Executive assessment
Oracle can plausibly produce several more years of negative free cash flow while continuing to pay its debt. The current evidence supports a financing-dependence thesis more strongly than an imminent inability to pay interest. Theseus should focus on the combination of persistent investment deficits, weaker customer funding and impaired access to capital. Negative free cash flow alone does not establish a default or an attractive CDS purchase.
The sustained-spending scenario projects free cash flow of negative $37bn in FY2027, negative $30bn in FY2028 and negative $12bn in FY2029, recovering to positive $5bn in FY2030. These are analyst assumptions, not management guidance or probability-weighted forecasts. The FY2027 investment assumption deliberately exceeds the latest spending guidance; a guidance-aligned sensitivity remains negative but is substantially easier to finance.
The decisive funding statement
Starting with reported August liquidity, the sustained-spending model falls below a $10bn operating buffer in February 2027 and reaches a $5.045bn cash deficit in May 2027 if no new external funding or corrective action occurs. It needs $15.045bn to retain the buffer. A hypothetical $20bn net financing package covers that need and leaves $14.955bn. The severe case still has a $5.045bn cash deficit after that package. These are conditional modeled balances, not predicted missed payments.
What the debt analysis establishes
Q1 operating income covers reported interest expense approximately 4.71 times. This is a useful earnings test, not a cash debt-service ratio or a contractual covenant calculation. The covenant uses a different earnings definition and net cash interest. Principal repayment depends on cash left after investment, liquidity and refinancing, so the interest cushion does not resolve the funding problem. [1, 4]
Investment judgment
Oracle fits the Theseus thesis if infrastructure commitments remain difficult to reduce after AI customer economics deteriorate. Its established earnings, access to equity and long maturity profile are material counterarguments. The stronger near-term expression could be spread widening, which can occur without default. A protection position still requires evidence that the expected deterioration exceeds what its premium already prices.
Public reporting supports the existence of an Oracle CDS market, with five-year protection near 300 basis points on October 15. That is a dated market indication, not a firm executable quote. At that level, meaningful credit stress is already priced. The analysis supports continued underwriting, not an unconditional buy recommendation. [7]
Reported cash flow and the advance funding effect
| Reported cash flows [1] | FY2026 | Q1 FY2027 |
|---|---|---|
| Operating cash flow | 31.977 | 23.103 |
| Gross cash capital expenditures | (55.663) | (28.499) |
| Free cash flow | (23.686) | (5.396) |
Q1 customer prepayments with a significant financing component were $11.363bn. Reported cash flow less this specified item is $11.740bn. Q1 net cash outlay for capital expenditures was $17.966bn, incorporating $0.830bn of additional short-term financing outflow. The equity ATM raised $19.909bn net during the quarter. [1]
Avoid counting customer money twice
Reported operating cash flow already includes the advances. Subtracting net capex from that operating cash flow would overstate cash remaining. The consistent calculation is $23.103bn less $28.499bn gross capex less $0.830bn financing outflow, giving negative $6.226bn before other investing and financing items. Equivalently, $11.740bn of operating cash excluding those advances less $17.966bn net capital outlay gives the same result.
Removing advances is a diagnostic, not a claim that the receipts are fictitious or refundable on demand. The residual still contains ordinary working-capital changes, deferred revenue, taxes and interest. It is not recurring cash earnings. The forecast therefore models the specified advances separately while leaving other operating cash movements inside the residual.
Why cash today can create obligations tomorrow
An advance funds construction before service delivery. It helps liquidity now but cannot be collected again when the associated revenue is recognized. If replacement advances slow while previously contracted service costs rise, revenue and EBITDA can grow faster than fresh cash receipts. Equally, reliable advances and customer-owned hardware can reduce Oracle’s capital burden and weaken the protection thesis.
Current liquidity and commitments
| August 31 balance or commitment [2] | Amount |
|---|---|
| Cash and equivalents excluding restricted cash | 36.369 |
| Current and noncurrent notes and borrowings | 125.337 |
| Recognized operating lease liability | 34.621 |
| Recognized finance lease liability | 9.185 |
| Additional leases not yet commenced | 288.000 |
| Purchase and certain other obligations of at least one year | 34.150 |
Additional leases are expected to start from Q2 FY2027 through FY2029, generally lasting 15–19 years. They are not all payable today. [2] The model excludes restricted cash and marketable securities from its opening buffer. Lease obligations, purchase commitments and capex cannot simply be added together: spending classifications overlap, and a project-level reconciliation is needed.
