Oracle Revealed a $638 Billion AI Backlog — and Cloud Infrastructure Jumped 93%

With the AI Rally Wobbling, Oracle Shoved Real Demand Onto the Table

Here's the deal: on June 10, 2026, Oracle reported FY2026 Q4 — and it was an answer to the market's doubt. The week before, a Broadcom guidance miss had crushed AI chip stocks and left "is AI-infrastructure demand real or a bubble?" hanging over everyone. Right in the middle of that, Oracle's numbers screamed "the demand is real."

Start with the headline figures. Q4 total revenue was $19.2 billion (+21%), total cloud revenue $9.9 billion (+47%), and cloud infrastructure (IaaS) revenue $5.8 billion — up a stunning 93%. The compute needed to run AI training and inference is selling like crazy.

But the real shock was remaining performance obligations (RPO) — future revenue that's contracted but not yet booked. It hit $638 billion at quarter-end, up 363% year over year, a $85 billion jump from the prior quarter's $553B in just three months. Oracle said most of that increase was large AI contracts — deals where customers prepaid for GPUs or supplied their own, with the prepaid/customer-supplied hardware portion alone totaling $75 billion.

The Cast — Oracle, OCI, and Safra Catz

First, Oracle — once known only as "the database company," now reinvented as an "AI infrastructure provider." The weapon is OCI (Oracle Cloud Infrastructure). Long treated as the perennial No. 4 behind AWS, Azure, and Google Cloud, OCI rode the AI compute explosion to become one of the fastest-growing clouds. IaaS +93% is the proof.

Second, the RPO metric itself. For a stock, "how much is contracted for the future" matters more than "how much you booked now." RPO of $638B means $638 billion of contracts queued to be recognized as revenue over time. Up 363% says AI-infrastructure demand isn't a passing fad — it's getting locked into long-term contracts.

Third, Safra Catz, Oracle's CEO — famously conservative and numbers-strict. Oracle disclosing a $638B AI backlog under her carries weight. And underpinning it all is the whole AI-infrastructure supply chain: Nvidia supplying GPUs, big-tech customers buying them. Oracle's results are a window showing that entire chain is alive.

The Quarter, by the Numbers

Item	Figure	YoY
Q4 total revenue	$19.2B	+21%
Total cloud revenue	$9.9B	+47%
Cloud infrastructure (IaaS)	$5.8B	+93%
RPO (backlog)	$638B	+363%
RPO quarterly increase	+$85B	—
Q4 EPS (non-GAAP)	$2.11	+24%
Full-year IaaS revenue	$18.1B	+77%

The key point: current revenue is strong, but future revenue is stronger. IaaS +93% proves "selling well now"; RPO +363% proves "years of work are locked in." That customers prepaid for GPUs makes it credible — not mere "interest" but confirmed demand paid for in advance.

The timing was uncanny. The prior week (June 3–4), a Broadcom guidance miss erased over $1 trillion of AI chip value and the market was gripped by "bubble" fears. Oracle's results were the first test of that doubt — and by proving "demand is real," they became a decisive catalyst for the chip/infrastructure rebound.

Who Gains What

For Oracle, this is the knockout punch that sheds the "perennial No. 4" label. Riding the AI compute explosion, OCI now grows faster than the Big Three in places. RPO of $638B means revenue is contractually set to climb for years, lifting both multiple and negotiating leverage.

For Nvidia and the chip chain, Oracle's results are a "demand certificate." IaaS up 93% with customers prepaying for GPUs means money flows to Nvidia, Broadcom, and memory makers selling those chips. After the prior week's "is AI chip demand cooling?" panic, Oracle showing "no — we're getting prepaid" reassured the whole chain.

For enterprise AI customers, it cuts both ways. Good: aggressive infrastructure buildout improves compute availability. Heavy: "GPU prepayment" becoming standard shows just how scarce infrastructure is — AI compute remains a supplier's market where you pay upfront and stand in line.

