Summary

The state difference: financing effective, spending conditional

Corporate announcements are cheap; balance-sheet entries are not. By the quarter ended August 31, 2026, Oracle had converted two announcements into irreversible facts. It completed the sale of $20 billion of common stock (before commissions) through an at-the-market equity program during Q1 FY2027, as part of its disclosed capital investment program (https://investor.oracle.com/investor-news/news-details/2026/Oracle-Announces-Q1-Results-Driven-by-Triple-Digit-Growth-in-Cloud-Infrastructure-Revenues/default.aspx). One granular secondary reconstruction reports 141 million common shares sold for $19.909 billion after issuance costs, a figure the filing-level review in this article treats as unverified beyond the issuer's own headline (https://digidai.github.io/2026/09/13/oracle-stock-sale-ai-restructuring-plan/).

The debt side was executed earlier and in more detail. On February 2 and 4, 2026, Oracle issued $25.0 billion aggregate principal of senior unsecured notes in eight tranches maturing between 2029 and 2066: $500 million of floating-rate notes at Compounded SOFR plus 1.11% due 2029; $3.0 billion of 4.550% notes due 2029; $3.5 billion of 4.950% notes due 2031; $3.0 billion of 5.350% notes due 2033; $5.0 billion of 5.700% notes due 2036; $2.25 billion of 6.550% notes due 2046; $5.0 billion of 6.700% notes due 2056; and $2.75 billion of 6.850% notes due 2066 (https://www.sec.gov/Archives/edgar/data/1341439/000119312526035603/d33906d424b2.htm; https://www.sec.gov/Archives/edgar/data/1341439/000119312526037840/d23732dex41.htm; https://www.ifre.com/bonds/2379081/oracle-raises-us25bn-in-bond-offering-as-ai-funding-spree-continues). Estimated net proceeds were about $24.9 billion. The notes rank equally with existing senior unsecured debt and are structurally junior to subsidiary liabilities — a term that matters for anyone reasoning about which creditors stand where in a stress scenario. The long-end coupons tell the market-clearing story: money lent to Oracle for 30 years prices at 6.70%, for 40 years at 6.85%, against a 4.55% coupon at the three-year point.

Everything else in Oracle's AI buildout remains conditional. The Form 10-Q for the quarter ended August 31, 2026, filed September 11, 2026, disclosed $288 billion of additional lease commitments, substantially all data-center arrangements, not yet on the balance sheet and generally expected to commence between Q2 FY2027 and FY2029 with 15-19 year terms; Oracle ended the quarter with roughly $125 billion of notes payable and other borrowings — a figure that excludes lease liabilities, as the filing-derived summary makes explicit (https://www.ibtimes.sg/oracle-stock-falls-4-ai-data-center-faces-18b-debt-cost-crisis-94226; https://last10k.com/sec-filings/orcl/0001193125-26-389274.htm).

What the February plan committed

The plan Oracle announced on February 1, 2026 was unusually specific for a capital program: raise $45-50 billion of gross cash proceeds during calendar 2026, split roughly evenly between debt and equity, to maintain what the company called a solid investment-grade balance sheet. Equity would come from an initial mandatory convertible preferred issuance plus a new at-the-market program of up to $20 billion. Debt would come from "a single, one-time issuance of investment-grade senior unsecured bonds early in 2026," with no additional bond issuance expected during calendar 2026. Oracle named AMD, Meta, NVIDIA, OpenAI, TikTok and xAI among the largest OCI customers whose contracted demand the financing was to support. Goldman Sachs led the bond; Citigroup led the equity (https://investor.oracle.com/investor-news/news-details/2026/Oracle-announces-Equity-and-Debt-Financing-Plan-for-Calendar-Year-2026/default.aspx).

