Oracle’s Cloud Business Begins Paying Off Despite Heavy Capital Spending

Oracle’s latest quarter set records for revenue and marked an unusual milestone: for the first time in its five-decade history, the first quarter of one fiscal year was larger than the fourth quarter of the preceding fiscal year. This differs from the typical cycle in which sales teams concentrate deals in the fourth quarter, often making the following first quarter weaker.

Sequential growth was modest at 0.8 percent, but the change in pattern is significant. Oracle ended the quarter with a revenue backlog of $664 billion, suggesting that its cloud infrastructure business could grow at an accelerating rate as backlog converts into sales.

“We would expect acceleration to continue in the remainder of fiscal 2027 as we convert more RPO into revenues,” Oracle chief financial officer Hillary Maxson told Wall Street analysts during a call covering the first quarter of fiscal 2027. “We now expect around half of our RPO to convert into sales over the next 36 months.”

A simple average would imply approximately $27.7 billion per quarter, although the amount would be lower in fiscal 2027 and higher in fiscal 2029. The remaining half of the backlog would extend beyond fiscal 2029.

Revenue and capital spending

Oracle reported quarterly revenue of $19.18 billion, up 29.6 percent year over year and slightly higher sequentially. Operating income increased 57.3 percent to $6.73 billion, while net income rose 62.6 percent to $4.76 billion.

Under more typical conditions, Oracle’s improving net-income-to-revenue ratio would likely attract considerable attention. However, investors are also focused on the company’s planned capital spending of $90 billion to $95 billion in fiscal 2027.

Oracle raised $19.9 billion through floating stock. It also received $11.4 billion in prepayments, largely from cloud customers, contributing to quarterly cash flow of $23 billion. Capital expenditures reached $28 billion, leaving cash flow negative by $5 billion.

Cloud growth leads Oracle’s results

Oracle’s cloud division generated $11.61 billion in revenue, up 61.5 percent year over year and 17.1 percent sequentially.

Software license revenue followed its usual first-quarter decline, falling 65.2 percent sequentially and 14.5 percent year over year as customers continue moving from perpetual licenses and support contracts to cloud subscriptions. Software support revenue, an important recurring source of income, was $4.9 billion, down 1.2 percent.

Hardware revenue increased 15.5 percent year over year to $774 million, although it declined 16.2 percent sequentially. In the same quarter a year earlier, hardware revenue had risen 6.3 percent sequentially to $670 million.

Hardware operating margins are under pressure because component prices have increased during the generative AI boom. Demand has exceeded supply across much of the semiconductor market. Hardware operating income reached a low of 63.7 percent of revenue, compared with an average above 73 percent during the two years before component prices began rising.

Margins in the Cloud and Software group are also under pressure, likely reflecting higher prices for the equipment Oracle purchases for its cloud infrastructure and rising costs across data center construction.

Services revenue, which includes implementations for Oracle’s Fusion and NetSuite applications, along with support for legacy application environments, rose 4.8 percent to $1.41 billion.

Infrastructure growth and AI workloads

Oracle’s cloud infrastructure, or IaaS, revenue doubled to $7.4 billion during the quarter and increased 27.6 percent sequentially. SaaS revenue grew 9.2 percent to $4.2 billion.

Oracle’s cloud database business increased 26 percent to $900 million. Although the database could be viewed as SaaS because it is software, it also operates as systems software on highly tuned Exadata hardware. Within this business, the multicloud implementation of the Oracle database sold as an infrastructure service grew 4.5 times year over year, although it likely remains a small portion of total IaaS database revenue.

CPU and GPU capacity accounted for approximately 88 percent of Oracle’s cloud IaaS revenue, generating $6.5 billion, up 2.5 times year over year. This growth reflects the conversion of AI-related backlog into active workloads.

During the first quarter, Oracle added 850 megawatts of infrastructure capacity, including more than 300,000 GPUs and an unspecified number of CPUs available for hourly rental. That was three times the approximately 280 megawatts Oracle delivered in the fourth quarter of fiscal 2026.

