Oracle's $638 Billion AI Backlog Can't Mask a Balance Sheet Under Siege

Oracle's AI pivot yields record $638B backlog but triggers credit downgrade, $23.7B negative cash flow, and 59% stock decline; half the backlog depends on OpenAI.

Oracle once epitomized the steady-eddie of enterprise software: fat margins, loyal corporate customers, and minimal drama. That identity has been shattered by a breakneck transformation into a hyperscaler of AI infrastructure, leaving the company with a record order book but a balance sheet that is increasingly groaning under the strain. Shares traded around €116 in mid-July, barely above the 52-week low of €111.58 hit on July 14, and more than 59% below the September 2025 peak of €280.70. The stock has shed 30.6% year to date and 28.6% in the past month alone.

At the heart of the Oracle story sits a staggering number: $638 billion in remaining performance obligations (RPO). On paper, that backlog underscores seemingly insatiable demand for AI compute capacity. But credit rating agencies see it differently. On July 9, S&P Global Ratings downgraded Oracle's long-term issuer rating to BBB-, just one notch above junk territory, explicitly citing the company's aggressive pivot toward AI infrastructure. The downgrade carries a stable outlook, but the company now has no buffer before slipping into speculative grade.

The risk concentration is the key concern. Roughly half of that $638 billion RPO is tied to a single customer: OpenAI. The ChatGPT developer burns through enormous amounts of capital with no end in sight, and any wobble in its own financing would leave Oracle with few alternatives. Rivals like AWS, Google, and Microsoft run massive in-house AI workloads that provide a natural hedge against customer churn; Oracle lacks that cushion almost entirely.

That concentration risk is compounded by a cash flow picture that has turned deeply negative. In fiscal 2026, Oracle reported negative free cash flow of $23.7 billion, driven by capital expenditures of $55.66 billion on new data centers and chips. The spending spree shows no sign of abating: analysts project capex of $90 billion to $95 billion in fiscal 2027, which could push free cash flow to a deficit of around $42 billion. Total debt already stands at roughly $167 billion, a load that makes the company look more like a capital-intensive utility than a traditional software firm.

To fund the build-out, Oracle plans to tap capital markets. A $20 billion equity raise is slated for fiscal 2026, and media reports suggest the company may need an additional $40 billion in debt and equity in fiscal 2027 — including a further $20 billion equity component that has sparked dilution fears among shareholders. S&P itself noted that the planned equity measures demonstrate Oracle's commitment to preserving its investment-grade rating, but the sheer scale of the financing has rattled investors.

Not everyone is bearish. Cloud infrastructure revenue surged 93% year over year to $5.8 billion in the latest quarter, driven by AI demand. The RPO grew 363% over the same period, giving Oracle what its proponents call "unprecedented revenue visibility." The Wall Street analyst community remains largely constructive: the average price target stands at €220.32, implying roughly 90% upside from current levels, and 37 of 43 analysts maintain a buy rating. The relative strength index (RSI) at 30.4 signals that the stock is technically oversold.

The bull case rests on the idea that Oracle is a deep-value hyperscaler — cheaper than Microsoft or Amazon relative to its growth trajectory — and that once the peak of the investment cycle passes in fiscal 2027, the cash generation will come roaring back. The company is also pushing to monetize its software stack with new "Fusion Agentic Applications," a suite of AI-native development tools unveiled in partnership with Accenture and Deloitte. The goal is to shift focus from raw compute capacity to an "agentic" era of autonomous business processes. But the market has so far shrugged off the announcement; the balance sheet concerns overshadow the technology vision.

On the other side, the risks are real and mounting. S&P has laid out specific triggers for a further downgrade: if Oracle's debt-to-EBITDA ratio stays persistently above 4.5x, if the company fails to generate positive free cash flow by fiscal 2029, or if the agency takes a dimmer view of Oracle's AI strategy or the AI sector as a whole. Even a single notch below BBB- would send the company into junk territory, driving up borrowing costs and potentially tripping covenants in existing loan agreements.

The OpenAI concentration is the most acute vulnerability. The start-up has already postponed its planned initial public offering to 2027. SoftBank recently reduced a loan backed by OpenAI equity from $10 billion to $6 billion due to valuation uncertainties. If OpenAI's own financial position deteriorates further, Oracle's massive RPO could quickly lose its sheen.

For now, the stock is pricing in a great deal of pessimism. But the path to recovery hinges on a single, high-wire act: converting that $638 billion backlog into actual cash before the funding costs spiral and the rating falls. The investment cycle is expected to peak in fiscal 2027. It is only after that cliff that investors will learn whether Oracle emerges as a leading AI infrastructure provider — or as a cautionary tale of a software giant that leveraged itself to the hilt chasing a technology that may not deliver on time.