Future free cash flow under explicit assumptions
All values below are Theseus scenarios. Operating cash is after interest, taxes, operating lease payments and ordinary working capital. “Advances” means the separately identified customer financing component, not every deferred-revenue movement. Capex is gross cash spending, including refresh investment and cash settlement of equipment payables. No acquisition or asset-sale proceeds are assumed.
| Sustained spending | FY2027 | FY2028 | FY2029 | FY2030 |
|---|---|---|---|---|
| Operating cash excluding specified advances | 43 | 50 | 58 | 65 |
| Specified customer advances | 25 | 20 | 15 | 10 |
| Total operating cash | 68 | 70 | 73 | 75 |
| Gross cash capex | (105) | (100) | (85) | (70) |
| Free cash flow | (37) | (30) | (12) | 5 |
The FY2027 residual is below four times Q1’s diagnostic residual; subsequent growth is approximately 16%, 16% and 12%. These assumptions allow a substantial improvement in operating cash generation while new advances decline. Capex declines only gradually because delivery and equipment refresh continue. The scenario does not require Oracle’s entire operating business to collapse.
Management’s September call maintained gross FY2027 capex of $90–95bn and net cash capex no higher than $70bn. [6] Our $105bn gross assumption is $10bn above that upper end. Holding operating cash at $68bn but using $95bn capex gives negative $27bn FCF. Thus negative FCF does not depend on the spending overrun; the size and timing of the funding gap do.
| Alternative inputs and results | FY2027 | FY2028 | FY2029 | FY2030 |
|---|---|---|---|---|
| Defensive residual operating cash | 45 | 58 | 70 | 80 |
| Defensive advances | 25 | 25 | 20 | 15 |
| Defensive gross capex | (95) | (90) | (80) | (75) |
| Defensive free cash flow | (25) | (7) | 10 | 20 |
| Severe residual operating cash | 38 | 40 | 42 | 45 |
| Severe advances | 20 | 10 | 5 | 0 |
| Severe gross capex | (115) | (110) | (95) | (80) |
| Severe free cash flow | (57) | (60) | (48) | (35) |
Interpretation and model limits
The defensive case combines guidance-level FY2027 spending with stronger later cash conversion. The severe case combines weaker collections and advances with slow spending adjustment. No probabilities are assigned. The scenarios specify cash outcomes rather than a full revenue, margin and working-capital model; forecast accuracy depends especially on advance timing and cancellable versus committed spending. Future interest is embedded in the residual assumptions, not independently solved from the eventual financing mix. Every additional $1bn of cash interest or lease expense lowers FCF by $1bn unless offset elsewhere.
Quarterly liquidity and the financing requirement
The bridge begins August 31, 2026, not the research date. Subsequent undisclosed cash movements are modeled. Q1 actuals plus the three forecast quarters reconcile to the FY2027 sustained-spending assumptions. Quarterly maturities are allocations of the annual estimate, not verified payment dates; the first buffer-breach quarter is sensitive to actual timing.
| USD billions | Q2 Nov 2026 | Q3 Feb 2027 | Q4 May 2027 |
|---|---|---|---|
| Operating cash excluding specified advances | 9.000 | 10.000 | 12.260 |
| Specified advances | 5.000 | 4.000 | 4.637 |
| Gross capex | (26.000) | (25.000) | (25.501) |
| Free cash flow | (12.000) | (11.000) | (8.604) |
| Dividends assumption | (1.600) | (1.600) | (1.600) |
| Principal repayment allocation | (1.500) | (1.500) | (1.110) |
| Other net cash uses assumption | (0.300) | (0.300) | (0.300) |
| Closing cash without new funding | 20.969 | 6.569 | (5.045) |
| Illustrative new funding each quarter | 0.000 | 10.000 | 10.000 |
| Closing cash with that funding | 20.969 | 16.569 | 14.955 |
Cash formula: opening cash + FCF − dividends − debt principal − other net uses + new funding. Other uses allow for finance-lease principal, financing fees and miscellaneous investing/financing; future net supplier financing is assumed zero. The assumption may understate those uses. Interest is already in operating cash and is not deducted again.
