Oracle's default insurance costs have moved past their 2008 crisis peak, and the message from credit markets is blunt: the AI buildout is no longer just an equity story.
Oracle's five-year credit default swaps closed at 198.23 basis points on Friday, July 17, according to Bloomberg's report citing ICE Data Services. That narrowly passed the 198.18 basis point record set on March 27 and topped the company's December 2008 peak. That's the story. Investors now want more to insure Oracle's debt against default than they did during the worst banking crisis in modern memory.
The move has been fast. The cost of protection has roughly quadrupled since mid-2025, Bloomberg reported. That is not a drift. It is a market repricing a company that used to be treated as a steady software cash machine and is now borrowing like an infrastructure developer with an AI customer list.
Oracle's AI Bill Is Getting Harder to Ignore
Oracle didn't arrive here by accident. In September 2025, the company sold $18 billion of investment-grade bonds, according to its SEC filing, with notes running from 2030 out to 2065. Oracle said the proceeds would be used for general corporate purposes, including capital expenditures. Everyone understood what that meant. The money was going into cloud capacity, power and data centers for AI customers.
The company has since put hard numbers around the strain. Oracle's fiscal 2026 results show capital expenditures of $55.7 billion, up from $21.2 billion a year earlier. Operating cash flow was $32 billion, which left free cash flow at roughly negative $23.7 billion after capex. CNBC reported that Oracle had about $130 billion in debt and had just suffered its worst week for the stock since August 2001, with shares down 19% in the week ended June 26.
The debt is real. So is the demand.
Oracle's own June results said remaining performance obligations reached $638 billion, a huge backlog tied heavily to cloud infrastructure. But S&P Global Ratings cut Oracle's long-term issuer credit rating to BBB-, one notch above junk, after warning about the capital requirements and customer concentration behind that backlog. Investor's Business Daily reported that OpenAI accounts for nearly half of those remaining performance obligations. You don't have to be bearish on AI to see the problem. One customer can be very valuable and still make lenders nervous.
Oracle has tried to show discipline. In February, it announced a 2026 funding plan of $45 billion to $50 billion, split between debt and equity, and said it wanted to preserve an investment-grade balance sheet. In June, Oracle said it had raised $43 billion in debt financing and $5 billion in equity financing during fiscal 2026, and expected to raise about $40 billion more through debt and equity in fiscal 2027. That is a lot of financing for a company whose old pitch was database durability.
The OpenAI Bet Cuts Both Ways
OpenAI is the name under this whole trade. In July 2025, OpenAI said it and Oracle had agreed to develop 4.5 gigawatts of additional Stargate data center capacity in the U.S. Two months later, OpenAI said Oracle-linked Stargate sites included Shackelford County, Texas, Doña Ana County, New Mexico, and a Midwest site later identified as Wisconsin, alongside the Abilene, Texas flagship campus. Oracle's own January update also named Michigan as part of its 2026 AI infrastructure work with OpenAI.
That footprint is enormous. It also has to work on time.
Credit investors are not only asking whether AI demand exists. It plainly does. Morgan Stanley said in a July 16 credit-market discussion that AI-related debt issuance was already close to $250 billion this year and could reach about $500 billion for 2026. The harder question is whether construction schedules, power supply, chips and customer payments line up cleanly enough to support all the debt being created around them.
Here's the thing: Oracle is now a clean way for traders to express that worry. Bloomberg has described Oracle as the largest non-bank issuer in the Bloomberg U.S. High-Grade Corporate Bond Index. When traders want to hedge AI infrastructure credit risk, Oracle is sitting right there.
The stock market has caught up late. Oracle shares closed at $126.41 on July 17, according to Bloomberg Línea market data, down about 35% for 2026 and more than 63% below the 52-week high reached in September 2025. Some of the newest pressure has come from China, where Reuters reported that Moonshot AI unveiled Kimi K3, a 2.8 trillion-parameter open-weight model, on July 17. If cheaper open models keep improving, investors will ask tougher questions about how much proprietary AI capacity companies need to rent at premium prices.
None of this means Oracle will default. A CDS spread is a price, not a verdict. But it is a vote, cast daily by the bond market, and right now that vote says Oracle's AI wager is carrying more visible credit risk than at any point in the company's modern trading history. The next numbers to watch are plain ones: capex, free cash flow, debt issuance, and whether OpenAI's demand keeps arriving quickly enough to make the buildout look less like a leap.
Also read: Lorenzo Protocol's BANK Token Rockets 517% After a Mysterious Wallet Move • Moonshot's Bet on a Bigger Kimi K3 Is Paying Off in China's AI Race • Google Shares Rise After Report of a Chip Designed Around Gemini
