5 Things To Watch Amid Underwhelming Oracle Quarter, Massive AI Buildout

When Oracle’s AI buildout translates into revenue and what it does if OpenAI fails to pay are among the questions.

Oracle shows no signs of halting the billions of dollars needed for data center buildouts to meet artificial intelligence demand.

But for solution providers looking for whether the AI revolution means money to be made today or whether the technology needs more time to make business sense, Oracle’s spending and experience turning business pipeline into revenue offers some insight.

Oracle’s AI Data Center Buildout

When Oracle expects to see revenue from its AI buildout, the ultimate cost of the AI buildout, whether it is experiencing company-specific execution issues and contingency plans should one of its biggest AI customers, OpenAI, fail to meet payments for its $300 billion contract with Oracle are some of the questions that remain as Oracle and its ecosystem enter 2026.

Austin, Texas-based Oracle, of course, isn’t alone in massive capital expenditures to meet AI demand and the promise of future AI revenue. Amazon has spent $89.9 billion so far this year, most of that going toward Amazon Web Services for AI and core services and Trainium and custom chips. Google expects to spend more than $90 billion this year in CapEx, and Microsoft expects around the same in spending in 2026, according to CNBC.

Here is more on what the channel should watch when it comes to Oracle’s AI buildout.

When Will AI Buildout Yield Revenue?

While Oracle dazzled a quarter ago with a large amount of remaining performance obligation—contracted business pipeline set to become future revenue—its new RPO of $523 billion came in below Wall Street’s expected $526 billion-plus, according to a KeyBanc report.

That RPO is up fivefold year on year and 15 percent—$68 billion—up quarter on quarter.

Spending to meet AI demand should become AI revenue for Oracle next year, according to a Bank of America report. The investment firm said that now is the heaviest phase of Oracle’s AI infrastructure buildout and that the infrastructure has to get built for Oracle to deliver capacity.

Although Oracle Co-CEO Clay Magouyrk said that Oracle could try to bring down the cost of AI buildouts through vendor financing, chip leasing and customers bringing their own chips to Oracle’s data center, these strategies undermine confidence in Oracle converting that backlog into revenue if it’s not reselling the underlying GPUs, according to a Morgan Stanley report.

Oracle usually spends capital a quarter or a little more before it captures the revenue from that spending, Bernstein said in a report. It is possible Oracle spends most of the $12 billion capital expenditures expected for fiscal year 2026 in the final quarter, with revenue appearing in the first or second quarter of fiscal year 2027.

How Expensive Is The AI Buildout?

Higher-than-expected spending on the AI buildout ate into Oracle’s earnings before interest and taxes, with the measure not using GAAP coming in 40 basis points below the 41.9 percent KeyBanc expected, according to a report by the investment bank.

Oracle’s negative free cash flow of about $10 billion was worse than the $5.2 billion expected due to higher capital spending, Bernstein said in a report.

Oracle spent $12 billion in capital during the quarter, $3.6 billion more than Wall Street expected, according to KeyBanc. That amount is higher than what Wall Street modeled for the remaining quarters in Oracle’s fiscal year.

Although Oracle’s leaders said to expect faster revenue recognition on Oracle Cloud Infrastructure—the vendor increased fiscal year 2027 revenue expectations by $4 billion, reaching about $89 billion—the cost of that work is a $15 billion increase in capital expenditures than originally forecast.

Oracle went from $35 billion in expected CapEx this year to about $50 billion, about 75 percent of fiscal year 2026 revenue. Oracle’s capital spending in fiscal year 2027 is now expected to hit $51.4 billion and then $68.7 billion in fiscal year 2028, according to KeyBanc.

That $15 billion in CapEx generates $2.5 billion a year in amortized assets given the six-year useful life of GPU and CPU servers, according to a Bernstein report. The firm estimates that Oracle could generate $24 billion in revenue and $7.2 billion in profit from a six-year rental. The total contract value could even reach upward of $30 billion.

Magouyrk said Oracle should need substantially less than the $100 billion in additional capital he has seen some analysts toss out as needed to fund AI infrastructure commitments. The implication there is that Oracle pursues equity raises and non-balance-sheet funding vehicles, William Blair said in a report.

It could take time for Oracle executives to give yearslong detailed cost models due in part to the fast pace of change for AI infrastructure technology, with uncertainty around what GPU will be deployed in three years’ time, Bernstein said in a report. It’s also likely that the cost of AI buildouts comes down over time and that Oracle is constantly negotiating equipment and capital supply contracts.

