Oracle's $225 Billion Revenue Forecast and Financing Cap Spark 11% Rally

Oracle shares surged roughly 11% on Wednesday after the company used its annual Investor Day in Las Vegas to answer the three questions that have been dragging the stock down for more than a year: When does the spending peak? When does free cash flow turn positive? And how much more money will Oracle need to raise? The answers — capex peaks in fiscal 2027-28 and then steps down, free cash flow returns in the second half of fiscal 2028, and new capital raises are capped at $20 billion through fiscal 2030 — reversed a decline that had taken Oracle from a September 2025 peak near $328 to a 52-week low of $121.50 in mid-July.

The stock move was not about a surprising revenue quarter. Oracle did not release quarterly results on Wednesday; its second-quarter fiscal 2027 earnings are not expected until December. The rally was a relief trade, driven specifically by the $20 billion cap on additional capital raises and a hard timeline for free cash flow recovery. After a year in which Oracle accumulated $43 billion in debt and $5 billion in equity in a single fiscal year, issued bonds that trade at junk-adjacent yields of roughly 6.5%, and absorbed an S&P downgrade to BBB- in July, investors needed a number that said the borrowing would eventually stop. Wednesday delivered one, as Oracle disclosed in a press release filed with the SEC.

The Debt Decline That Set Up the Relief

Oracle entered its Investor Day carrying approximately $117 billion in long-term debt — second only to Amazon by outstanding bonds among non-financial US corporate issuers — and a credit rating one notch above speculative grade. Moody's has a negative outlook on Oracle and warned that leverage could temporarily approach five times adjusted EBITDA. Morgan Stanley credit analysts flagged the possibility that Oracle could end the calendar year with two low-BBB ratings, raising what bond investors call fallen-angel risk: a move to high-yield territory forces institutional mandates that prohibit junk bonds to sell, compressing prices regardless of the underlying business.

The proximate cause of the debt load is straightforward. Oracle's fiscal 2026 capital expenditure reached $55.7 billion — roughly 83% of total revenue of $67.4 billion — as the company raced to build out GPU cluster capacity for AI customers including OpenAI, Meta, NVIDIA, AMD, and xAI. Oracle's fiscal 2026 earnings release showed free cash flow falling to negative $23.7 billion for the fiscal year. With analysts at FactSet estimating capex could jump another 65% to roughly $91.9 billion in fiscal 2027, the open question was how much larger the deficit would grow before it started narrowing — and how many more bond and equity offerings Oracle would need to fund the gap.

The S&P downgrade in July 2026 crystallized the concern. By mid-July the stock had fallen roughly 63% from its peak. Bernstein analysts, in a note ahead of Investor Day, observed that Oracle had lost about 60% of its value since its last major investor day, even as the contracted revenue backlog kept climbing.

Three Commitments That Moved the Market

At the Investor Day, co-CEOs Clay Magouyrk and Mike Sicilia and CFO Hilary Maxson made three specific commitments that the prior financial narrative had not contained.

First, Oracle defined a free cash flow timeline. The company expects to reach positive free cash flow during the second half of fiscal 2028 — the period ending November 2028 — and projects free cash flow of at least $30 billion in fiscal 2029. Net debt relative to adjusted EBITDA is targeted below 3.5 times in fiscal 2029, down from roughly 4.3 times today.

Second, Oracle specified that capex will peak in fiscal 2027 and fiscal 2028 and then step down in fiscal 2029. That capex decline directly enables the FCF recovery: once infrastructure spending plateaus and declines, the operating cash flow Oracle generates from its contracted backlog flows through to free cash flow rather than being consumed by construction. As of the first quarter of fiscal 2027, Oracle's remaining performance obligations stood at $664 billion, representing contracted multi-year commitments from customers.

Third — and the specific detail that most directly addressed the stock's core concern — Oracle committed to limiting additional capital-market fundraising through fiscal 2030 to $20 billion total, split roughly evenly between equity and debt. This is incremental to the $20 billion at-the-market equity offering Oracle completed in the first quarter of fiscal 2027, and it excludes refinancing of existing debt. The commitment tells bond and equity investors that Oracle will not return to capital markets for another $40–50 billion round like fiscal 2026's. The mechanism that makes this possible: vendor financing, customer-provided hardware, and customer prepayments are expected to absorb a significant share of the remaining buildout cost, reducing Oracle's upfront cash requirement. Oracle has disclosed that many of its large recent AI contracts involve customers either prepaying for GPUs so that Oracle can purchase the chips on their behalf, or customers buying the GPUs outright and supplying them to Oracle — a structural cost shift that was underappreciated in prior discussions of Oracle's financing needs.

Why AI Customers Built Their Workloads on OCI

Understanding why OpenAI, Meta, and a roster of frontier AI labs chose Oracle rather than AWS or Azure requires understanding what Oracle Cloud Infrastructure (OCI) actually built. Conventional multi-tenant cloud infrastructure isolates customer workloads through software-defined virtual networks layered on shared physical switching infrastructure. OCI's second-generation architecture takes a different approach: it isolates customer networks at the physical hardware level, with no shared virtual switches between tenants. Each customer gets a dedicated network fabric.

