The market's enthusiasm for artificial intelligence (AI) has lifted just about every name with a credible foothold in the build-out. Alphabet and Amazon have hit fresh all-time highs as their cloud divisions accelerate. AI chipmaker Nvidia is close to a $5 trillion market capitalization, and Broadcom and Intel have joined the rally; Intel alone climbed 114% in April.
And then there is Oracle (ORCL 0.25%). Shares have rebounded some in the past week, rising about 10%, but they remain far below their 52-week high of more than $345. And even with the stock's recent rebound, it's down more than 6% year to date as of this writing.
That is an unusual look for a company sitting on more contracted future revenue than just about any rival.
So what gives? A heavy debt load funding a build-out largely tied to one anchor customer -- a posture that contrasts with the more diversified backlog Amazon described in its latest earnings call.
Image source: The Motley Fool.
A backlog like nothing the company has seen
The most recent quarter -- Oracle's fiscal third quarter of 2026 (the period ended Feb. 28, 2026), reported in March -- was, by the database giant's own description, the strongest in more than 15 years. Total revenue rose 22% to $17.2 billion, with cloud revenue jumping 44% to $8.9 billion. The standout was infrastructure-as-a-service -- Oracle's offering that rents out raw computing power for AI training and inference. That business grew 84% year over year -- an acceleration from 68% in the prior quarter.
Then there is the contract backlog, which Oracle calls remaining performance obligations, or RPO. It ended the quarter at $553 billion, more than four times the level of a year earlier and a $29 billion step-up from just three months prior. For perspective, that backlog is larger than Oracle's entire market capitalization of about $525 billion as of this writing.
Additionally, Oracle's AI infrastructure revenue alone surged 243% in the quarter.
Turning to profitability, Oracle's gross margin on that capacity came in at 32% -- well ahead of the 30% management had guided to. Further, non-GAAP (adjusted) earnings per share rose 21% to $1.79, and the company nudged its fiscal 2027 revenue forecast to $90 billion from $89 billion. It was Oracle's first quarter since 2009 in which both organic revenue and adjusted earnings per share grew by 20% or more in U.S. dollars.
A build-out funded with debt -- and one anchor tenant
But here's where many investors likely have concerns.
Most of that swelling backlog ties back to a small group of AI customers, and one looms very large. Oracle's roughly $300 billion, five-year compute deal with OpenAI -- announced last September as part of the Stargate joint venture-- is the linchpin of the bull case and, increasingly, the source of the bear case too.
When the Wall Street Journal reported in late April that OpenAI had missed internal revenue and user targets, and that the start-up's chief financial officer had voiced concern about its ability to fund future compute commitments, Oracle shares unsurprisingly fell sharply.
NYSE: ORCL
Key Data Points
The other complication is how the build-out is being paid for.
Oracle is on track to spend roughly $50 billion in capital expenditures in fiscal 2026, more than double its fiscal 2025 spending.
It's funding this build-out with the help of significant debt.
Oracle's long-term debt sat at about $125 billion exiting the quarter -- and trailing-12-month free cash flow has swung deeply negative as the company pours money into data centers ahead of the revenue.
With this said, Oracle executive Doug Kehring has emphasized that much of the new AI capacity is supported by customer prepayments or customer-supplied chips, which softens the blow to the balance sheet.
But this is still a risky setup.
That backdrop helps explain why the stock has lagged its soaring AI peers. Cloud rivals like Alphabet and Amazon are funding their AI buildouts mostly from internal cash flow, while chip suppliers like Nvidia and Broadcom are selling into the boom rather than borrowing into it.
Oracle is doing something ambitious -- and arguably riskier -- by leveraging up to lock in long-dated compute contracts whose economics may or may not play out as projected.
For investors who believe the AI build-out is still in its early innings, this tension could be the opportunity. After all, the backlog is real -- and demand from AI customers has not yet shown signs of cooling in Oracle's own results. Further, shares now sit well below their September peak last year. But for investors who believe in Oracle and want to buy this dip, being measured makes sense given the risks.