Two parts of that plan deserve scrutiny. First, the single-bond commitment: the company does not expect to issue additional bonds in calendar 2026, yet Fitch forecast incremental debt and equity issuance through at least fiscal 2028, and CreditSights estimated roughly $70 billion of debt issuance needs across fiscal third-quarter 2026 through fiscal fourth-quarter 2028 (https://www.fitchratings.com/research/corporate-finance/fitch-rates-oracle-senior-notes-offering-bbb-affirms-idr-outlook-stable-02-02-2026; https://know.creditsights.com/insights/oracle-top-20-client-faqs/). These statements reconcile only if further issuance falls outside calendar 2026 or arrives through non-bond financing. That reconciliation is unresolved, and it is one of the article's open questions: the observable test is whether Oracle enters 2027 with a new bond mandate.

Second, the balance-sheet target. Maintaining investment grade is not a preference; it is a market-access condition. S&P's July 9, 2026 action moved Oracle to BBB-, the lowest investment-grade notch (https://www.heise.de/en/news/S-P-downgrades-Oracle-to-BBB-only-one-notch-above-junk-level-11363472.html; https://cbonds.com/news/4003647/). Moody's affirmed Baa2 with a negative outlook on February 2, 2026 (https://www.newsquawk.com/headlines/moody-s-affirms-oracle-s-orcl-baa2-rating-on-equity-announcement-outlook-remains-negative-02-02-2026), and Fitch held BBB/Stable while explicitly modelling negative free cash flow (https://www.fitchratings.com/research/corporate-finance/fitch-rates-oracle-senior-notes-offering-bbb-affirms-idr-outlook-stable-02-02-2026). The agencies disagree on trajectory but agree on the mechanics: Oracle's financial flexibility is below that of Microsoft, Google or Amazon, in S&P's judgement, and the margin above the high-yield boundary is now one notch on one axis.

The quarter's numbers and what they mean

Oracle's Q1 FY2027 results were strong on demand and weak on conversion — which is exactly the combination the financing-stack thesis predicts. Total revenues were $19.3 billion, up 30%; cloud revenues rose 62% to $11.6 billion, with Cloud Infrastructure up 121% to roughly $7.4 billion; GAAP EPS was $1.56 and non-GAAP EPS $1.92, ahead of the $1.74 LSEG consensus on earnings (https://investor.oracle.com/investor-news/news-details/2026/Oracle-Announces-Q1-Results-Driven-by-Triple-Digit-Growth-in-Cloud-Infrastructure-Revenues/default.aspx; https://www.cnbc.com/2026/09/10/oracle-orcl-q1-earnings-report-2027.html). Remaining performance obligations reached $664 billion after more than $30 billion of additional AI cloud contracts, exceeding StreetAccount's $630.6 billion consensus. Oracle delivered 850 megawatts of additional data-center capacity and more than 300,000 GPUs in the quarter (https://investor.oracle.com/investor-news/news-details/2026/Oracle-Announces-Q1-Results-Driven-by-Triple-Digit-Growth-in-Cloud-Infrastructure-Revenues/default.aspx; https://www.fool.com/earnings/call-transcripts/2026/09/11/oracle-orcl-q1-2027-earnings-call-transcript/; https://www.cnbc.com/2026/09/10/oracle-orcl-q1-earnings-report-2027.html).

The cash statement is the counterweight. Capital expenditures were $28,499 million for the quarter versus $8,502 million a year earlier; the net cash outlay for capital expenditures was $17,966 million; operating cash flow was $23 billion, up 184%; free cash flow was negative about $5 billion on the issuer's framing, or negative $5.4 billion per CNBC, against negative $362 million a year earlier (https://investor.oracle.com/investor-news/news-details/2026/Oracle-Announces-Q1-Results-Driven-by-Triple-Digit-Growth-in-Cloud-Infrastructure-Revenues/default.aspx; https://www.cnbc.com/2026/09/10/oracle-orcl-q1-earnings-report-2027.html).

