Of all the major technology companies riding the generative AI wave, Oracle’s financial transformation is the most extreme, and its accounting blind spots the most treacherous for retail and institutional investors alike. For decades, Oracle was a high-margin, incredibly sticky software business, generating robust, predictable free cash flow through enterprise database licenses and recurring support fees. Today, management is effectively executing a leveraged buyout of the AI physical infrastructure market. The announcement in early 2026 that Oracle intends to raise up to $50 billion in debt and equity to fund data center buildouts for clients like OpenAI, Meta, and xAI fundamentally alters the company’s risk profile. Oracle is rapidly transitioning from a software provider into an outsourced, highly levered infrastructure landlord for the largest tech companies in the world.
The disconnect between Oracle’s cash flow reality and its income statement has reached critical, unsustainable levels. In fiscal 2025, Oracle’s depreciation and amortization (D&A) remained completely flat year-over-year at $6.17 billion. Yet, in that exact same period, capital expenditures tripled to $21.22 billion, driving the firm’s free cash flow negative. Because the physical servers and massive gigawatt data centers take significant time to construct and be placed into service, the depreciation hit is severely delayed. Investors looking solely at Oracle’s 2025 operating margins are viewing a financial illusion; they are seeing the revenue benefits of early cloud adoption without bearing the expense weight of the infrastructure that supports it.
When the $21 billion from 2025 and the incoming $50 billion from 2026 hit the balance sheet and enter service, the subsequent depreciation expense will be staggering. Visible Alpha estimates that Oracle’s depreciation will skyrocket to $33 billion by 2029, representing an estimated 18% of total revenue. Because Oracle embeds this depreciation within its “Cloud and software” operating expenses, the headline gross margins of the cloud division are destined for severe compression. S&P Global Ratings has already noted that Oracle’s leverage metrics will remain stretched, forecasting adjusted leverage to reach the high-4x range in fiscal 2027 and 2028.
The bullish counter-argument, heavily promoted by management, is found in Oracle’s Remaining Performance Obligations (RPO). Oracle reported a staggering RPO backlog of $523 billion in late 2025, fueled by massive multi-year commitments from hyperscalers who desperately need power and space. Management argues that the $50 billion capital raise is entirely de-risked because the data centers are essentially pre-leased to blue-chip tenants on ironclad long-term contracts. However, the operational execution risk remains immense. Oracle is taking on massive debt to buy rapidly depreciating assets (GPUs and servers) that will be technologically obsolete in three to five years. If Oracle is depreciating these assets over their maximum useful lives to protect GAAP earnings, it runs the precise risk that Amazon is currently trying to avoid: a balance sheet bloated with carrying values that far exceed their true economic worth. If any of the major AI startups default, or if the next generation of AI chips renders the current infrastructure too power-inefficient to run profitably, Oracle’s leveraged bet will result in catastrophic asset impairments. Oracle is no longer a traditional software stock; it is a highly levered, macro bet on the physical plumbing of the AI economy.