle came back with revenue up 22% to $17.2 billion, OCI up 84% to $4.9 billion, cloud revenue up 44% to $8.9 billion, and RPO at $553 billion — and the market, once again, decided it liked the plot. The stock was up around 10% in Wednesday’s trading.

Nothing cleans up an AI earnings story like a good utility bill.

The company’s case barely moves: more cloud, more AI demand, more enterprise data inching toward the models, more backlog, more ambition. What keeps changing is Wall Street’s verdict. In September, the numbers read like a prophecy. In December, those same numbers read like a very expensive habit. In March, they read like a business again.

This time, the market decided it liked one detail in particular: Oracle said that much of the latest RPO jump came from large AI contracts where it “does not expect to have to raise any incremental funds” because the equipment is either “funded upfront” by customers or bought by the customers themselves and handed over. Pair that with unchanged fiscal 2026 capex guidance of $50 billion and a raised fiscal 2027 revenue target of $90 billion, and Wall Street suddenly decided the same story sounded like it had a happily ever after once again.

The post-earnings reaction sounded more relieved than smitten. Wedbush called the report a “huge relief” for the broader tech sector and said the company remains “a bedrock for the AI Revolution.” Citi called the earnings “very solid.” Morgan Stanley $MS’s read was that the quarter could start rebuilding investor confidence after December’s stumble. J.P. Morgan, meanwhile, upgraded the stock to Overweight from Neutral and said “blind pessimism” had gotten too extreme and created a more attractive risk-reward setup. Same Oracle story, new chapter, and a choose-your-own ending.

Oracle’s long game still trades quarter to quarter

The same Oracle story that looked thrilling in September and alarming in December is, at heart, a data story. Oracle isn’t pitching AI as a novelty. No one really is anymore. How can they when it’s the heart of today’s market? Instead, Oracle is pitching AI as something that gets more valuable — and more billable — once it gets close to records, workflows, compliance, and all the dreary corporate machinery that already runs inside Oracle’s walls. That’s why the quarter’s loudest number, Oracle Cloud Infrastructure’s (OCI) 84% growth, matters. It’s also why the quarter’s sneakiest number, multicloud database revenue up 531%, may matter more.

Co-CEO Mike Sicilia said on the earnings call that “data gravity matters here” and that “mission-critical data gravity matters” could be even more crucial. That’s the whole Oracle pitch in one sentence: If companies want AI agents to do work that matters — approve payments, move inventory, manage schedules, touch healthcare records, pull from finance systems — those agents need to get close to the systems of record. Oracle would very much like to be the landlord in that arrangement.

Fellow co-CEO Clay Magouyrk pushed the same point from the infrastructure side. Early in the AI boom, he said, many customers thought they’d train their own private large language models. That has “largely proven” not to be the case. What they want instead is “the best models” combined “in a private way with their private data.” That line is doing a lot of work to shift the quarter’s story toward the part that matters: who owns the data, who runs the links, and who gets paid when AI starts doing actual work.

The company keeps saying that AI won’t live in some hermetically sealed toy box. It’s going to seep into payroll, hospitals, banks, supply chains, customer records — and whatever other grimly profitable systems already keep the lights on.

On the call, Oracle said it has already delivered well over 1,000 embedded agents inside its horizontal and industry applications. In the earnings release, it said AI code generation is letting it build more software “in less time with fewer people.” Same long game. Same insistence that AI belongs inside the old plumbing, not floating above it. Same bet that the boring layer ends up being the lucrative one.

The company isn’t reinventing its story every quarter. It’s rewriting a few lines here and there. Cloud is now 52% of revenue. Database growth is accelerating inside the multicloud. AI coding tools are changing product economics. Demand for AI training and inference, Oracle said, still grows “faster than supply.” Oracle keeps bringing the same argument to the market and finding out, every three months, whether investors think it sounds bankable and what the AI mood ring says this time around.

