Oracle stock surged 10% after its latest earnings report, and the reason is simple: the company's AI cloud infrastructure business is growing faster than almost anyone expected. Analysts predicted 79% growth in the infrastructure segment this quarter. Oracle delivered 84%. That kind of beat — on a segment that Wall Street is obsessed with right now — is exactly what sends a stock printing a god candle. And with a remaining performance obligation (RPO) backlog now sitting at $553 billion, the demand story isn't slowing down anytime soon.
Why Did Oracle Stock Jump 10% After Earnings?
The short answer: Oracle beat on the metric that matters most right now — cloud infrastructure growth — and raised its full-year revenue guidance to $90 billion for the fiscal year beginning in June, above the analyst estimate of $86.7 billion. But beyond the topline numbers, the market was watching for signs that Oracle's massive infrastructure buildout wasn't going to blow up in Larry Ellison's face.
02:15
Oracle's infrastructure segment growth: 84% actual vs 79% analyst estimate — the beat that moved the stock
Watch at 02:15 →
There had been real fear heading into this print. Could Oracle actually deliver on its aggressive timelines? Were the deals with OpenAI and Meta real, or just headlines? Was Larry going to get caught holding the bag on billions of dollars of GPUs he couldn't monetize fast enough? This earnings answered those questions decisively: 90% of capacity deliveries are on or ahead of schedule, customers are happy, and gross margins actually came in above guidance at 32% versus the expected 30%.
The stock was still well below its September highs — Oracle was briefly approaching a $1 trillion valuation before selling off sharply — but this earnings report signals the company is back on a credible upward trajectory.
How Fast Is Oracle's Cloud Infrastructure Growing?
The infrastructure segment is the engine of Oracle's entire bull case, and the growth numbers are genuinely impressive. Last quarter showed 68% growth. Analysts modeled 79% for this quarter. Oracle printed 84%.
07:40
Capex breakdown: Oracle spent $18.5B in Q3 vs the $14B analyst estimate
Watch at 07:40 →
To put that in context, Oracle is essentially attempting to build a third major hyperscale cloud platform — something in the same category as AWS, Azure, and GCP — at a time when demand for AI compute is unlike anything the industry has ever seen. The company spent $18.5 billion in capex in the fiscal third quarter alone, well above the analyst estimate of $14 billion. Total capex for the current fiscal year is tracking toward $50 billion.
Larry Ellison is very clearly all-in. The pocketbook is open. The couch cushions have been checked for coins. And so far, the bet is paying off.
Is Oracle's AI Infrastructure Actually Profitable?
This was one of the most important questions heading into earnings, and the answer is yes — Oracle's AI infrastructure is profitable the moment it comes online. There was a wave of bearish commentary earlier this year suggesting Oracle was in financial trouble because of the economics of its infrastructure bets. The argument went something like: Oracle has to pay Nvidia for GPUs a year in advance, it takes time to rack and deploy them, and in the meantime you've got a cash flow squeeze with no revenue to show for it.
That's technically accurate as a description of how infrastructure buildouts work. It's also business 101. And the actual data shows it isn't the disaster the bears were predicting. Gross margins came in at 32%, above the 30% guidance. The infrastructure is profitable on day one of deployment. Oracle is not being forced into dangerous financial engineering — no massive equity issuance, no spiraling debt load. The structure of the deals, heavily weighted toward long-term contracted revenue (more on that below), gives the financials a stability that makes the aggressive capex spending look disciplined rather than reckless.
14:20
The S-curve problem: consumer AI user growth vs exponential token consumption growth
Watch at 14:20 →
What Is Oracle's RPO and Why Does It Matter?
RPO stands for Remaining Performance Obligations — essentially the value of contracts Oracle has signed that haven't been recognized as revenue yet. Think of it as the order book. It's one of the most forward-looking indicators of where a company's revenue is actually heading, and Oracle's RPO number is staggering.