The bridge between those two lines is the mechanism this article tracks. Management reiterated full-year FY2027 capex guidance of $90-95 billion with not more than $70 billion in net cash capex — the difference being customer prepayments and other short-term financing netted against gross spend (https://www.fool.com/earnings/call-transcripts/2026/09/11/oracle-orcl-q1-2027-earnings-call-transcript/). Oracle received $11.4 billion of customer prepayments carrying a significant financing component in the quarter (https://digidai.github.io/2026/09/13/oracle-stock-sale-ai-restructuring-plan/). Management said the vast majority of new Q1 RPO was contracted via prepay, bring-your-own-hardware or similar mechanics that "will not require incremental capital from Oracle" — and then the CEO corrected the record: the structures avoid Oracle having to find additional cash, not the capital expenditure itself (https://www.fool.com/earnings/call-transcripts/2026/09/11/oracle-orcl-q1-2027-earnings-call-transcript/). CFO Hilary Maxson also said Oracle has not given a timeframe for a return to positive free cash flow, while asserting that each project reaches roughly 100% free-cash-flow conversion to post-tax EBITDA shortly after ramp (https://www.fool.com/earnings/call-transcripts/2026/09/11/oracle-orcl-q1-2027-earnings-call-transcript/). Management additionally said new RPO booked in Q1 will not affect capex or revenues until fiscal 2028 or beyond, and that about half of RPO should convert to sales over the next 36 months (https://www.fool.com/earnings/call-transcripts/2026/09/11/oracle-orcl-q1-2027-earnings-call-transcript/).

That is the operating mechanism stated plainly: demand converts to revenue only after capacity commences; capacity requires the lease queue to begin on schedule; and the queue's funding requires Oracle to keep both prepayments flowing and investment-grade market access intact. Each link showed a measurable signal in September 2026.

Link one: credit-market access

The rating agencies' February-to-July sequence frames the constraint. Moody's rated Oracle Baa2 negative as far back as its December 2025 credit outlook, when RPO stood at $523 billion, up $68 billion in a quarter, with counterparty risk and required spending named as the primary drivers; it affirmed Baa2 negative on February 2, 2026, on the equity announcement (https://www.moodys.com/web/en/us/insights/credit-risk/outlooks/mco-15-december-2025.html; https://www.newsquawk.com/headlines/moody-s-affirms-oracle-s-orcl-baa2-rating-on-equity-announcement-outlook-remains-negative-02-02-2026). Fitch assigned BBB to the February bonds and affirmed the IDR at BBB/Stable, forecasting EBITDA leverage above 3.5x in FY2026, EBITDAR peaking above 4x, negative pre-dividend free cash flow of more than $26 billion in FY2026 and $18 billion in FY2027, and incremental debt and equity issuance through at least fiscal 2028 (https://www.fitchratings.com/research/corporate-finance/fitch-rates-oracle-senior-notes-offering-bbb-affirms-idr-outlook-stable-02-02-2026).

S&P moved first to the boundary: on July 9, 2026 it downgraded Oracle from BBB to BBB-, the lowest investment-grade notch. In that action S&P forecast a fiscal 2027 free-operating-cash-flow deficit of almost $42 billion financed with debt and equity, noted that Oracle had raised FY2027 spending guidance to $90-95 billion against S&P's prior $60 billion assumption — an increase the agency attributed to rising GPU and networking component costs — and described OpenAI as a "central credit risk," attributing roughly half of $638 billion of contracted-but-undelivered service volume to OpenAI (https://www.heise.de/en/news/S-P-downgrades-Oracle-to-BBB-only-one-notch-above-junk-level-11363472.html). Contemporaneous accounts including Cbonds and heise describe the post-downgrade outlook as stable; S&P's own page was not directly accessible to verify the body text, so the outlook characterization rests on those secondary accounts (https://cbonds.com/news/4003647/; https://www.spglobal.com/ratings/en/regulatory/article/-/view/sourceId/101694744). Reuters framed the sequence on August 4, 2026: Oracle as an outlier among top software firms, forced by limited cash reserves to borrow heavily and sign roughly $260 billion of data-centre leases at fiscal 2026 year-end — a disclosure since superseded by $288 billion at August 31, 2026 — with S&P the first agency to move to BBB- (https://www.reuters.com/business/oracle-corp-goes-high-stakes-ratings-gamble-ai-strategy-2026-08-04/).