The trend looks durable. The trust doesn’t

This quarter, Oracle brought the answer the Street had been chasing since December: somebody else helping cover the chip bill. Many of the contracts behind that swollen RPO pile are structured so customers prepay for equipment or bring the hardware themselves. For a company that has spent months spooking investors with the size of its AI buildout bill — looking like the most debt-exposed player in AI infrastructure — that’s the part that kept this from turning into another backlog sermon.

“Oracle’s quarter is a beat and a stress test result for the AI trade,” eMarketer analyst Jacob Bourne said, according to Reuters. “As the most debt-exposed major player in AI infrastructure, Oracle is the canary in the coal mine, and this report suggests there's underlying health in AI spending beyond the hype.” But investors still want proof that all this GPU capacity and AI cloud expansion will become earnings and free cash flow, not just giant contracted revenue and giant capital bills.

Magouyrk added more concrete detail on the call. Oracle, he said, signed more than $29 billion in contracts using bring-your-own-hardware and upfront-customer-payment models. It delivered more than 400 megawatts to customers in Q3, and 90% of that committed capacity was delivered “on or ahead of schedule.” That’s the kind of operating receipt Wall Street has been begging for since December — not another promise about some glorious, lucrative future, but proof that steel, power, cash, and timing are lining up in the same quarter. 

The margin story helped, too. Oracle said the gross margin on AI capacity delivered in Q3 came in at 32%, above its 30% guidance. Magouyrk reiterated that accelerator rentals still run in the 30% to 40% range, while 10% to 20% of customer spending in OCI goes to other services, including database work that can carry 60% to 80% gross margins. Put all that together, he said, and the margin profile “continues to strengthen.” 

Same Oracle story again — only this time, investors could see the scaffolding.

None of that means the trust issue has vanished. It just lets Oracle spend a quarter talking about growth before the balance-sheet questions take over again. December’s quarter still happened. Oracle still missed expectations on guidance then, still said spending would rise by $15 billion versus earlier estimates, and still watched the shares slide. Heading into this week’s report, the stock was more than 50% below its September highs, and analyst commentary was still laced with nerves about debt load, OpenAI concentration, and whether all this construction would turn into profits fast enough.

The job this quarter was to make the same sprawling future sound expensive, yes, but survivably so. That’s a different task, and it’s a much harder one. Plenty of companies can sell growth. Fewer can sell growth, giant capex, giant contracts, and a plausible answer to who’s paying for this whole thing.

Same thesis, new referendum

That’s what keeps making Oracle such a strange quarterly ritual. The underlying thesis barely budges, yet the market keeps reacting as though somebody swapped out the script between acts. September’s version sounded like the company had cracked open a gold mine. December’s sounded like somebody had found the invoice. March’s sounded, for the first time in a while, like an operating business with actual discipline. 

Oracle didn’t change religions. Wall Street changed moods.

Morgan Stanley still wants proof that Oracle’s GPU-as-a-service push will turn into earnings and free cash flow. Oracle’s earnings will continue to come with lingering nerves around debt, data-center spending, and reliance on a small circle of very large AI customers.

The company is still riding one of tech’s biggest secular waves, still piling up cloud demand, still arguing that AI gets truly valuable once it sits close to corporate data and old enterprise plumbing. What changes is the read-through. One quarter, the numbers sound like destiny. The next, they sound like overhead. This week, Oracle found a new version of the old story that the market was willing to read with less suspicion. 

There was one genuinely new thing for the market to grab onto, though. This was Oracle’s first quarter in more than 15 years with both revenue and earnings growth above 20%. That might just give investors a cleaner reason to treat this round of Oracle déjà vu as something more substantial than another backlog sermon.

Still, nobody should pretend the loop is broken. Oracle keeps handing Wall Street the same giant future — more cloud, more AI demand, more data moving toward the models, more contracts already spoken for — and Wall Street keeps deciding, every three months, whether that future looks like growth, danger, or somebody else’s tab.

Oracle no longer gets points for telling a giant AI story — plenty of companies can do that. Larry Ellison can’t just say something grandiose on cue — plenty of tech billionaires can do that, too. Oracle’s trick is making the same story sound bankable every 90 days. This quarter, the answer tilted back toward growth. Next quarter, the company has to sell the same future all over again.