Oracle just increased its RPO backlog to $553 billion. For reference, when the initial excitement about Oracle's AI buildout first hit last year, the RPO number that got everyone excited was around $500 billion — which critics at the time compared to the size of AWS. Oracle has now grown that backlog even further, which means the demand picture is strengthening, not weakening. Customers aren't walking away. They're signing more contracts. That's the foundation of the bull case.
Why Are OpenAI and Meta Choosing Oracle Over AWS?
This is one of the most interesting strategic dynamics in the current AI infrastructure race. Amazon, Microsoft, and Google all have genuine hyperscale cloud platforms — AWS, Azure, and GCP. They offer true cloud infrastructure with all the flexibility, tooling, and global reach that comes with running a platform at that scale.
But OpenAI and Meta have massive, specific AI compute needs that go beyond what they can easily absorb through those platforms alone. Meta is a hyperscaler in the sense that it builds enormous data centers — but it doesn't operate a public cloud platform with the flexibility of AWS. OpenAI doesn't have its own cloud infrastructure platform at all. Both of them need burst capacity — a place to rapidly scale up compute for training and inference without having to build every data center themselves from scratch.
Oracle has positioned itself as exactly that: an extra burst of capex capacity in the system, purpose-built for the AI era. The deals Oracle signed with both OpenAI and Meta were viewed skeptically by some analysts who worried about execution risk. Those deals are now apparently going better than expected, which is a big part of why the market responded so strongly to this earnings report.
Why AI Agents Are Causing Token Consumption to Explode
One of the most important — and underappreciated — dynamics in the AI infrastructure story right now is the disconnect between user growth and compute growth. Consumer AI adoption is starting to look like a classic S-curve. OpenAI is reportedly past a billion monthly active users. You can't expect 10 billion ChatGPT users next year because there aren't 10 billion people with reliable internet access. Growth will decelerate.
But compute demand is behaving very differently. Here's why:
- Reasoning models like OpenAI's o1 use roughly 10x more tokens per query than earlier LLMs.
- GPT-5's release drove significantly higher usage rates of those more compute-intensive reasoning models.
- AI agents — tools like Codex, Claude Code, and the open-source agent frameworks — represent another step-change in token volume. Instead of one person typing one query, you have one person running an agent that works autonomously all day, generating a continuous stream of tokens.
The result is a situation where user growth is decelerating while compute demand continues to grow exponentially. GPU fleets are being melted. Oracle's infrastructure pipeline is full. This is the core reason the infrastructure buildout story remains so compelling even as the consumer AI narrative enters a more mature phase.
Who Is Larry Ellison and Why Is He Betting Everything on AI?
To understand why Oracle is moving so aggressively, you have to understand Larry Ellison — a man who in 1997 was heckling Michael Jordan from the seventh row at the United Center, had Rupert Murdoch making him coffee on a racing sailboat (Murdoch famously lost a fingertip in a rope accident during a race off Australia and still made the afterparty), and was simultaneously trying to take over Apple Computer to use it as a weapon against Bill Gates.
Ellison has always operated at full tilt. He's not a man who hedges when he believes in something. And right now, he clearly believes that the demand for AI compute infrastructure is the defining economic opportunity of this era — essentially that the market needs another AWS built at speed, and Oracle is going to be the one to build it.
There's also an interesting family dynamic at play. While Larry is going all-in on the infrastructure side through Oracle — betting on the picks-and-shovels of the AI gold rush — his son David Ellison has been assembling a massive media IP portfolio through his acquisition of Paramount and Warner Brothers Discovery. Larry is betting on the future of compute. David is betting on the enduring value of intellectual property. Together, the Ellison family is arguably long both generative AI and the content that generative AI might threaten.
Whether that's a hedge or just two very large, very confident bets running in parallel is a matter of perspective. But based on Oracle's latest earnings report, at least the infrastructure side of that family trade is looking very good right now.
