Oracle at BBB-: The Credit Market's Verdict on the AI Infrastructure Bet

On July 9, 2026, S&P Global Ratings downgraded Oracle Corporation's long-term issuer credit rating from BBB to BBB-, the lowest rung of investment grade. The outlook was set to stable. One notch below BBB- lies BB+ — junk.

This is not a routine one-notch adjustment. It is the credit market's most explicit verdict yet on the AI infrastructure buildout: a verdict that says the largest cloud contract in history, when financed on a single investment-grade balance sheet, is straining the boundaries of investment-grade credit metrics.

The equity market's muted reaction — Oracle shares initially gained on the downgrade day before settling — is better understood as exhaustion than indifference. The stock is down roughly 28% year-to-date and nearly 60% from its 52-week high of $345.72. The credit deterioration was already largely priced in. But the credit market has been signaling distress for months. Oracle's 5-year CDS spread exceeded 198 basis points in March 2026, the widest level in the company's history. The company's 4.125% notes due 2045 trade at 68.6 cents on the dollar. This is the profile of a credit that the bond market treats with considerably more caution than the rating agencies' bottom-of-investment-grade badge suggests.

Here is what happened, why, and what to expect.

The Rating Action in Context

S&P's downgrade was not a surprise in direction — the agency had placed Oracle on negative outlook in July 2025 — but the paradox inside the decision is what matters. S&P actually improved several of its underlying credit metric forecasts even as it cut the rating. Peak leverage for fiscal 2027 was revised down from 4.8x to 4.4x. Adjusted EBITDA for 2027 was raised from $53.76 billion to $56.72 billion. Free operating cash flow is projected to turn positive by fiscal 2029 at $6.2 billion.

The agency improved its numbers and cut the rating anyway. This tells you the downgrade was driven by a reassessment of structural risk — not deteriorating metrics. The stable outlook S&P assigned alongside the downgrade reinforces this: the agency sees the 4.4x peak leverage as within its tolerance for the BBB- category, and does not expect a further cut in the next 6-24 months, provided Oracle executes on its financing plan.

Two structural drivers prompted the cut:

First, capital expenditure has escalated far beyond prior expectations. S&P's prior FY2027 capex forecast was $60 billion; it is now $90-95 billion. The resulting free operating cash flow deficit — projected at negative $42 billion for FY2027 — is 75% larger than the prior estimate of negative $24 billion.

Second, customer concentration risk of a kind no other investment-grade technology company carries. Both are direct consequences of the $300 billion, five-year contract Oracle signed with OpenAI in September 2025 to build 4.5 gigawatts of Stargate AI data center capacity.

The other rating agencies are positioned differently. Moody's holds Oracle at Baa2 (equivalent to BBB) with a negative outlook — meaning Moody's next move is more likely down than up. A cut to Baa3 would align Moody's with S&P at BBB-. Fitch maintains a mid-range BBB with a stable outlook, making it the most lenient of the three. This split rating — where agencies disagree by a notch or more — is itself a warning signal for credit investors, as it typically precedes convergence at the lower level.

The Numbers: A Cash Flow Deficit Without Modern Precedent

Oracle's fiscal 2026 results (year ended May 31, 2026) tell a story of extraordinary growth colliding with extraordinary capital intensity.

Several things are worth noting about this table. First, the gap between GAAP and non-GAAP operating income is $8.3 billion — a 28.7% adjustment that is heavily driven by stock-based compensation, a real economic cost that the non-GAAP figure obscures. For credit investors, GAAP operating income ($20.6B) is the more conservative measure of cash-generating capacity.

Second, the $638 billion in remaining performance obligations — up 363% year-over-year and up $85 billion sequentially in Q4 alone — is the largest contracted backlog in the cloud industry. Oracle's FY2027 guidance of $90 billion in total revenue implies a 33.5% growth rate. Non-GAAP EPS guidance of $8.05 implies approximately 16.4x forward earnings at the current ~$132 share price.

Third, the single most important number is negative $42 billion: S&P's forecast for Oracle's FY2027 free operating cash flow deficit. To contextualize: Oracle must finance a $42 billion annual cash shortfall — on top of the $23.7 billion deficit already recorded in FY2026 — through a combination of debt and equity issuance while protecting a credit rating that sits one notch above junk.