Oracle added 1.16 gigawatts of capacity during the previous year. This included 400 megawatts in the third quarter of fiscal 2026, with another 480 megawatts installed during the first half of fiscal 2026.

OpenAI infrastructure deployments

The OpenAI Stargate buildout includes Oracle infrastructure in Abilene and Shackleford, Texas, with 3.2 gigawatts and 2.2 gigawatts respectively, as well as sites in Dona Ana, New Mexico, with 2.2 gigawatts, Port Washington, Wisconsin, with 1.3 gigawatts, and Saline, Michigan, with 1.4 gigawatts. Together, the sites are expected to total 8.1 gigawatts.

Oracle says it has secured 10 gigawatts of power for three years through fiscal 2029. These figures refer to data center power. Infrastructure power consumption would therefore be lower, estimated in the source at approximately 70 percent, or about 5.7 gigawatts.

If the capacity used Nvidia systems exclusively, delivering the infrastructure could cost approximately $227 billion, based on an estimated $40 billion per gigawatt discussed by Nvidia chief executive officer Jensen Huang. If the deployment combines OpenAI Jalapeno accelerators for inference with Nvidia GPUs for training, an estimated cost of $30 billion per gigawatt would put the hardware total at approximately $170 billion.

Clay Magouyrk, Oracle’s co-chief executive officer and head of Oracle Cloud Infrastructure, said 131,000 GPUs, almost certainly Nvidia “Blackwell” B200 and B300 systems, were installed at the OpenAI Abilene data center during the quarter. Another 69,000 were installed in the fourth quarter of fiscal 2026.

OpenAI has taken delivery of 618 megawatts, or 75 percent of the total infrastructure capacity Oracle is deploying at Abilene. The GPT-6 Astra model was trained on this infrastructure. The Shackleford data center is being equipped with Nvidia “Vera-Rubin” systems, with OpenAI expected to receive the initial capacity during the second quarter of fiscal 2027.

GPU utilization and pricing

GPU economics currently favor Oracle because demand for GPUs and other accelerators exceeds supply. Most GPUs in Oracle’s fleet, which may now contain 1 million or more units, are at least four years old. Nevertheless, they are commanding higher prices because demand for computing capacity remains greater than available supply.

The equipment could remain in service for three, four, or more years, and the source estimates that the GPUs may pay for themselves within three years. Across all GPU renewals during the first quarter, renewed contracts were priced 20 percent higher than the original agreements.

GPU utilization reached 97.9 percent in the first quarter, a level comparable to utilization associated with IBM mainframes.

Oracle’s use of AI in applications

Mike Sicilia, Oracle’s co-chief executive officer responsible for the company’s software business, said Oracle would announce a new agentic AI accelerator at its AI World event the following month. The system is intended to use generative AI to deploy Oracle applications more quickly and at lower cost than human implementation teams can achieve today. ERP suite implementation costs can be substantial.

“Working alongside Oracle and customer teams, AI agents will automate and orchestrate implementation at an unprecedented scale, compressing SaaS deployments from years to months and months to weeks,” Sicilia said. “It is really the power of these things together that reinforce my belief that the growth of our applications business is only going up from here.”

Customers running Oracle applications on OCI used embedded AI capabilities 150 million times during the first quarter, a 42 percent sequential increase. AI agents were launched more than 3.5 million times, nearly double the figure from the fourth quarter of fiscal 2026.

According to Sicilia, customers have more than 2,300 agents in production on Oracle applications. Customers using the high-end Fusion ERP suites consumed 900 billion tokens.

Oracle’s focus is on applying AI to its SaaS applications and expanding their capabilities. Sicilia described the intended shift in how employees interact with enterprise software:

“AI changes this dynamic. Rather than asking every employee to navigate and execute a process exactly as a system expects, AI agents can perform tasks using the organization's established workflows and business rules. Employees then shift to overseeing agents, resolving exceptions, and applying human judgment where it matters most.”