Oracle’s AI Infrastructure Drives 93% Infra Growth, $638B Backlog

For the AI community, Oracle’s Q4 report is a landmark signal: enterprises are locking in long-term GPU capacity deals at a staggering pace. The $75 billion in prepaid GPU commitments underscores the critical shortage of AI compute and Oracle’s emergence as a serious alternative to the hyper-scale clouds. But the heavy capital outlay and margin pressure raise the question: can Oracle sustain this AI infrastructure arms race without diluting returns, or will it become a cautionary tale of over-expansion?

Oracle Corporation reported its fourth quarter fiscal year 2026 results on June 14, 2026, posting revenue and earnings that surpassed consensus estimates, yet shares dipped as investors fixated on margin pressure and the massive capital expenditure required to fuel its cloud infrastructure expansion. Revenue climbed 21% year over year to $19.18 billion, exceeding the $19.1 billion analyst consensus compiled by LSEG. The standout performer was cloud revenue, which surged 47% to $9.9 billion. Within this, cloud infrastructure (IaaS) revenue skyrocketed 93% to $5.8 billion, while cloud application (SaaS) revenue grew a more modest 10% to $4.1 billion. Traditional software revenue, however, continued its slow erosion, falling 2% to $6.8 billion. Adjusted earnings per share rose 20% to $2.03, beating the $1.96 consensus.

“

Within this, cloud infrastructure (IaaS) revenue skyrocketed 93% to $5.8 billion, while cloud application (SaaS) revenue grew a more modest 10% to $4.1 billion.

The most staggering figure was the cloud computing backlog, which ballooned 363% to $638 billion. Management noted that most of the recent backlog increase came from customers prepaying for or supplying their own GPUs, a clear indicator of insatiable AI-driven demand. This prepaid GPU portion now accounts for $75 billion of the backlog. While the backlog points to robust future revenue, it also underscores the enormous upfront costs Oracle is incurring—building data centers, acquiring cutting-edge hardware, and depreciating those assets over time. This dynamic initially depresses gross margins, as costs are recognized before the full revenue stream kicks in, fueling the bear case against the stock.

Investor reaction was swift and negative, sending Oracle shares lower despite the beats. The stock has become a battleground in the AI narrative, with bulls cheering the extraordinary backlog and bears warning of a cash-burn spiral and uncertain returns on infrastructure investments. Year-to-date, Oracle is down only about 5% and up roughly 5% over the past year, reflecting this deep divide.

Looking ahead, Oracle maintained its full fiscal year 2027 revenue forecast of $90 billion and issued upbeat guidance for fiscal Q1 2027: revenue growth of 27–29% and cloud revenue growth of 57–63%. These targets suggest that management expects the capacity buildout to begin paying off rapidly. Yet the market remains skeptical, having been burned before by capital-intensive cloud transformations that take longer than promised to reach profitability.

Contextualizing these results within the broader cloud industry, Oracle is carving a unique path. Unlike the hyperscale triumvirate of AWS, Microsoft Azure, and Google Cloud, Oracle is leveraging its entrenched enterprise database and application footprint to offer high-performance, AI-optimized cloud services, often with a multi-cloud twist. The autonomous database and Exadata infrastructure give it a differentiated value proposition, particularly for mission-critical workloads. The $638 billion backlog—an almost surreal number—suggests that large enterprises are committing to Oracle’s vision en masse, especially for AI training and inference.