What covers the gap
Oracle announced approximately $40bn of FY2027 debt and equity financing, including the $20bn ATM subsequently completed in Q1. [1, 5] The remaining approximately $20bn is an intention, not a committed facility or cash already received. It must not be added alongside the entire $40bn plan. The bridge assumes $20bn net new proceeds, slightly more favorable than $20bn gross after fees.
Without new financing, $5.045bn is needed just to avoid a negative May balance, and $15.045bn to retain the assumed $10bn buffer. In the severe case those needs rise to $25.045bn and $35.045bn respectively. A hypothetical fully available $10bn revolver alone would cover the sustained case’s zero-cash deficit but leave a $5.045bn shortfall to the buffer. A draw replaces liquidity risk with additional debt and interest; it is not a loss absorber.
The strongest immediate falsification test
Using $95bn capex instead of $105bn, with all other sustained-case assumptions unchanged, improves May cash by $10bn. Before new financing, cash would be $4.955bn; only $5.045bn is needed to retain the buffer. Timely financing, lower capex or stronger advances therefore materially postpone the stress. A forecast that ignores these remedies would overstate the default case.
Debt exposure and principal serviceability
| Contractual principal at May 31 [3] | USD bn |
|---|---|
| FY2027 | 7.210 |
| FY2028 | 10.145 |
| FY2029 | 5.500 |
| FY2030 | 7.250 |
| FY2031 | 9.750 |
| Thereafter | 90.250 |
| Total | 130.105 |
This is a principal schedule, not August carrying debt. The quarterly model conservatively allows $4.110bn for remaining FY2027 principal, subtracting only $3.1bn of Q1 scheduled repayments from $7.210bn. Q1 also repaid $1.1bn of commercial paper. [2, 3] To the extent that paper was included in the May maturity schedule and not reissued, the allowance overstates remaining payments by about $1.1bn; funding needs fall dollar for dollar. Exact residual maturities need lender-level confirmation. Current balance-sheet debt covers a rolling twelve months and is not the same as the remaining fiscal-year amount.
The relevant legal exposure
Oracle Corporation senior unsecured credit is the target for a consolidated funding thesis. Its notes rank with other parent unsecured borrowing; creditors at operating subsidiaries have claims ahead of distributions to the parent. [8] A project’s financing shortfall or GPU impairment is not automatically a parent payment default. The signed guarantee, borrower and cross-default provisions determine transmission. Equity issuance dilutes shareholders but can preserve creditors.
The $10bn unsecured revolving facility matures March 6, 2031 and was undrawn at May 31. [8] Availability on the research date is not assumed proven. Commercial paper capacity and revolver capacity should not be treated as two independent buffers where the facility backs the paper. Mandatory convertible preferred equity should not be modeled as a conventional bond principal maturity; its dividend still consumes cash.
Repeated funding even when interest is paid
| Sustained case after FY2027 financing | FY2028 | FY2029 | FY2030 |
|---|---|---|---|
| Opening cash | 14.955 | 10.000 | 10.000 |
| Free cash flow | (30.000) | (12.000) | 5.000 |
| Dividends | (6.400) | (6.400) | (6.400) |
| Existing scheduled debt principal | (10.145) | (5.500) | (7.250) |
| Other net uses | (1.000) | (1.000) | (1.000) |
| New funding required for $10bn closing cash | 42.590 | 24.900 | 9.650 |
The requirement includes refinancing existing maturities; it is not all incremental debt. Funding can be debt, equity or an operating remedy. Future debt issued to meet these needs could create additional maturities absent here. These are conditional financing requirements with the specified after-interest cash flows held fixed, not a fully integrated debt issuance forecast. A shift to more expensive debt would increase them.
Interest coverage and the covenant distinction
Q1 GAAP operating income was $6.728bn and interest expense $1.428bn. [1] Dividing gives 4.71 times. Annualizing both yields $26.912bn and $5.712bn; this is a run-rate sensitivity, not a forecast of annual earnings or cash coupon payments. It shows why negative FCF can coexist with an ability to pay interest before funding investment.
| Annualized EBIT scenario | Interest 5.712 | Interest 6.912 | Interest 9.712 |
|---|---|---|---|
| 26.912 unchanged | 4.71x | 3.89x | 2.77x |
| 20.000 weaker earnings | 3.50x | 2.89x | 2.06x |
| 15.000 severe earnings stress | 2.63x | 2.17x | 1.54x |
The middle interest column adds $20bn of new borrowing at an assumed 6%; the last adds a further $40bn at 7%. Both are full-year incremental interest sensitivities, not forecasts of financing amounts, coupons or issue dates. They ignore retired coupons and therefore isolate the cost of financing new deficits. Equity funding would avoid this coupon burden. Existing fixed-rate debt does not all reprice when market spreads rise.
