Oracle says AI customers are footing the bill as it posts fastest growth in a decade

Oracle Corp. delivered one of its strongest quarters in years—and used it to argue that the soaring costs of the artificial intelligence boom are increasingly being shouldered by its customers rather than its balance sheet.

Strong quarter, bigger forecast

Oracle on March 10 reported fiscal third-quarter 2026 revenue of $17.2 billion, up 22% from a year earlier, its fastest organic growth in more than a decade. Cloud infrastructure sales surged 84% to $4.9 billion, driven by AI workloads, and the company raised its revenue target for fiscal 2027 to $90 billion.

Shares rose roughly 8% to 10% in after-hours and next-day trading, recouping some losses from late last year when heavy AI spending and cautious guidance prompted concern that Oracle’s build-out was becoming too costly.

Customers prepay—or bring their own GPUs

Behind the headline numbers, Oracle described a funding approach that differs from the traditional hyperscale model in which cloud providers finance and own most of the underlying hardware.

In a filing with the Securities and Exchange Commission, Oracle said much of the increase in its remaining performance obligations—contracted revenue not yet recognized—came from “large scale AI contracts,” and that it does not expect to raise “incremental funds” to support many of those deals.

“Most of the equipment needed is either funded upfront via customer prepayments so Oracle can purchase the GPUs, or the customer buys the GPUs and supplies them to Oracle,” the company said.

The result is a bring-your-own-hardware (or pay-ahead) structure that shifts a meaningful portion of the upfront infrastructure burden to customers.

A record backlog under new terms

Oracle ended the quarter with $553 billion in remaining performance obligations, up 325% from a year earlier and $29 billion higher than the prior quarter. The company said “most” of that increase stemmed from AI contracts.

On a conference call, co-Chief Executive Clay Magouyrk said Oracle and its partners have secured more than 10 gigawatts of power and data-center capacity over the next three years, with “greater than 90% of that capacity fully funded through our partners.”

Magouyrk added that since the previous quarter, Oracle has signed more than $29 billion of contracts using new commercial structures combining customer prepayments and customer-supplied hardware. Those agreements, he said, allow the company to expand AI capacity “without any negative cash flow from Oracle Corporation.”

In the three months ended Feb. 28, Oracle delivered more than 400 megawatts of AI capacity, with roughly 90% turned on as scheduled or earlier. Magouyrk put gross margins on AI capacity at about 32%, above prior guidance of more than 30%.

The approach stands in contrast to providers such as Amazon Web Services, Microsoft Azure and Google Cloud, which typically fund and own most data-center equipment and then sell compute by the hour or second.

It also represents a change for Oracle. In the prior quarter, the company disclosed negative free cash flow of around $10 billion amid a spike in capital spending for AI infrastructure.

Funding a capex surge

Oracle is still preparing for a heavy spending cycle, reiterating that it expects capital expenditures of about $50 billion in fiscal 2026, spanning data centers, networking and AI accelerators.

To help fund the build-out, Oracle in February launched a program to raise up to $50 billion through a mix of debt and equity securities. Within days, it sold roughly $30 billion of investment-grade bonds and mandatory convertible preferred stock, describing demand as “substantially oversubscribed.” The company said it does not expect to issue additional bonds in calendar 2026 beyond that program.

Executives argue that customer prepayments, partner financing for power and real estate, and customer-supplied GPUs reduce the amount Oracle must commit per dollar of AI revenue.

The strategy comes as cloud and AI infrastructure spending rises sharply. Some analysts estimate combined capital expenditures by the largest cloud and AI providers could exceed $600 billion in 2026. Microsoft has reported quarterly capex topping $30 billion as it builds AI capacity, while Alphabet has guided for as much as $185 billion in capex in 2026, much of it tied to AI chips and data centers.

Oracle’s spending is smaller and it remains behind the top three hyperscalers in market share. But if its model holds, it could expand capacity at a lower net cash cost than some larger rivals.

A long-term bet on AI demand—and SaaS staying power

Oracle now expects to generate about $67 billion in revenue this fiscal year and about $90 billion in fiscal 2027, up from $57.4 billion in fiscal 2025.

Executives also described a “halo effect” from AI workloads: customers training or running models on Oracle Cloud Infrastructure are more likely, they said, to adopt Oracle databases and business applications.

Oracle reported cloud database services revenue grew 35% in the quarter, and said multicloud database revenue—Oracle databases running in other providers’ data centers—rose more than fivefold from a year earlier.

Addressing concerns that generative AI could erode demand for traditional software-as-a-service products, co-CEO Mike Sicilia said customers are not seeking to replace mission-critical systems with improvised AI tools.

“We have not met customers who plan to replace mission-critical systems with cobbled-together AI tools,” he said, citing core banking platforms, hospital electronic health records and large retail merchandising systems. Instead, he said, companies want AI embedded “out of the box.”

Sicilia said Oracle has deployed about 1,000 AI agents inside its Fusion applications suite and hundreds more inside its banking software as part of routine upgrades. Executive Chairman and Chief Technology Officer Larry Ellison told analysts that “the SaaS apocalypse applies to others, but not to us,” arguing AI features will strengthen Oracle’s software franchise.

Power, politics and the cost of growth

Oracle’s expansion also carries implications for energy systems and policy. Building and operating 10 gigawatts of data-center capacity—roughly comparable to the output of several large nuclear power plants—requires vast amounts of electricity amid intensifying debate over how AI-driven demand should be met and who should bear the cost.

In March, Oracle signed what it called a voluntary “ratepayer protection pledge,” committing to structure power costs so they are not recovered through higher electricity rates for residential customers—an apparent response to scrutiny from utility regulators and local officials.

Oracle is also involved in politically sensitive projects, including a stake in the restructured U.S. operations of TikTok and participation in an AI infrastructure initiative known as Stargate announced in 2025. Alongside sovereign cloud offerings for governments, those efforts place Oracle at the intersection of technology, national security and regulation as it scales AI.

A test of a new funding model

For now, Oracle has persuaded customers to sign long-term AI contracts under terms that shift more upfront cost to them, while investors appear encouraged by the company’s growth and guidance.

Whether the model endures may hinge on factors outside Oracle’s control, including AI adoption rates, next-generation chip availability, the health of corporate and government IT budgets, and regulators’ response to data centers’ growing strain on power grids.

Oracle’s message is that it can compete in the AI arms race without matching the most balance-sheet-intensive spending of its larger rivals. The next few years will test whether customer prepayments and customer-supplied hardware can keep pace with the company’s ambitions.