The falsifiable test for this link is price, not rating: CreditSights observes Oracle trades in high-BB territory, roughly two notches below its mid-BBB ratings, with spreads primarily reflecting what it calls binary OpenAI risk (https://know.creditsights.com/insights/oracle-top-20-client-faqs/). The February bond's curve — 4.55% at three years against 6.85% at forty — already encodes doubt about the long end. A further downgrade, or a new issue pricing wider than the February curve at comparable maturities, would confirm that market access is narrowing.

Link two: customer concentration

S&P's OpenAI characterisation is the sharpest articulation of the concentration risk, but the mechanism predates it. Moody's named counterparty risk a primary driver of its December 2025 negative outlook (https://www.moodys.com/web/en/us/insights/credit-risk/outlooks/mco-15-december-2025.html), and CreditSights' framing of spreads as primarily binary OpenAI risk predates the July downgrade (https://know.creditsights.com/insights/oracle-top-20-client-faqs/). Oracle's own February financing plan names OpenAI among the largest OCI customers whose contracted demand the $45-50 billion raise supports (https://investor.oracle.com/investor-news/news-details/2026/Oracle-announces-Equity-and-Debt-Financing-Plan-for-Calendar-Year-2026/default.aspx).

What would falsify the concern? Observable conditions: OpenAI raising external capital on terms that fund its Oracle obligations, or Oracle's RPO composition shifting materially toward diversified counterparties. Management's statement that about half of RPO converts to sales within 36 months (https://www.fool.com/earnings/call-transcripts/2026/09/11/oracle-orcl-q1-2027-earnings-call-transcript/) is itself a testable claim — if the conversion rate lags, the contracted volume that S&P treats as undelivered exposure will not de-risk on schedule.

Link three: lease commencement timing

The lease queue grew and drifted simultaneously. At February 28, 2026 the 10-Q disclosed $261 billion of additional not-yet-commenced lease commitments, substantially all data-center arrangements, generally expected to commence between Q4 FY2026 and FY2028 with 15-19 year terms, alongside $130.9 billion of senior notes and other long-term borrowings outstanding (estimated fair value $118.4 billion), $43.0 billion of senior notes issued in the first nine months of FY2026, $11 billion of unconditional purchase obligations primarily for data-center power, a $2.2 billion guarantee of a lessor's borrowing maturing September 2026, and a March 6, 2026 replacement of the $6.0 billion revolver with a new $10.0 billion five-year unsecured revolving facility priced off Oracle's long-term senior unsecured credit rating (https://www.sec.gov/Archives/edgar/data/1341439/000119312526101045/orcl-20260228.htm).

By August 31, 2026, the same disclosure class had grown to $288 billion and the expected commencement window had shifted rightward to Q2 FY2027 through FY2029 (https://www.ibtimes.sg/oracle-stock-falls-4-ai-data-center-faces-18b-debt-cost-crisis-94226). A growing queue with a later window is the signature of capacity arriving later than contracted — which pushes the point at which projects reach the roughly 100% post-ramp free-cash-flow conversion management describes further into the future.

The September 24, 2026 force majeure report makes the timing link concrete. Bloomberg News, via Reuters, reported that Oracle sent a force majeure notice to developer Blue Owl on the New Mexico data-center campus known as Project Jupiter, aiming to delay payments if the site fails to come online in 2028 rather than exit as main tenant; Oracle and Blue Owl did not immediately comment to Reuters; Oracle separately said the project remains on its planned schedule and that it is fully committed to New Mexico; Oracle shares fell about 4% in premarket trading (https://www.reuters.com/business/oracle-cites-force-majeure-shield-itself-controversial-data-center-bloomberg-2026-09-24/). This is an interim, precautionary contractual move — not a merits outcome — and Reuters could not independently verify the notice. Per a source cited by Channel News Asia, securing power for the New Mexico site is Oracle's responsibility under the contract; Oracle reportedly cannot terminate the lease under any circumstances; Blue Owl holds about $3 billion of equity in the project with Oracle responsible for paying the debt costs; the 1,400-acre campus secured $18 billion of loans from a consortium of about 20 banks and forms part of Oracle's agreement with OpenAI; Blue Owl earns a 9% development-stage equity yield rising to about 11% levered at completion; and Blue Owl said the notice does not change financial commitments. Bloomberg Law adds that the project debt was trading below 90 cents on the dollar and notes it is not certain force majeure would free Oracle from previously agreed financial obligations (https://www.channelnewsasia.com/business/oracle-triggers-force-majeure-data-center-project-over-power-delays-source-says-6408541; https://news.bloomberglaw.com/artificial-intelligence/oracle-cites-force-majeure-to-shield-itself-on-big-data-center).