The actual revolver test
The agreement requires at least 3.0 times Consolidated EBITDA to Consolidated Net Interest Expense. Its EBITDA definition includes specified adjustments and stock compensation; net interest uses cash interest and subtracts cash interest income. The earnings measure covers four quarters, with specified acquisition adjustments. [4] The EBIT/gross-interest table above cannot establish a breach of that covenant.
A proper compliance calculation requires the last four quarters under the signed definitions, including cash versus noncash interest, lease interest, qualifying adjustments and netting. Passing the covenant does not prove adequate cash after capex. Failing it can restrict facility access or accelerate obligations following the applicable contractual process, but it is not automatically the CDS credit event.
Four distinct meanings of serviceability
Interest service: the latest earnings proxy offers a meaningful cushion. Principal service: scheduled maturities require refinancing or cash reserves because projected FCF is negative. Investment funding: expansion competes with debt service for the same cash. Long-run solvency: completed assets must earn enough cash after maintenance and replacement to support creditors without perpetual subsidized funding. These tests can give different answers at the same time.
What a credible debt failure would require
The damaging combination is lower operating cash, fewer customer advances, inflexible spending and impaired financing access. Rating pressure can then raise new coupons and floating-rate margins, while falling cash reduces interest income. That feedback can erode net-interest coverage even before operating earnings fall sharply. The present evidence does not establish that all these conditions will occur together.
Failure modes remedies and the Theseus thesis
| Failure mode | Transmission to creditors | Evidence that would strengthen it |
|---|---|---|
| Customer funding weakens | Fewer advances and slower receipts while delivery costs persist | Advance receipts below plan; aging receivables; renegotiations |
| Construction or power delay | Cash spent before service receipts; potential penalties | Project completion slippage; delayed acceptance |
| Investment cannot be reduced | Committed spending crowds out liquidity | Noncancellable payments exceed available funds |
| Equipment economics deteriorate | Lower utilization and prices; higher refresh burden | Falling realized yields and renewal margins |
| Capital markets become restrictive | Higher cost or unavailable refinancing | Failed issuance; reduced facility availability |
| Software cash generation weakens | Less internal support for infrastructure | Cash margins fall outside OCI as well |
These are mechanisms to test, not reported events. A diversified, profitable company can absorb considerable project disappointment. The thesis becomes much stronger if customer stress and financing stress are correlated: the same market shock removes advances, equity issuance and debt capacity while contractual payments remain fixed.
The contractual work that matters most
For each major data center, identify the lessee, tenant, borrower and guarantor; delivery and acceptance milestones; when rent begins; remaining construction payments; delay remedies; termination rights; and whether reserves are unrestricted or project-specific. Map purchase commitments into capex and operating spending. Do not subtract the full long-term lease commitment from today’s liquidity or assume an unverified parent construction guarantee.
For customer contracts, obtain the prepayment timetable, credit support, minimum payment terms, performance conditions and refund or offset rights. A large backlog helps only insofar as Oracle can deliver and collect profitable cash on schedule. Future service revenue supported by advances already spent is not equivalent to a new future receipt.
Remedies and their limits
Capex deferral improves immediate liquidity, but only the cancellable portion is available and delayed capacity can delay receipts. Lower common dividends conserve cash but cannot close very large recurring investment deficits alone. Equity preserves debt service without adding coupons, at a cost to shareholders. Asset sales require time and may sacrifice future cash flow. Borrowing bridges timing but raises the future interest burden. None should be assumed unlimited; none should be ignored.
Evidence that would invalidate the protection case
The thesis weakens if contracted expansion converts to cash faster than spending, customers fund more hardware, the company repeatedly secures equity on acceptable terms, or infrastructure spending becomes flexible enough to protect liquidity. The defensive scenario’s return to positive FCF illustrates this path. A short-credit thesis must explain why those defenses fail, not merely why AI investment looks expensive.