For AI training, which requires moving gradient updates between hundreds of thousands of GPU nodes with deterministic, low-jitter latency, physical network isolation matters. OCI combines this isolation with RDMA (Remote Direct Memory Access) networking over a converged Ethernet fabric — a technology that allows GPU memory on different servers to communicate directly without CPU involvement, at throughput approaching InfiniBand. This is the same architecture that HPC supercomputers use for tightly coupled scientific workloads. Oracle has packaged it into what it calls OCI Superclusters, which the company says can scale to more than 100,000 NVIDIA GB200 Grace Blackwell Superchips in a single contiguous cluster.

A second differentiation is Oracle's multicloud database strategy. Oracle's Exadata infrastructure — the hardware substrate for Oracle Database — can now be deployed natively inside AWS, Azure, and Google Cloud, through partnerships Oracle calls Oracle Database@AWS, @Azure, and @Google Cloud. This means enterprises whose transaction systems, ERP workflows, and analytics data reside in Oracle Database can run AI inference and agentic workloads directly against that data without moving it across clouds. Multicloud database revenue grew 531% year over year in Oracle's fiscal third quarter of 2026 — the fastest-growing component of the cloud portfolio.

OCI GPU utilization reached 97.5% in fiscal Q4 2026 — a figure that suggests demand has outpaced capacity rather than capacity sitting idle. Oracle signed $67 billion in new cloud contracts in Q4 alone, and delivered more than 1.2 gigawatts of compute capacity to customers across the fiscal year.

The $225 Billion Revenue Math

Oracle's fiscal 2030 target of at least $225 billion in total revenue implies a compound annual growth rate above 35% from the fiscal 2026 base of $67.4 billion. Oracle projects non-GAAP earnings per share of at least $21 and a non-GAAP operating margin of at least 40% in fiscal 2030.

Three revenue streams carry the forecast. Cloud infrastructure — OCI — has previously been projected to reach $166 billion in fiscal 2030, roughly three-quarters of the total target, up from $18.1 billion in fiscal 2026. AI-powered database and data platform services — Oracle's Autonomous Database and the broader AI Data Platform — are targeted at $20 billion in fiscal 2030, up from roughly $3 billion in fiscal 2026. Cloud applications (ERP, HCM, SCM via Oracle Fusion) and other revenue fill the remainder.

The revenue ramp has been accelerating. Oracle's cloud infrastructure business grew 77% in fiscal 2026, then 84% in Q3 FY26, then 93% in Q4 FY26, then 121% in Q1 FY27. Revenue growth — total, not just cloud — hit 30% in the quarter ending August 31, 2026. Oracle guided 30–34% total revenue growth for its second fiscal quarter of 2027. The $664 billion RPO backlog as of Q1 FY27, representing contracted multi-year commitments from customers, provides meaningful forward revenue visibility.

Risks the Investor Day Did Not Erase

The commitments Oracle made on Wednesday are forward-looking and company-stated. Several risks remain unresolved.

The most technically specific is the depreciation schedule controversy. Oracle depreciates servers and networking equipment over an estimated useful life of up to six years — consistent with how AWS, Microsoft, and Google treat similar assets. Short-seller Jim Chanos has argued that the practical useful life of AI GPUs is closer to two to three years, given NVIDIA's annual chip release schedule and the corresponding obsolescence of prior-generation hardware: rental rates for Hopper-generation chips have already fallen roughly 28% year over year as Blackwell arrived. Michael Burry has estimated Oracle will overstate earnings by 26.9% by fiscal 2028 if the depreciation schedule holds. Oracle's 10-Q for the quarter ending August 31, 2026 shows $77.8 billion in gross property, plant and equipment for computers and networking equipment, with $25.8 billion in total accumulated depreciation across all fixed asset categories — a significant asset base whose real economic value depends entirely on the GPU obsolescence cycle playing out at the six-year pace Oracle assumes.

Customer concentration is a second unresolved risk. OpenAI is estimated by S&P and independent analysts to account for roughly half of Oracle's $664 billion RPO backlog — a reported multi-year contract worth in the vicinity of $300 billion. OpenAI itself is not yet profitable and continues to require its own external financing. S&P specifically flagged this concentration as a key credit risk in its July downgrade decision. If OpenAI were unable to fund its commitments on schedule, the RPO would need to be revised and Oracle's self-funding premise would be tested.

A third risk is the thin near-term margin on AI hardware rentals. A report from The Information, cited by SemiWiki, indicated that Oracle lost approximately $100 million from renting Blackwell chips in its most recent quarter, generating only about 14 cents of gross profit per dollar of revenue from that activity. The long-duration contracts may eventually be profitable, but the current economics of renting next-generation GPUs suggest Oracle's cloud infrastructure margins will remain under pressure through the peak capex years.

Oracle's BBB- credit rating from S&P remains in place. The $20 billion raise cap and FCF timeline announced Wednesday are company commitments, not rating actions. S&P, Moody's, and market participants will evaluate whether Oracle executes on schedule — and the next meaningful data point arrives with second-quarter fiscal 2027 results expected in December 2026. That quarter's numbers will show whether the 30–34% total revenue growth Oracle guided is materializing and whether Q1's 121% OCI growth rate is holding, decelerating, or accelerating further. The FY29 free cash flow thesis will not be testable for two years, but the near-term execution trend is what rating agencies will use to decide whether the BBB- floor holds.