Read together: the lease structure concentrates both power-delivery risk and debt-service obligation on Oracle while the yield accrues to the developer — and the first public signal of distress management of that structure arrived as a force-majeure notice whose legal effect is itself uncertain. The observable confirmation conditions for this link are whether Project Jupiter commences on the 2028 schedule, and whether any further lease in the $288 billion queue slips its commencement window.

What the off-balance-sheet queue hides and what it does not

The $288 billion is a commitment, not a liability — yet. On-balance-sheet lease liabilities at August 31, 2026 were roughly $34.6 billion operating and about $9.2 billion finance per one filing summary, with another secondary summary reporting $9.85 billion finance leases and future undiscounted payments of $48.050 billion operating and $15.195 billion finance; the two sources disagree on the finance-lease figure, so the range should be treated as unverified against the filing (https://rundownsai.com/oracle-force-majeure-288b-data-center-leases/; https://digidai.github.io/2026/09/13/oracle-stock-sale-ai-restructuring-plan/). The filing does not attribute the entire $288 billion to AI workloads (https://rundownsai.com/oracle-force-majeure-288b-data-center-leases/).

The economically relevant distinction is commencement. Lease commitments enter the balance sheet when leases commence, converting a disclosure into fixed obligations with 15-19 year terms. The queue's size relative to the $125 billion of notes payable and other borrowings — roughly 2.3x — means the commencing portion in any quarter materially determines leverage. Fitch's EBITDAR-above-4x forecast and CreditSights' net lease-adjusted leverage of 3.7x by FY2028 both model exactly this conversion (https://www.fitchratings.com/research/corporate-finance/fitch-rates-oracle-senior-notes-offering-bbb-affirms-idr-outlook-stable-02-02-2026; https://know.creditsights.com/insights/oracle-top-20-client-faqs/).

Management's own framing, stated precisely

Oracle's investor-facing arithmetic deserves restating without the promotional gloss. The company said the new contract structures have no incremental impact on its capital-raising plans (https://investor.oracle.com/investor-news/news-details/2026/Oracle-Announces-Q1-Results-Driven-by-Triple-Digit-Growth-in-Cloud-Infrastructure-Revenues/default.aspx) — true as stated, because the $45-50 billion plan was sized for demand already contracted. It also guided FY2027 capex of $90-95 billion with no more than $70 billion net cash (https://www.fool.com/earnings/call-transcripts/2026/09/11/oracle-orcl-q1-2027-earnings-call-transcript/), which concedes that $20-25 billion of the year's spend is someone else's cash — customer prepayments — arriving with financing components attached (https://digidai.github.io/2026/09/13/oracle-stock-sale-ai-restructuring-plan/). Customer-funded buildout is not costless capacity; it is capacity financed by obligations to deliver future service at contracted prices. Whether those prices leave adequate returns depends on the conversion economics management asserts — roughly 100% free-cash-flow conversion to post-tax EBITDA shortly after ramp — which remains a company assertion without a stated date for aggregate positive free cash flow (https://www.fool.com/earnings/call-transcripts/2026/09/11/oracle-orcl-q1-2027-earnings-call-transcript/).

The market's verdict so far is measurable: Oracle shares had dropped 22% year-to-date as of the September 10, 2026 close while the S&P 500 gained roughly 11% (https://www.cnbc.com/2026/09/10/oracle-orcl-q1-earnings-report-2027.html). The company nonetheless raised FY2027 guidance to at least $90 billion of revenue and $8.10 non-GAAP EPS, guided Q2 FY2027 to 30-34% revenue growth and $1.85-$1.93 non-GAAP EPS, scheduled an investor day for October 2026, and set Q2 results for December 14, 2026 (https://investor.oracle.com/investor-news/news-details/2026/Oracle-Announces-Q1-Results-Driven-by-Triple-Digit-Growth-in-Cloud-Infrastructure-Revenues/default.aspx; https://www.fool.com/earnings/call-transcripts/2026/09/11/oracle-orcl-q1-2027-earnings-call-transcript/).