Protection economics and investment decision
Reuters reported Oracle five-year CDS near 300bp on October 15. [7] For illustration, a 300bp par spread corresponds to about $300,000 annual premium-equivalent per $10m notional while the contract survives. Actual trading may use a standardized running coupon plus upfront payment; the reported spread is not itself the confirmation’s cash schedule.
| Assumed recovery | Approximate annual hazard at 300bp | Five-year implied event probability |
|---|---|---|
| 20% | 3.75% | 17.1% |
| 40% | 5.00% | 22.1% |
| 60% | 7.50% | 31.3% |
Illustrative constant-hazard approximation: hazard = spread / (1 − recovery); five-year probability = 1 − exp(−5 × hazard). These are risk-neutral pricing approximations, not physical default forecasts. They ignore curve shape, discounting, accrued premium, liquidity and counterparty effects. Higher assumed recovery requires a higher event probability to justify the same spread.
At 40% recovery, a $10m position would receive approximately $6m after a covered credit event, subject to settlement terms. Five years of uninterrupted premium-equivalent at 300bp would total $1.5m before discounting; premiums normally stop at a covered event. A spread-widening exit can make money without default, but its value depends on remaining risky duration, liquidity, bid–ask cost and timing.
Contract selection before execution
Request a firm two-way quote on Oracle Corporation senior unsecured USD protection, with legal entity identifier, exact reference obligation, maturity, notional and governing documentation. Verify covered events, restructuring treatment, succession, deliverability, auction settlement, upfront amount, running coupon, collateral and closeout terms. Also compare three-year and five-year tenors to the modeled funding horizon. No dealer has been contacted and no executable quote is established here.
Research priority and conclusion
First reconcile monthly project cash uses and customer advances against actual unrestricted cash. Second verify contractual maturity dates, committed facility availability and the covenant calculation. Third price the exact instrument and estimate expected spread paths and loss severity. The quarterly funding result is useful for directing those checks but cannot substitute for them.
Under the sustained-spending assumptions, Oracle needs $15.045bn by May 2027 to preserve a $10bn buffer, and the modeled $20bn package covers it. Under the severe assumptions, that same package leaves a $5.045bn cash deficit before remedies. Reported interest coverage provides a real defense, while repeated future funding requirements create the credit vulnerability. Oracle is consistent with Theseus as a conditional financing and spread-widening thesis; imminent default and favorable entry valuation remain unproven.
Sources and calculation notes
Public information reviewed through 17 October 2026. Latest financial period used: quarter ended 31 August 2026. The sources establish reported facts and contract terms; forecast inputs and sensitivities are Theseus analytical assumptions. Dollar rounding does not imply forecast precision.
[1] Oracle Q1 FY2027 earnings release, September 10, 2026. Cash flow, net capital outlay, income statement and ATM proceeds. Open source
[2] Oracle Q1 FY2027 Form 10-Q, September 11, 2026. Balance sheet, leases, purchase obligations and liquidity discussion. Open source
[3] Oracle FY2026 Form 10-K principal maturity table, May 31 balances. Open source
[4] Oracle revolving credit agreement dated March 6, 2026, Exhibit 10.17. Defined terms and financial covenant. Open source
[5] Oracle FY2026 earnings release, June 10, 2026. Fiscal 2027 financing plan. Open source
[6] Oracle Q1 FY2027 management call, September 10, 2026, transcript hosted by Roic AI. Gross and net capex guidance. Open source
[7] Reuters, AI credit worries send default insurance to record high, October 15, 2026; syndicated by Investing.com. Open source
[8] Oracle FY2026 Form 10-K, June 22, 2026. Borrowing arrangements, ranking and revolving facility. Open source
Model reproduction
FCF = residual operating cash + specified advances − gross cash capex. Required funding = max(0, minimum cash − opening cash − FCF + dividends + principal + other uses). The minimum cash assumption is $10bn and is not a contractual requirement. Future interest sensitivity must be fed back into residual operating cash when the funding mix is known.
No formal default probability, recovery valuation or investment sizing is assigned. Project recourse, covenant headroom and market access require instrument-level verification before capital is committed. The latest public cash balance should not be presented as current-day liquidity.