Past Parallels — Cloud Transitions

The archetype of success is Amazon AWS: an e-commerce company that rode the cloud wave first to dominance, with AWS profit powering all of Amazon. Oracle's OCI is the same pattern on the "AI compute" wave — connecting an existing strength (databases) to AI infrastructure is working.

But plenty missed the wave too — IT veterans that failed to transform in the mobile/cloud era and faded. The difference was whether they laid infrastructure ahead of demand. The opposite risk exists too: over-invest on overconfident demand and weak utilization becomes a crushing fixed-cost burden. A huge RPO means nothing if you stumble converting it to revenue.

The lesson: $638B of RPO is "promised future revenue," not "money already made." The real test is converting that backlog into revenue and profit without a hitch. AWS pulled it off; some over-invested and wobbled. Whether Oracle secures GPUs, power, and data centers in time to cash this backlog is the thing to watch.

Competitor Counterplay

The most direct rivals are AWS, Microsoft Azure, and Google Cloud. If Oracle touts "fastest AI-infra growth," the Big Three counter with overwhelming scale, custom silicon (Google TPU, Azure's chips), and broad enterprise bases. Their "full-stack" play — bundling their own AI models with cloud — lets them defend with ecosystem lock-in even if Oracle pushes on price and availability.

Emerging GPU-cloud players (the CoreWeave type) are a wildcard. They grew fast as "AI-native clouds"; Oracle entering at greater scale and enterprise trust intensifies competition. Their counter: more aggressive pricing, faster GPU supply, workload-specific optimization. It's all a fight over "who supplies AI compute cheaper, faster, more reliably."

The interesting twist: this may not be zero-sum. AI-infra demand is growing so fast that Oracle's RPO surge also signals the pie itself is expanding — Big Three, newcomers, and Oracle can all grow at once. The winner will be decided by how long that growth lasts and who converts backlog to revenue most efficiently.

So What Changes — By Reader

For investors, you've got first evidence that AI-infra demand is real. After the prior week's crash stoked bubble fears, RPO +363% and GPU prepayments show confirmed demand. But the crux is converting RPO into revenue and profit — track how smoothly the backlog cashes out each quarter. Promised and realized revenue are not the same.

For cloud/AI-infra professionals, the takeaway is that AI compute remains a supplier's market. Customers prepaying for GPUs means continued demand across data centers, power, cooling, and GPU supply — worth checking where your niche sits in that flow.

For everyone else, the big picture: the AI boom is reaching not just chipmakers but infrastructure providers in real earnings. Nvidia sells chips, Oracle deploys and services them, and that demand loops back into chip orders. Whether the cycle persists depends on whether AI actually generates enough profit — and Oracle's results are one snapshot that it's still alive.

🥄 Three Things You're Probably Wondering

— So what does this mean for me? No direct impact. But if you hold Oracle, Nvidia, or a chip ETF, read it as a signal the prior week's crash was overdone. And if you use AI services, compute infrastructure is growing fast, with gradual effects on availability and price.

— If RPO is $638B, isn't that just money already made? No. RPO is future revenue that's contracted but not yet booked — recognized over years as services are delivered. So the real question is the ability to convert that backlog into actual revenue and profit without a hitch. Promise and realization differ.

— Does this end the AI-bubble debate? Too early. Oracle strongly showed "demand is real," but the bubble debate's core is whether this massive investment returns enough profit. Confirmed demand and proven profitability are separate — we need a few more quarters of conversion results for an answer.

References

• Oracle Announces Record Q4 and FY 2026 Results — Oracle Investor Relations
• Oracle Q4 Earnings Land June 10: Record $553 Billion AI Backlog Faces Its First Conversion Test — TechTimes
• Oracle Earnings Pose Next Test for Suddenly Shaky AI Stock Rally — Bloomberg
• Oracle Announces Record Q4 and FY 2026 Results — StockTitan

Numbers and criteria are as of announcement and may change. Investment calls are yours to make!