The leverage trajectory tells the rest of the story. Oracle's adjusted debt-to-EBITDA ratio was 3.3x as of May 2025. S&P expects this to peak at approximately 4.4x in fiscal 2027. While S&P's framework is not governed by a rigid threshold, leverage in the mid-4x range is at the upper end of what the agency considers consistent with a BBB- profile — particularly for a company with the concentration risks described below.

The critical question is when cash flow turns. S&P projects FOCF turning positive by fiscal 2029, reaching approximately $6.2 billion. That projection depends on three assumptions: (1) the OpenAI contract ramps toward its full billing rate (approximately $60 billion per year at scale, though revenue will follow an S-curve as data centers come online rather than a linear ramp), (2) capex normalizes as the initial buildout completes, and (3) no material deterioration in Oracle's legacy software and applications businesses. If any of these assumptions slip by 12-18 months, the inflection point moves from 2029 to 2030 or beyond.

One partial mitigant: Oracle disclosed in its FY2026 earnings release that $75 billion of its large AI contracts involve customer prepayments or customer-supplied hardware. This substantially reduces the capital Oracle must raise — though S&P's -$42B FOCF forecast presumably already incorporates this offset.

The OpenAI Problem: Concentration as Structural Credit Risk

S&P described OpenAI as a central credit risk for Oracle — language that is unusual for a rating agency and that signals the analysts concluded Oracle's creditworthiness cannot be assessed independently of OpenAI's business viability.

The concentration is extreme. OpenAI accounts for approximately half of Oracle's $638 billion in remaining performance obligations. By comparison, the largest customers of AWS, Google Cloud, and Microsoft Azure each represent low single-digit percentages of revenue. Oracle has effectively made a single startup the cornerstone of its growth strategy and its credit profile.

The structural problem extends beyond revenue concentration to asset specificity. According to industry analysis, an estimated $135 billion in AI systems hardware is being deployed for the OpenAI contract — purpose-built GPU clusters (predominantly NVIDIA GB200 racks) configured for OpenAI's workload requirements. If OpenAI falters, cannot meet its payment obligations, or renegotiates terms, Oracle is left with highly specialized data center capacity that cannot be easily repurposed or transferred to other customers on comparable economic terms.

Moody's characterized the Stargate project as effectively one of the world's largest project financings — a framing that is precise and alarming. Traditional project finance uses a special-purpose vehicle (SPV) that ring-fences infrastructure risk from the corporate balance sheet. The SPV borrows to build, then repays lenders from operating cash flows. Oracle is performing this function for OpenAI without the ring-fencing. Corporate credit risk and infrastructure project risk have been collapsed onto a single investment-grade issuer's balance sheet.

The risk is not theoretical. OpenAI has postponed its planned IPO to 2027. SoftBank recently reduced an OpenAI share-backed loan from $10 billion to $6 billion due to valuation uncertainties, according to reporting in The Decoder. The AI industry's path to profitability remains unproven for most participants. If OpenAI's ability to raise capital falters — if its models lose competitive ground to open-source alternatives, if compute demand plateaus, if its corporate restructuring encounters regulatory friction — Oracle's $300 billion backlog becomes a fraction of its face value, and the specialized hardware becomes a stranded asset.

A further structural disadvantage: unlike its larger cloud rivals, Oracle lacks substantial internal AI workloads that could absorb excess data center capacity if a major customer defected. AWS runs Amazon's own AI services, Google Cloud runs DeepMind and Gemini, and Azure runs Microsoft Copilot across its product suite. These first-party workloads provide a natural hedge against customer churn. Oracle has no equivalent buffer — its AI infrastructure is built for rent, not for use.

The Capital Structure: Equity Dilution as the Last Line of Defense

Oracle is doing something unprecedented in its 49-year history: funding its growth primarily through equity dilution rather than debt.

In fiscal 2026, Oracle raised $43 billion in debt financing and $5 billion through equity (including a mandatory convertible preferred stock issuance in February 2026). For fiscal 2027, the company plans to raise approximately $40 billion through a combination of debt and equity, including a previously announced $20 billion at-the-market equity issuance. Management has stated it does not expect to issue additional debt in calendar year 2026 — a significant constraint for a company with a $42 billion annual cash flow deficit.