Oracle contracts actually include protection clauses if technology or pricing negatively affect margins, according to the investment firm.

Is Execution An Issue?

Oracle only recently set up expectations going into the quarter following gigantic new cloud contracts including from ChatGPT maker OpenAI, potentially calling into question whether Oracle can execute on the revenue promise of AI, KeyBanc said in a report.

Oracle’s results in its latest quarterly results missed in multiple areas, according to a KeyBanc report. Total revenue for the second fiscal quarter was $16 billion when KeyBanc expected closer to $16.1 billion, according to a report by the investment firm. With Oracle reiterating near-term revenue forecasts, this adds greater pressure onto Oracle delivering the next two quarters.

The miss could be due to delays at Oracle’s data centers in Abilene, Texas, William Blair said in a report.

Infrastructure as a Service reached $4.1 billion and not $4.2 billion as Wall Street expected. Cloud revenue came in at $8 billion when Wall Street expected $8.1 billion—33 percent growth year on year when investors wanted 34 percent, KeyBanc said. Oracle’s expected cloud revenue for the next quarter is about 39 percent, 290 basis points below the estimate of 41.9 percent Wall Street wanted.

Cloud and software gross margin of 72.3 percent was below KeyBanc’s 72.8 percent estimate and will continue to serve as a key metric as Oracle layers in the cost of AI buildouts.

Still, the missed revenue measures could just be an investment curve issue rather than a reflection of changing fundamentals, Bank of America said in a report. Growing AI demand means growing AI spending to have capacity for customers. The vendor’s various financing options for the buildout while pledging to maintain investment-grade credit also eased the investment firm’s concerns.

Oracle executives pointed to a sales force reorganization worldwide as a way to encourage cross-selling and moving on-premises workloads to the cloud and higher revenue lift, but a reduced sales capacity could get in the way of Oracle selling more first-party products as opposed to training and inference compute, according to a Morgan Stanley report.

NetSuite’s 13 percent growth year on year marks a deceleration from 15 percent the prior quarter. Cloud database services growth year on year of 30 percent marks a deceleration from 32 percent growth year on year the prior quarter, according to the investment firm.

What If OpenAI Fails To Pay?

As concerns of an AI bubble form heading into 2026, Oracle’s leadership said it feels comfortable with OpenAI’s ability to meet its $300 billion contract with the cloud provider.

And should OpenAI fail to meet its contract terms, Oracle can give its data center capacity to another customer, Magouyrk said on the call.

Oracle can spin up a bare-metal computer in a few minutes and then recycle the computer for another customer in less than an hour, the co-CEO said. The large model providers might take two or three days to spend the capacity Oracle gives them, reflecting the need for quick buildouts.

The vendor usually doesn’t purchase data center hardware, a project’s largest cost, until months before revenue comes in. If any major customer experiences issues, Oracle could sublease or warehouse the data center shell, redeploy capacity to other customers or cancel the hardware outright before going live, according to a Bernstein report.

What Are The Positives?

Still, Oracle executives had some good news to share, especially for solution providers still banking on increased cloud work ahead.

Current remaining performance obligation (cRPO) was $53.1 billion, up 40 percent year on year. That marks an acceleration over the 21 percent growth year on year reported the prior quarter, which saw cRPO of about $45.5 billion. The implication is that Oracle saw cRPO bookings of $23.7 billion, up 65 percent year on year and an acceleration over the 17 percent growth year on year from the prior quarter, according to Morgan Stanley.

Cloud infrastructure revenue grew 66 percent year on year. Multi-cloud database revenue grew ninefold and Oracle is at work on another 25 sites.

The vendor saw GPU-related revenue more than double and revenue from non-AI segments of new enterprises, sovereigns, dedicated regions that run Oracle Cloud Infrastructure racks on-premises and deployments of the Alloy cloud infrastructure platform at partner-managed sites, according to a William Blair report.

The investment firm said Oracle’s total cloud growth of 35 percent accelerated from about 25 percent for the same period a year ago, and William Blair expects the growth to accelerate again to 42 percent in the next quarter with more capacity available.

Oracle did also show some progress in diversifying its cloud computing customer list, with leaders discussing contracts with Facebook parent Meta and chip giant Nvidia on the call. Those two companies contributed to the $68 billion added to Oracle’s RPO.

Oracle’s application revenue accelerated to 11 percent, up 200 basis points, potentially showing the early signs of an AI halo effect, Bank of America said in a report.