The chain, stated as conditions

The mechanism from announced capex to OCI cash flow runs through five links, each now carrying an observable stress signal:

  1. Demand contracts exist — $664 billion of RPO, verified (https://investor.oracle.com/investor-news/news-details/2026/Oracle-Announces-Q1-Results-Driven-by-Triple-Digit-Growth-in-Cloud-Infrastructure-Revenues/default.aspx).
  2. Capacity must commence on schedule — the $288 billion queue's window has already drifted rightward once (https://www.ibtimes.sg/oracle-stock-falls-4-ai-data-center-faces-18b-debt-cost-crisis-94226), and the first public timing dispute arrived as the September 24 force majeure report (https://www.reuters.com/business/oracle-cites-force-majeure-shield-itself-controversial-data-center-bloomberg-2026-09-24/).
  3. Commencement requires power and construction delivered on Oracle's side of private lease structures — Project Jupiter places power responsibility on Oracle contractually, per the CNA-sourced account (https://www.channelnewsasia.com/business/oracle-triggers-force-majeure-data-center-project-over-power-delays-source-says-6408541).
  4. Funding requires investment-grade market access — now one notch above the high-yield boundary at S&P, with Moody's negative (https://www.heise.de/en/news/S-P-downgrades-Oracle-to-BBB-only-one-notch-above-junk-level-11363472.html; https://www.newsquawk.com/headlines/moody-s-affirms-oracle-s-orcl-baa2-rating-on-equity-announcement-outlook-remains-negative-02-02-2026).
  5. Cash conversion requires prepayments to continue and customer credit to hold — S&P calls OpenAI a central credit risk with roughly half of undelivered contracted volume (https://www.heise.de/en/news/S-P-downgrades-Oracle-to-BBB-only-one-notch-above-junk-level-11363472.html).

The falsification conditions are specific: a further one-notch downgrade, a new bond pricing materially wider than the February curve, a missed lease commencement in the queue, or an OpenAI funding failure. Any one of them would confirm a break at its link. Absent them, the thesis is that Oracle's buildout is a leveraged bet on conversion speed — one whose financing is now effective and whose returns are not yet.

The next dated checkpoints are the October 2026 investor day and the December 14, 2026 Q2 FY2027 report, where management must either show the lease queue commencing on schedule or explain the drift (https://www.fool.com/earnings/call-transcripts/2026/09/11/oracle-orcl-q1-2027-earnings-call-transcript/).

Boundaries of this analysis

Several figures in this article come from secondary summaries of filings rather than filings read in full: the $288 billion lease-queue figure and $125 billion notes payable derive from filing-based secondary reporting (https://www.ibtimes.sg/oracle-stock-falls-4-ai-data-center-faces-18b-debt-cost-crisis-94226; https://last10k.com/sec-filings/orcl/0001193125-26-389274.htm); the finance-lease split is inconsistent between two such summaries (https://rundownsai.com/oracle-force-majeure-288b-data-center-leases/; https://digidai.github.io/2026/09/13/oracle-stock-sale-ai-restructuring-plan/); the S&P outlook after July 9 rests on contemporaneous accounts because S&P's own page was not accessible (https://cbonds.com/news/4003647/; https://www.spglobal.com/ratings/en/regulatory/article/-/view/sourceId/101694744); and the Project Jupiter contractual terms come from an unnamed source and unpublished private documents (https://www.channelnewsasia.com/business/oracle-triggers-force-majeure-data-center-project-over-power-delays-source-says-6408541; https://news.bloomberglaw.com/artificial-intelligence/oracle-cites-force-majeure-to-shield-itself-on-big-data-center). These are marked as attributed rather than verified wherever they appear. The force majeure notice is an interim contractual step reported by Bloomberg News and not independently verified by Reuters; nothing here treats it as a decided outcome.