The pivot to equity is not a strategic preference. It is a necessity. At approximately $167 billion in total debt and a leverage ratio heading toward 4.4x, the bond market has effectively closed to Oracle at investment-grade pricing. (Note: the $43B in FY2026 debt financing figure from Oracle's press release does not fully reconcile with the $18B September 2025 and $38B October 2025 debt raises cited in litigation filings — a combined $56B. The difference likely reflects fiscal year mapping, repayments, or project-level vs. corporate-level debt distinctions. Oracle's FY2026 ended May 31, 2026.)

Additional debt issuance at scale would risk pushing leverage past the threshold that would trigger a further downgrade. A fall to BB+ — the first junk rating — would raise borrowing costs, narrow the investor base to high-yield funds, and force index-tracking investment-grade funds to liquidate their Oracle holdings. Investment-grade corporate bonds are typically covenant-lite (negative pledge, limitations on liens), so the mechanical risk is less about covenant breaches and more about index ejection and repricing.

The equity dilution is the price of preserving investment grade. Oracle has approximately 2.8 billion shares outstanding. A $20 billion at-the-market issuance at current prices (~$132) would add roughly 150 million shares — approximately 5.4% dilution. Management has signaled expectations to raise tens of billions more over the following three years, implying cumulative dilution that could reach 15-20% before the capex cycle normalizes.

Larry Ellison, Oracle's chairman and CTO, who built the company on a philosophy of debt financing and share buybacks for decades, is now presiding over the largest equity dilution in the company's history. The same executive who made Oracle a serial buyback machine is issuing shares at decade-low prices to fund a bet that the bond market has yet to fully endorse.

Credit Market Signals: What the Bond Market Is Pricing

The rating agencies may have Oracle at the bottom of investment grade, but the bond market has rendered its own judgment — and it is harsher.

CDS spreads. Oracle's 5-year CDS spread exceeded 198 basis points in March 2026 — an all-time high for the company. For context, the average BBB- industrial company trades at approximately 130-150 basis points. Oracle is pricing 50-70 basis points wider than its rating category. (Note: Oracle's CDS was significantly tighter during the 2008 financial crisis, but that comparison conflates different risk profiles — the 2008 spread reflected systemic stress on an underlevered, pre-cloud company, while the 2026 spread reflects company-specific leverage and concentration risk.)

Cash bond prices. Oracle's 4.125% senior notes due 2045 trade at 68.6 cents on the dollar, yielding approximately 7.2%. With 19 years to maturity, the deep discount partly reflects the below-market coupon (4.125% vs. current investment-grade yields of ~5-5.5%) — some of the price decline is pure duration and coupon effect, not credit distress. But the 7.2% yield still embeds a credit spread of roughly 270 basis points over the 19-year Treasury, which is consistent with a lower-tier BB credit, not BBB-. The shorter-dated 5.5% notes due August 2035 were recently quoted at $94.43, implying a yield of approximately 6.3% and a tighter spread. This normal upward-sloping spread curve — wider at the long end — reflects the market's view that Oracle's credit risk is concentrated in the medium-to-long term, when the capex cycle either proves out or doesn't.

Analyst view. Barclays' credit strategy team maintains an Overweight recommendation on Oracle bonds, according to reporting by MacroStream. They argue that credit spreads already reflect substantial pessimism and note that Oracle's spreads are wider than those of Charter Communications (rated BBB-/BB+ in a split rating). The thesis is essentially mean-reversion: if Oracle's metrics improve as forecast, current spread levels offer value. The risk is asymmetric — from BBB-, there is exactly one notch of cushion before junk.

Peer Comparison: Oracle vs. the Hyperscaler Credit Complex

\* Cloud revenue figures are not strictly comparable across companies — Microsoft includes Azure (IaaS+PaaS) plus cloud SaaS, Google includes GCP plus Workspace, Amazon is AWS (IaaS+PaaS). Oracle's figure is IaaS ($18.1B) + SaaS ($15.9B). The comparison is directional.

