Oracle Debt and OpenAI Contracts: What Rising CDS Spreads Signal in 2026

Oracle Borrowed Its Way Into the AI Trade. The Credit Market Is Now Pricing the Risk.

August 13, 2026August 13, 2026By Dominick Bianco, Editor-in-Chief
0
Oracle Debt and OpenAI Contracts: What Rising CDS Spreads Signal in 2026
Share This:

Introduction

Oracle Corporation has spent the past year converting a database and enterprise software business into one of the largest debt-financed infrastructure builds in corporate history. The bet is anchored to a single customer: OpenAI.

The equity market has already repriced that bet. Oracle shares peaked at $345.72 in the third quarter of 2025 and touched a 52-week low of $121.50 on July 17, 2026  a decline of roughly 63% from the all-time high, according to Investing.com. But the more consequential signal is coming from the credit market, where the cost of insuring Oracle’s debt against default has moved to levels the company has never seen.

Oracle credit-default swaps traded near 200 basis points in early August 2026, compared with approximately 78 basis points for Nvidia and 93 basis points for Meta Platforms, Reuters reported. An index tracking investment-grade credit-default swaps traded near 53 basis points.

This article examines whether the arithmetic behind Oracle’s AI buildout holds up when the anchor tenant is itself losing money at scale.

Background

OpenAI signed a contract with Oracle to buy $300 billion worth of computing power over five years, one of the largest cloud deals ever struck, as first reported by The Wall Street Journal. The contract begins in 2027 and requires 4.5 gigawatts of power capacity.

The deal transformed Oracle’s reported order book. Remaining performance obligations grew $85 billion in the fourth quarter of fiscal 2026 alone, from $553 billion to $638 billion, according to Oracle’s Q4 and FY 2026 earnings release. Fiscal 2026 total revenue reached $67.4 billion, up 17%, with cloud revenue up 39% to $34.0 billion.

In simple terms: remaining performance obligations, or RPO, is contracted revenue Oracle has signed but has not yet delivered or recognized. It is a promise, not cash.

The concentration is the issue. S&P Global estimates that OpenAI accounts for roughly half of Oracle’s $638 billion in remaining performance obligations. By a Financial Times count cited in February 2026, 58% of Oracle’s contractual backlog was tied to the ChatGPT maker, making the company what the paper described as a publicly traded proxy for the AI startup’s fortunes.

Oracle funded the buildout primarily with borrowed money. In fiscal year 2026, Oracle raised $43 billion in debt financing and $5 billion in equity financing. For fiscal year 2027, the company expects to raise approximately $40 billion through a combination of debt and equity, including a previously announced $20 billion at-the-market equity issuance.

Paying for It: 30,000 Jobs

Oracle did not fund the buildout from the capital markets alone. It also cut payroll, in the largest restructuring in the company’s history.

As NexfinityNews reported in April, employees across the United States, India, Canada, Mexico and other countries received termination emails from “Oracle Leadership” at approximately 6 a.m. local time on March 31, 2026, with no prior warning from HR or their direct managers. System access was cut immediately.

Up to 30,000 employees were let go that day  roughly 18% of Oracle’s global workforce  to free up an estimated $8 billion to $10 billion in cash flow for AI data center construction. India was hit hardest, with approximately 12,000 terminations out of Oracle’s roughly 30,000-person Indian workforce.

That $8 billion to $10 billion figure is the number to hold against the capital program. Set beside fiscal 2027 capital expenditure guided at up to $95 billion, the largest workforce reduction in Oracle’s history covers roughly a tenth of a single year’s build.

The cuts were not distributed evenly. Divisions focused on legacy software maintenance, on-premises support and traditional SaaS operations absorbed the deepest reductions  Revenue and Health Sciences and SaaS/Virtual Operations Services each lost approximately 30% of staff  while teams working on Oracle Cloud Infrastructure, AI services and next-generation data center technology were largely spared, and in some cases were actively hiring.

Two details from that reporting bear on the credit picture. All unvested restricted stock units were forfeited immediately upon termination, meaning the cost of the reduction fell partly on employees rather than on Oracle’s cash. And federal data showed Oracle filed more than 3,100 H-1B visa petitions for fiscal year 2026 while the reductions were underway.

The external record corroborates the scale. CNBC reported on April 1 that Oracle had begun telling its 162,000-strong workforce that thousands would be affected, citing two people familiar with the matter; Oracle declined to comment. Luke Yang, an analyst at Morningstar, told USA TODAY that the reductions could account for about 30,000 positions and that the move should improve operating efficiency and boost revenue per headcount. As of March 31, Oracle had not filed a recent Worker Adjustment and Retraining Notification (WARN) Act notice, which requires employers with 100 or more employees to give 60 days’ advance notice of mass layoffs.

Oracle’s headcount stood at approximately 152,764 as of June 1, 2026.

The sequencing is what matters for the credit question. Oracle treated the AI buildout as a fixed obligation and operating expense as the variable it could adjust. That tells you which side gives first if OpenAI’s spending ramp slips.

