The OpenAI order is the future and the debt is huge! Oracle (ORCL.US) default risk reached a record high, and CDS surpassed the 2008 peak

The Zhitong Finance App learned that according to the latest statistics, the risk of default in database software and cloud computing supergiant Oracle (ORCL.US) has soared to the highest level in history, surpassing the historical peak during the 2008 global financial crisis. The company's five-year credit default swap (CDS) spread — a key measure of the cost of insuring its risk of debt default — has risen to a record high, indicating that investors are increasingly concerned about the real risk of Oracle's credit default.

A credit default swap (CDS) is a financial contract that protects against the risk of a borrower being unable to pay a debt. Higher CDS spread data means that investors require higher risk premiums before they are willing to provide default insurance for related debts. This reflects the rising credit risk perceived by the market.

Oracle CDS surpasses the peak set in 2008

In less than a year, Oracle went from being a global leader in AI cloud computing to a “garbage cliff” company with a record high risk of default, surpassing the peak of the 2008 financial crisis in one fell swoop. Standard & Poor's, one of the top three international credit rating agencies, previously downgraded Oracle to BBB- on July 9, 2026, which is only one step away from the most pessimistic rating of “junk.”

At a time when Oracle's CDS is rising sharply, the company is increasing AI infrastructure spending and OpenAI, which is not yet profitable, to become one of Oracle's largest cloud computing order customers. This includes Oracle investing billions of dollars in cloud computing capacity and AI data centers to meet the increasing AI computing power needs of OpenAI and other customers.

Oracle is increasingly reliant on debt financing and a single AI application client to support the expansion of its AI cloud computing infrastructure business; as AI capital expenditure continues to rise, investor concerns about its balance sheet have also intensified. OpenAI-related contracts bring huge forward computing power requirements to Oracle, but Oracle must bear the cost of data center land contracts, chips, electricity, and financing in advance.

Oracle turned the long-term AI computing power needs of major customers such as OpenAI into data center construction, capital expenses, and AI financing burdens ahead of time, and whether orders can be fulfilled still depends on customer financing capacity, AI revenue growth, and contract execution. Oracle's FY2026 RPO has reached $638 billion, but high capital expenses, approximately $129.5 billion in debt, and long-term lease commitments have repriced its credit risk.

These concerns have continued to put selling pressure on its share price. Oracle's stock price has fallen 27% since this year, with a decline of nearly 44% over the past year. The core reason is that investors are re-evaluating the company's aggressive AI investment strategy and its rising level of credit leverage.

From AI cloud computing leader to eye of credit default storm

Oracle did not fall into a credit crisis because the cloud computing business related to AI computing power completely failed; it was precisely because AI orders grew much faster than its own capital carrying capacity.

In fiscal 2026, Oracle's cloud computing infrastructure business, Oracle Cloud Infrastructure (OCI business), saw revenue growth of 77%, cloud business revenue growth of 39%, and remaining performance obligations (RPO) surged 363% to US$638 billion, proving that it has become an important bearer of global AI training and inference infrastructure; however, in order to meet these future revenues, the company's annual capital expenditure jumped from US$21.2 billion to US$55.7 billion.

Even with a record operating cash flow of $32 billion, free cash flow turned negative at $23.7 billion. In other words, Oracle has achieved an “explosion of demand orders for AI cloud computing infrastructure,” yet it has yet to complete the transition from huge contracts to sustainable cash returns. As a result, the commercial victory of AI cloud computing evolved into a stress test for the balance sheet.

What really frightens the credit market is the combination of leverage, term, and counterparty risk: by the end of fiscal year 2026, Oracle's debt was about US$129.5 billion, equivalent to about 4.3 times EBITDA, and it also signed long-term data center lease commitments of about US$260 billion; these leases usually last 15 to 19 years, while the longest computing power contracts for some customers are only about five years, creating a typical balance and liability mismatch. More importantly, OpenAI accounts for about half of its $638 billion RPO, making Oracle actually use its own investment-grade credit to build long-term fixed assets ahead of schedule for AI laboratories that are still consuming large amounts of cash.

As a result, S&P downgraded Oracle to BBB- on July 9, 2026, which is only one step away from garbage; by the beginning of August, its five-year CDS had risen to about 215 basis points, and bond yields had reached 7% to 8%, which is usually close to high-yield bonds. This does not mean that Oracle is about to default, but rather that the credit market is putting an unprecedented price on tail risks such as “data center delays, insufficient OpenAI compliance, rapid chip depreciation, or rising refinancing costs.”

The biggest difference between Oracle and Microsoft, Alphabet, and Amazon is not that AI demand is weaker, but rather that when it enters the construction cycle, it already has significantly higher leverage, weaker free cash flow, and greater customer concentration. Free cash flow conversion rates, customer advance payment ratios, tenancy and customer contract period limits, adjusted debt/EBITDA, CDS, and credit ratings are becoming core indicators for investors to examine the real winners of the “AI Super Bull Market.”

As a result, Oracle became the first credit litmus test for the entire AI debt cycle. Castle Securities estimates that by 2028, AI chip purchases alone may generate more than 500 billion US dollars in additional debt; this “computing power securitization” can lock in AI computing power cluster purchases, optical communications/optical interconnections, data center CPUs, high-performance Ethernet switching infrastructure, data center HBM/DRAM/NAND storage components, and data center power chain orders in advance, but it also transforms commercial risks that originally depended on terminal AI revenue into nested credit risks between bonds, SPVs, long-term leases, and supplier guarantees.