\\ Approximate, based on most recent reported segments.

The gap is stark. Microsoft and Alphabet carry AAA/AA+ ratings with negligible leverage and massive free cash flow generation. They fund AI infrastructure buildout from operating cash flow. Amazon, with higher leverage, still generates positive free cash flow and carries an AA rating.

Oracle's distinguishing risk is visible in the "Largest Customer" row. No other hyperscaler has a single customer representing anywhere near 50% of its contracted backlog. AWS, Azure, and Google Cloud benefit from extreme customer diversification — losing any single customer would be material but not existential. For Oracle, losing or impairing the OpenAI contract would be catastrophic.

A new competitive dynamic has emerged as well. SpaceX — now public following its June 2026 IPO — has begun renting compute capacity to Anthropic and Alphabet, leveraging its own infrastructure investments. Combined with Meta's announcement that it will sell excess AI computing capacity to enterprise customers, these moves signal that compute supply may be catching up to demand. If hyperscalers have enough capacity to start selling the surplus, pricing power shifts from providers to buyers — and Oracle's contract economics deteriorate.

The Litigation Wildcard

In January 2026, the Ohio Carpenters' Pension Plan filed a class-action lawsuit against Oracle in New York state court, alleging that the company misled investors who purchased $18 billion in senior notes during a September 2025 bond offering.

The timing is the core of the allegation. Oracle announced its $300 billion OpenAI contract in September 2025. Two weeks later, it issued $18 billion in bonds. The offering documents described additional debt as something Oracle "may" need. Seven weeks after that, Oracle borrowed an additional $38 billion to fund data center projects — specifically two facilities developed by Vantage Data Centers as part of the Oracle-OpenAI agreement, according to Data Center Dynamics.

The plaintiffs argue that the "may" language was misleading because management was already planning the $38 billion raise at the time of the September offering. When news of the additional borrowing broke in October, Oracle's senior notes sold off sharply — a $4 billion segment of the September offering declined to approximately 95.75 cents on the dollar, representing roughly $170 million in losses on those securities.

Named defendants include Oracle, Chairman Larry Ellison, CEO Safra Catz, and EVP/Chief Accounting Officer Maria Smith, along with the underwriters of the bond offering. CoreWeave faces a similar class action filed the same week, and the Quinn Emanuel law firm has published a client alert on emerging litigation risks in AI data center financing — signals that the plaintiffs' bar sees an industry-wide pattern.

The lawsuit's outcome is uncertain, but its existence adds pressure on Oracle's financing strategy. If the plaintiffs prevail or discovery reveals internal communications suggesting management knew the $38 billion raise was imminent during the September offering, it could result in significant damages and tighter disclosure requirements for future offerings.

What to Watch: Catalysts and Triggers

The next 12-18 months will determine whether Oracle stabilizes at BBB- or faces further pressure. Specific catalysts to monitor:

1. The FY2027 capex trajectory. S&P's downgrade was driven by the escalation from $60 billion to $90-95 billion. Watch for quarterly capex disclosures and commentary about component cost inflation. If capex exceeds $95B, the FOCF deficit and leverage peak will be worse than forecast.

2. OpenAI's financial trajectory. OpenAI's ability to meet its contractual obligations depends on its ability to continue raising external capital at escalating valuations. Key signals: the IPO timeline (currently 2027), SoftBank financing activity, OpenAI's own revenue disclosures, and competitive dynamics in the foundation model market. If open-source models erode OpenAI's pricing power, the credit risk embedded in Oracle's RPO becomes acute.

3. The $20 billion equity issuance. The pace and pricing of the at-the-market equity program is the primary mechanism protecting the BBB- rating. If the offering struggles or requires a discount, it signals waning confidence.

4. RPO conversion rate. Oracle's $638 billion RPO is a promise, not a payment. The key metric is how quickly it converts to recognized revenue. In FY2026, Oracle recognized $67.4 billion in total revenue while RPO grew by $85 billion in Q4 alone. Once data centers come online and OpenAI begins drawing down capacity, revenue should accelerate — but any construction delays (power procurement, permitting, supply chain) push the revenue inflection further out.