What the Debt Numbers Actually Show

Reported totals vary depending on what is counted and as of which date.

Oracle’s long-term debt stood at $122.34 billion for the fiscal quarter ending May 31, 2026, according to Zacks Investment Research data. Counting current and non-current obligations together, total debt reached approximately $156 billion, with an additional $261 billion in data-center lease commitments. A separate analysis put total debt at roughly $167.4 billion by the close of fiscal 2026.

The scale relative to the broader bond market matters. Oracle has approximately $117 billion of debt in the Bloomberg US high-grade corporate bond index, making it the largest issuer outside the financial sector.

Cash flow moved the other direction. Free cash flow for fiscal 2026 turned negative at $23.7 billion, driven almost entirely by construction of GPU clusters and data centers. Capital expenditure for fiscal 2027 is guided at up to $95 billion, which after customer repayments would leave a net cash outflow of around $70 billion.

The backlog also converts slowly. Only 12% is expected to convert within 12 months and 34% within three years, meaning most of that revenue is years away.

How Credit Default Swaps Work  and What 200 Basis Points Means

A credit default swap is a derivative that protects against the risk that a bond issuer fails to pay what it owes. Prices are quoted as a credit spread, measured in basis points charged for protection. Higher perceived risk means a wider spread. A CDS quoted at 100 basis points costs $1 a year to insure every $100 of debt.

In simple terms: it is default insurance, and the premium is the market’s running estimate of how likely trouble is.

Oracle’s premium has repriced sharply. The five-year spread stood at 144 basis points at the start of 2026. On July 24, 2026, it reached a record of approximately 203 basis points  an annual cost of about $203,000 to insure $10 million of debt. Oracle’s 2056-maturing bonds widened to 263 basis points at a 6.7% coupon. One market tracker recorded a subsequent move to roughly 215 basis points.

Oracle’s 200-basis-point price is almost four times the 53-basis-point investment-grade index, indicating that investors currently require substantially greater compensation to insure its debt. However, a wider CDS spread represents market pricing of perceived risk rather than a prediction that Oracle will default. Reuters noted that increasing protection costs can pressure bond prices and raise future borrowing expenses.

The activity is not confined to Oracle. Technology-linked CDS trading reached almost $650 million per day during the second quarter, increasing 20% sequentially and nearly 600% from one year earlier, according to Depository Trust & Clearing Corporation data.

The ratings agencies moved in the same direction. S&P Global Ratings cut Oracle’s credit rating to BBB- on July 9, one notch above junk status, citing extreme concentration risk from OpenAI alongside a projected free cash flow deficit of $42 billion in fiscal 2027.

The Counterparty: OpenAI’s Own Losses

Oracle’s contracted revenue is only as good as the customer behind it.

Audited figures leaked in June 2026 and verified by the Financial Times and Fortune showed $13.07 billion of booked 2025 revenue against a $20.92 billion operating loss, with $19.18 billion of R&D spend and $17.2 billion paid to Microsoft for Azure compute. Net loss reached $38.5 billion after a one-time, non-cash charge of roughly $41.55 billion tied to OpenAI’s nonprofit-to-for-profit conversion.

The 2026 trajectory has not reversed. OpenAI is on pace for roughly $14 billion in losses this year on approximately $25 billion of run-rate revenue, after burning $3.7 billion in the first quarter alone against $5.7 billion of quarterly revenue. Internal projections reported by the Financial Times and The Information show cumulative losses of roughly $115 billion through 2029 before the company turns cash-flow positive around 2029 to 2030.

The funding side is stronger than the operating side. OpenAI closed a $122 billion funding round on March 31, 2026 at an $852 billion post-money valuation, with Amazon committing $50 billion and Nvidia and SoftBank each putting in $30 billion. It filed a confidential S-1 with the SEC on June 8, 2026.

The timing gap is the structural problem. Oracle’s contract revenue ramps from 2027. OpenAI’s own projections do not show cash-flow breakeven until the end of the decade. Oracle carries the construction debt in the interim.

The Burry Position

Michael Burry, who shorted the subprime mortgage market before the 2008 collapse, has traded Oracle in both directions this year  a detail often lost in summaries of his position.

Burry originally disclosed put options on Oracle in a January 2026 Substack post, shortly after shutting down Scion Asset Management in November 2025. His thesis centered on Oracle’s aggressive AI infrastructure spending, what he viewed as questionable financing, and heavy customer concentration around OpenAI.

He then took profits, telling readers the January 2027 low-100s strikes had produced a substantial gain and that he had decided not to roll them, while leaving open the possibility of re-entering if volatility came down, according to Benzinga.

He re-entered on August 6, 2026. Burry disclosed a short position in Oracle shares at $144.63, along with a new short in Nebius, arguing that multiple companies are overextended with long-term lease obligations that far outweigh current sales metrics, Stocktwits reported.

