Oracle's Q4 2026 earnings were a record on almost every line, yet shares fell sharply after hours as investors focused on capital expenditures hitting $55.7 billion for the year and a financing plan that will keep the dilution and debt coming through fiscal 2027.
Oracle reported Q4 fiscal 2026 revenues of $19.2 billion, up 21% year-over-year, and non-GAAP earnings per share of $2.11, both clearing analyst estimates. Cloud infrastructure revenue came in at $5.8 billion for the quarter, up 93%. For the full fiscal year, Oracle posted $67.4 billion in total revenues, up 17%, and its remaining performance obligations climbed to a record $638 billion, rising from $553 billion just a quarter earlier. By any traditional measure, this was the earnings report Oracle's management team wanted to deliver.
And then the stock dropped more than 7% after hours.
The reason is straightforward: the capital expenditure number. Oracle spent $15.9 billion on capex in Q4 alone, pushing the full-year total to $55.7 billion. That exceeded the company's own prior guidance of $50 billion and generated $23.7 billion in negative free cash flow for the fiscal year. Depreciation nearly doubled to $7.62 billion. The message investors took away was that the cost of building Oracle Cloud Infrastructure is running significantly ahead of the revenue that infrastructure is actually generating right now.
To fund that buildout, Oracle has constructed one of the more unusual capital-raising programs in enterprise tech history. During fiscal 2026, the company raised $43 billion in debt and $5 billion in equity, totaling roughly $48 billion. Looking ahead to fiscal year 2027, Oracle has already signaled it expects to raise approximately $40 billion more, including through a previously announced $20 billion at-the-market equity issuance program, against guidance of $70 billion in net capital expenditure outlays for the coming year.
The split between debt and equity is deliberate. Oracle is using investment-grade senior unsecured bonds for the debt tranche and a combination of mandatory convertible preferred securities alongside the ATM equity program for the equity side. The mandatory convertibles are a classic tool for companies that need the economics of equity financing without the immediate dilution of a straight stock offering. By routing roughly half the raise through debt instruments rather than common stock, Oracle can preserve its investment-grade credit rating while still accessing the kind of capital that would normally require surrendering meaningful ownership.
This is the defining tension in Oracle's current story: it is trying to compete at hyperscaler spending levels while retaining an enterprise software company's financial profile. AWS, Microsoft Azure, and Google Cloud have balance sheets and recurring cash flows that can absorb $50 billion capex years without triggering investor alarm. Oracle does not yet have that cushion. So it is engineering its way around the problem through financing structure rather than organic cash generation.
The customer list is effectively a map of AI's current power structure
Oracle named the customers driving the cloud infrastructure buildout: OpenAI, Meta, Nvidia, xAI, AMD, and TikTok. That is not a list of companies shopping for competitive cloud pricing. That is the core roster of entities building or enabling the largest AI training and inference workloads on the planet in 2026. OpenAI's Stargate partnership with Oracle alone is targeting 4.5 gigawatts of U.S. data center capacity at an estimated $300 billion investment over time.
The presence of Nvidia on the customer list is worth pausing on. Nvidia sells the GPUs that go inside the data centers; it is also contracting for cloud capacity from Oracle. That circular relationship reflects the degree to which the AI infrastructure stack has become deeply interconnected. AMD's inclusion signals that Oracle is positioning OCI as a home for the challenger GPU ecosystem as well as Nvidia's dominant H-series hardware.
Whether OCI is becoming genuinely critical infrastructure or is primarily serving as overflow capacity for AWS, Azure, and Google is the open question. Multicloud database revenue grew 531% year over year, and management says demand for both multicloud and AI infrastructure continues to outstrip available capacity. That growth pattern is consistent with OCI acting as a preferred platform for AI workloads that the big three cannot accommodate at the speed or scale these customers require, rather than a replacement for them. The distinction matters for long-term pricing power: critical infrastructure commands margin; overflow capacity does not.
The remaining performance obligations figure of $638 billion is the bullish counter to that concern. That backlog represents contracted future revenue, and it grew $85 billion in a single quarter. As CNBC's coverage of the report makes clear, Bank of America analysts estimate more than half of that backlog comes from OpenAI alone, a concentration worth keeping in view. The conversion question, as investors are clearly signaling, is how long it takes for that backlog to become recognized revenue versus just building more data centers to chase the next tranche of contracts. Oracle's capital expenditures are running ahead of its depreciation and amortization schedules, which means the revenue recognition on these assets is still years out for much of what is being built today.
Oracle's management has consistently argued that the demand picture justifies the investment pace, and the backlog numbers provide some structural support for that argument. But the market's reaction to a record earnings report confirms that at $55.7 billion in annual capex and $40 billion more in financing to come, patience is no longer a given. What to watch next: whether IaaS revenue sustains growth above 80% into fiscal Q1 2027, and whether the early tranche of Stargate-linked capacity coming online can start narrowing that free cash flow gap even as the new $70 billion capex year gets underway.
Also read: Ramp's June AI Index reveals a 680-fold spending gap between AI leaders and everyone else • Mastercard's Agent Pay for Machines puts blockchain infrastructure at the center of the emerging AI transaction economy • Niteshift raises $7 million to be the cloud layer under every AI coding agent, not just the winning one