5. Moody's next move. Moody's currently rates Oracle Baa2 with a negative outlook. A downgrade to Baa3 would align Moody's with S&P at BBB-. Fitch would remain at BBB — one notch above — until and unless it also acts. The scenario that matters is if two of three agencies move to junk territory, which would trigger broad-based forced selling.

6. Macro AI demand signals. The thesis rests on continued growth in AI compute demand. Any signal of demand softening — enterprise AI budget cuts, model training efficiency improvements, regulatory constraints — would simultaneously pressure OpenAI's viability, Oracle's RPO conversion, and the broader hyperscaler capex cycle. The Bank for International Settlements has explicitly warned of parallels between debt-financed AI investments and the 2008 financial crisis.

7. Competitive dynamics. SpaceX entering AI compute leasing, Meta selling excess capacity, and potential new entrants signal supply may be catching up to demand. The entry of well-funded competitors into a space where Oracle is betting its balance sheet creates incremental pricing pressure.

8. The bondholder lawsuit. Any ruling on motions to dismiss, discovery revelations, or settlement discussions will affect Oracle's credit spreads and disclosure obligations.

Investment Implications

For Credit Investors

Oracle bonds at current prices offer a genuine risk-reward trade, but one with asymmetric downside. The Barclays Overweight thesis — that spreads have overshot — is defensible. The $75 billion in prepaid and customer-supplied hardware portions disclosed in the FY2026 earnings release substantially reduces the capital Oracle must raise, a positive not yet fully reflected in spread levels.

But the asymmetric risk is real. At BBB-, there is exactly one notch of cushion before junk. A downgrade to BB+ would trigger forced selling by investment-grade index funds, widening spreads dramatically. The 2045 notes at 68.6 cents already price significant distress, but a junk downgrade could push them toward 50-55 cents before finding a floor.

The optimal exposure is in the 3-7 year part of the curve, where the projected FOCF inflection point (2029) provides a plausible path to stabilization. Long-dated bonds carry duration risk on top of credit risk with limited additional spread compensation. Short-dated paper trades closer to par but offers less yield pickup.

For Equity Investors

The equity story is a leveraged play on the credit story. Oracle's $638 billion RPO, 77% IaaS growth, and $90 billion FY2027 revenue guidance are genuinely impressive. At roughly $370 billion market capitalization against $67.4 billion in FY2026 revenue, the stock trades at approximately 5.5x price-to-sales — not cheap for a company with negative free cash flow, but not egregious for one growing cloud infrastructure at 77% with a $638 billion backlog. Forward non-GAAP P/E of approximately 16.4x (at $8.05 EPS guidance) is the most bullish data point.

The risk is that equity is structurally subordinated to the credit. Every dollar of the $42 billion annual cash deficit must be financed, and management has chosen equity dilution as the primary mechanism. A $20 billion ATM issuance at current prices dilutes existing holders by approximately 5.4%. Additional raises over the following three years could push cumulative dilution to 15-20%.

The 21,000 job cuts (approximately 13% of workforce) implemented in the past twelve months represent a significant cost reduction that partially offsets cash burn. But reductions of this scale carry execution risk in a company attempting its most ambitious infrastructure expansion.

The Binary Core

The defining risk is the OpenAI contract. If it executes — data centers come online on schedule, OpenAI meets its payment obligations, FOCF turns positive by 2029 — Oracle stabilizes, the credit recovers, and the equity re-rates. If it stumbles — through OpenAI's own financial difficulties, demand softening, or construction delays — the credit goes to junk, the equity gets further diluted at distressed prices, and the specialized AI hardware becomes a stranded asset of historic scale.

S&P's stable outlook is the most important signal here. It means the agency believes the base case is stabilization, not deterioration — provided Oracle executes its equity financing plan and the OpenAI contract progresses. The market's wider spreads reflect a less benign view. The gap between S&P's stable outlook and the bond market's pricing is itself the trade: either the spreads tighten (bonds rally) or the outlook deteriorates (equity suffers further).

Watch the capex trajectory, the OpenAI ramp, the equity issuance pace, and the bondholder lawsuit. These four indicators will determine whether Oracle's BBB- rating is the floor or just a waypoint.