Oracle’s fourth-quarter fiscal 2026 results were objectively strong, with total revenue of $19.2 billion, up 21% year-over-year, and cloud infrastructure revenue up 93% to $5.8 billion. The short thesis is not about product demand. It is about whether capital expenditure required to service the backlog grows faster than revenue recognition.

Impact

Three effects follow from wider spreads.

Borrowing costs compound. Oracle still has roughly $40 billion to raise in fiscal 2027, and wider protection costs can pressure bond prices and raise the price of new issuance.

Equity dilution becomes the alternative. Oracle disclosed plans for an additional $20 billion equity issuance in calendar 2026, which would dilute existing shareholders at a price well below last year’s highs. The company has stated it does not expect to issue additional debt in calendar year 2026.

Concentration risk cuts both ways. If AI demand slows, funding conditions tighten, or OpenAI reduces spending, Oracle could be left with long-term data center lease obligations and specialized infrastructure that may be difficult to re-lease on attractive terms  a risk heightened because some customer contracts are shorter than the lease commitments Oracle has taken on to support them.

That mismatch has already surfaced once. In March 2026, CNBC reported that OpenAI was not planning to expand its partnership with Oracle at the Abilene, Texas Stargate site because it wanted clusters with newer generations of Nvidia chips. Oracle publicly called reports about the activity “false and incorrect,” but its statement addressed only existing projects, not expansion plans.

Analysis

The sustainability question does not resolve to a simple yes or no. It resolves to a timing question with two defensible readings.

The bull reading: the contracts are real, take-or-pay, and long-dated. Oracle trades at roughly 13.6 times forward earnings after falling 57% from its 52-week high, which looks inexpensive if the backlog converts to recurring revenue. OpenAI raised $122 billion in March and has filed to go public. Cloud infrastructure revenue is compounding at rates most large-cap software companies cannot match.

The bear reading: contracted revenue is not collateral. Customer concentration always deserves a discount because investors lose diversification. Oracle is carrying negative free cash flow, a BBB- rating, $261 billion in off-balance-sheet lease commitments, and a five-year CDS at four times the investment-grade index  while the customer responsible for half the backlog does not expect to be cash-flow positive until roughly 2030.

There is also a scale argument worth noting. The single-name CDS market represents approximately $9 trillion, compared with more than $150 trillion of global bonds outstanding  a relatively thin market in which a modest volume of protection buying can move a headline spread. A record CDS print is a signal, not a verdict. NexfinityNews has previously examined how concentrated flows in thin markets can drive outsized price moves in its analysis of the Citadel–Situational Awareness block trade and the July 2026 AI selloff.

What is not in dispute is that Oracle chose a financing structure no other hyperscaler chose. Microsoft, Alphabet, Amazon, and Meta are funding AI capacity substantially from operating cash flow. Oracle is funding it from the bond market. That leverage sits inside a broader pattern this publication has tracked, from record retail margin debt of $1.5 trillion to the scramble for data center power capacity.

Conclusion

Oracle’s AI buildout becomes a solvency question only under a specific scenario: OpenAI’s spending ramp slipping materially while Oracle’s construction obligations and debt service continue on schedule. Nothing in the current disclosures shows that happening today.

What the disclosures do show is a company with negative free cash flow of roughly $24 billion, total debt above $150 billion, a rating one notch above speculative grade, 30,000 jobs cut to protect cash flow, and half of its order book resting on a counterparty that lost approximately $21 billion at the operating line last year.

The credit market has moved from 144 basis points to roughly 200 in eight months. Oracle’s next earnings report and the pace of its fiscal 2027 capital raise will determine whether that repricing was early or correct.

Key Takeaways

  • Oracle’s total debt has climbed to roughly $156 billion to $167 billion depending on measurement date, with an additional $261 billion in data-center lease commitments not carried on the balance sheet.
  • Fiscal 2026 free cash flow was negative $23.7 billion; fiscal 2027 capital expenditure is guided at up to $95 billion.
  • Oracle cut up to 30,000 jobs  roughly 18% of its workforce  on March 31, 2026, freeing an estimated $8 billion to $10 billion in cash flow against fiscal 2027 capital expenditure guided at up to $95 billion.
  • OpenAI accounts for roughly half of Oracle’s $638 billion contracted backlog, of which only 12% converts within 12 months.
  • Oracle’s five-year CDS rose from 144 basis points in January 2026 to a record of about 203 basis points in July, against 78 for Nvidia, 93 for Meta, and 53 for the investment-grade index.
  • S&P downgraded Oracle to BBB- on July 9, 2026, one notch above junk, citing OpenAI concentration and a projected $42 billion fiscal 2027 free cash flow deficit.
  • OpenAI booked $13.07 billion in 2025 revenue against a $20.92 billion operating loss and projects roughly $115 billion in cumulative losses through 2029.
  • Michael Burry closed his original Oracle puts after taking profits, then re-entered with a direct equity short at $144.63 on August 6, 2026.
  • A wider CDS spread reflects the market’s price of perceived risk, not a forecast of default.

Related Coverage from NexfinityNews

Sources

Share This:
, , , , ,