Oracle Corporation (ORCL) is executing one of the most aggressive capital reallocation programs in enterprise technology history, targeting $50 billion in capital expenditures for fiscal year 2026 — up from $6.9 billion in FY2024 — to build AI cloud infrastructure for customers including OpenAI, AMD, Meta, NVIDIA, and xAI. The company has raised tens of billions through debt and equity markets to fund the buildout, backed by $130 billion in remaining performance obligations as of Q3 FY2026. But on April 28, 2026, the Wall Street Journal reported that OpenAI missed internal user and revenue targets, causing Oracle shares to tumble and raising a central investor question: how much of Oracle’s contracted backlog depends on a single customer that may be growing slower than expected?
Oracle’s fiscal year ends May 31. Q3 FY2026 covers December 2025 – February 2026, reported March 2026. FY2024 ended May 31, 2024.
The $50 Billion Bet: Oracle’s AI Infrastructure Commitment in Context
Oracle’s capital expenditure trajectory has no precedent in the company’s history. FY2024 capex was $6.9 billion; the FY2026 target is approximately $50 billion — a 7x increase in two years. The spending is directed toward Oracle Cloud Infrastructure (OCI) data centers capable of hosting tens of thousands of GPUs, which are the hardware essential for training and running large AI models.
Table 1. Oracle Capital Expenditure Trajectory
Fiscal Year
Capex
YoY Change
FY2024 (ended May 31, 2024)
$6.9 billion
Historical baseline
FY2025 (ended May 31, 2025)
$21.2 billion
3x FY2024
FY2026 (target, ends May 31, 2026)
$50 billion
~2x+ FY2025 pace
The investment is not speculative in management’s framing. Oracle’s remaining performance obligations — contracted revenue not yet delivered — stood at $553 billion as of Q3 FY2026, with AI workloads representing a substantial portion. For context, Oracle’s total revenue for fiscal year 2025 was $57.4 billion. The $553 billion backlog implies years of contracted demand, though the key variable is whether the customers behind that backlog — and particularly the largest one — actually consume what they have committed.
Capital Markets Backing: Debt, Equity, and the Backlog Behind the Spend
Oracle cannot fund $50 billion in annual capex from operating cash flow alone. The company’s non-GAAP adjusted free cash flow was negative $24.7 billion on a trailing four-quarter basis as of Q3 FY2026. To bridge the gap, Oracle turned to capital markets with a multi-pronged financing strategy.
In March 2026, Oracle issued approximately $15 billion in senior unsecured notes, one of its largest bond offerings on record, following a late 2025 debt raise. Oracle also began an at-the-money equity offering — selling shares gradually at market prices — targeting additional capital for the buildout.
The OpenAI Risk: Concentration in a Single Customer’s Success
The Oracle–OpenAI relationship is large enough to represent a single-customer concentration risk unusual for a company of Oracle’s size. Media reports have cited a multi-year contract valued at approximately $300 billion over five years; even at a fraction of that value, the partnership would represent a substantial portion of Oracle’s $553 billion RPO. OpenAI has simultaneously expanded an agreement with Amazon Web Services, reportedly valued at $138 billion, underscoring a multi-cloud strategy that could dilute Oracle’s share of OpenAI’s infrastructure spend.
Technological efficiency gains add a structural dimension to the demand uncertainty. OpenAI’s release of GPT-5.5 in April 2026 reportedly reduced inference costs, enabling the company to serve more users with less raw compute. As model efficiency improves, Oracle’s data center utilization projections may be overstated even if OpenAI’s revenue recovers.
Oracle does have structural mitigants. Large cloud infrastructure contracts typically include minimum usage commitments and termination clauses. Oracle is also pursuing customer diversification beyond OpenAI, with workloads from AMD, NVIDIA, Meta, and government agencies. And the company’s $553 billion RPO — if it converts to revenue as contracted — would represent 9.6 years of current annual revenue, providing a longer runway than the near-term sentiment suggests.
Workforce Restructuring: Funding the Future by Cutting the Past
In early-to-mid 2026, Oracle initiated one of the largest workforce restructurings in its history, with an estimated 20,000–30,000 positions eliminated, primarily from legacy software support, on-premises product lines, and administrative functions. The company is simultaneously hiring engineers and cloud specialists to support OCI expansion, including a $16 billion data center announced in Michigan in April 2026.
The restructuring creates a near-term gap: severance costs and retraining expenses will weigh on GAAP earnings while new hires ramp up, likely widening the spread between GAAP and non-GAAP reported results through late FY2026. In the European Union, collective redundancy rules require advance notice and employee consultation, introducing geographic timing variability in the cost recognition. The combination of these charges with the front-loaded capex depreciation schedule means Oracle’s GAAP income statement will face meaningful pressure before the AI revenue buildout fully materializes.
Key Signals for Investors
Oracle’s $553 billion RPO as of Q3 FY2026 represents the clearest indicator of AI demand absorption; the pace at which it converts to recognized revenue — particularly the portion attributable to OpenAI — is the single most important metric to track in Q4 FY2026 and FY2027.
The Wall Street Journal’s April 28 report on OpenAI missing internal targets introduced counterparty risk into Oracle’s valuation framework; investors should monitor OpenAI’s quarterly revenue disclosures and any amendments to contracted infrastructure commitments.
Oracle’s stock is down approximately 12% year-to-date as of May 1, 2026, despite the massive backlog — reflecting investor skepticism that the AI demand trajectory will be sufficient to justify the financial leverage being accumulated.
The domestic first-class EBIT cadence for legacy software and support margins will determine how much internal cash generation Oracle can redirect toward infrastructure without further credit deterioration.
