Oracle is a very profitable old software company. It has bolted on a huge, fast-growing and borrowed-money business renting out AI computers, mostly to OpenAI. At $137 the stock is priced about fairly for that mix. The upside if the AI contracts pay off is large. So is the downside if OpenAI or the bond market stumbles. That makes it a watch-and-ladder name, not a buy-now name.
| Item | Verdict |
|---|---|
| Score | 6.0 / 10 (first version, so no prior score). Demand, growth and paper profits are strong. The balance sheet, cash flow, dependence on one customer and the credit market all argue against. |
| Thesis status | ON WATCH. No thesis breaker has fired yet: the order book is still rising, the credit rating is still investment grade, and the OpenAI contract is intact. But on 9/24 the New Mexico "Project Jupiter" site went under force majeure (a delay clause for events outside Oracle's control). Default insurance hit a reported record on 9/25 UNVERIFIED single source. The rating is one notch above junk. |
| Tier | Conviction-spec, NOT Tier-1. The framework allows Tier-1 sizing only with a moat of 9/10 or better. Oracle's blended moat is 6/10: the database scores 8.5–9, the AI rental business 3.5–4 (§9). |
| Macro stamp (Stage 1) | YELLOW. The veto does not fire, but no full-size position. Oracle is not an "unprofitable small cap," so the hard rule doesn't trigger. It does burn cash after its building spend and needs about $40B a year of outside money. That is happening while the Fed is raising rates (3.75–4.00%, hiked 9/16/26) and the 10-year Treasury yield is at 5.18%, the highest since 2007. Those are the conditions the rule exists to protect against. |
| Sizing rule | Framework Stage 10: 10% or less per position is for Tier-1 only (moat 9 or better). Conviction-spec gets 5% or less of the portfolio at full ladder. That is the portfolio-manager (PM) house rule used for IREN, half the Tier-1 cap; the framework sets only the Tier-1 number. With macro YELLOW and 0 of 5 "flip to green" conditions met (§1b), no stock purchases at spot. Only Zone 2–3 cash-secured puts (§15), at starter size: 25% of the 5% cap, or 1.25% of the portfolio. Size everything to survive a 50% fall. |
| Next check | Q2 FY27 earnings, about 2026-12-10 after the close UNVERIFIED: yfinance calendar, not confirmed by Oracle investor relations. Before that: Analyst Day, 2026-10-28 UNVERIFIED: single-leg mention; the Fed meeting 10/27–28; Q3 13F institutional-holdings filings around 11/14. |
Oracle has $664B of signed contracts not yet billed. About half is OpenAI, roughly $300B by press reports; Oracle never names the figure. Oracle is racing to turn those contracts into revenue before its ability to borrow runs out. The old database-and-apps business, roughly $55B a year of sales, throws off the cash and the trust. The new AI rental business (OCI, Oracle Cloud Infrastructure) is growing 121% a year but earns about 34 cents of gross profit per dollar versus about 60 cents on the old mix. It is being built with $125B of bonds, $20B of new stock sold this summer, and $288B of data-center rent signed but not yet started, which appears nowhere on the balance sheet.
If OpenAI keeps raising money, Jupiter-style delays stay isolated and the rating holds, fiscal 2028 sales reach about $135B and the stock is worth about $241. If builds slip a year, it is worth about $152. If OpenAI stretches its commitments, about $73; the tail is much lower. The PM's probability-weighted value is about $146, only +6.6% from $137. The Red Team's is $129. No edge at spot. The edge starts in the $116–124 zone.
It is pricing a credit event, not a demand collapse.
It trades every OpenAI headline as if it were the whole order book. Meanwhile the non-OpenAI half of the backlog, around $330–364B EST, more than doubled in a year.
The economy is still growing fast, but the Fed is raising rates to fight inflation, and money is getting tighter. Oracle needs to borrow or sell stock every year to fund its build, so tighter money hits it harder than companies that pay their own way.
| Signal | Level (date) | Read |
|---|---|---|
| Fed funds rate (the Fed's policy rate) | 3.75–4.00%, raised 0.25 points on 9/16/26, 12–0 vote | Headwind |
| Odds of another hike Oct 27–28 | ~70–73% (CME via CNBC, 9/23) | Headwind |
| Fed's own rate projections (the "dot plot") | 4.00–4.25% at end-2026 and end-2027, so no cuts | Headwind |
| Treasury yields, 3-month / 5-year / 10-year / 30-year | 4.07 / 5.01 / 5.18 / 5.50% (9/25) | 10-year highest since 2007 |
| Consumer inflation (CPI), headline / core | 3.4% / 2.4% (Aug) | Oil shock keeps the Fed hawkish (inclined to keep raising rates) |
| Atlanta Fed live GDP estimate (GDPNow), Q3 | 5.0% (9/25) | Passes the framework's 2%+ test |
| ISM surveys (above 50 = growing), manufacturing / services | 54.6 / 55.4; price gauges 71.1 / 72.6 (Aug) | Late expansion |
| Bank reserves at the Fed | $2,930B, down $84B in a week (9/23) | Liquidity draining |
| Global liquidity (Michael Howell) | ~$196T; cycle peak put at Q4'25–Q1'26 | Past peak |
| Stock fear gauge (VIX) / bond fear gauge (MOVE) | 14.87 / ~104 | Stocks calm, bonds stressed. Bonds usually lead |
| Junk-bond extra yield over Treasuries (high-yield spread) | ~2.80 points, widening off the lows | Tight but turning |
This is the first full ORCL report, so the "prior version" is the 5-minute triage screen run the same morning. The deep dive corrected several of that screen's numbers, some worse and some better. It also logged the events of the last three months that moved the stock.
No earlier deep dive or tracker exists for ORCL. The baseline is the triage of 2026-09-26: 8.5/12, "worth the deep dive."
Struck-through red = the triage number the deep dive corrected. Green = held up.
| Triage claim | Deep-dive finding | Verdict |
|---|---|---|
| Cash after building (free cash flow) over the last 12 months was −$45.9B | −$28.7B reported ($46.94B operating cash minus $75.66B capex, the spending on buildings and equipment). −$46.2B only if customer prepayments are also removed. | MISSED overstated the burn |
| Revenue growth 29.6% | That is the latest quarter only. The full 12 months is +21.7% ($71.78B vs $58.99B). | MISSED overstated |
| Rule of 40 = −34 | −18.3 (21.7% growth + −40.0% cash margin). Still FAIL. | MISSED in size, KEPT as FAIL |
| Shares +7.0% a year | +6.4% at quarter end (3,024M vs 2,842M); +3.1% on average diluted shares. The driver is 141M new shares sold through the ATM (at-the-market) program. | Partly KEPT |
| Margins PASS | Downgraded to PARTIAL. Gross margin fell from 70.9% to 60.0% in 8 quarters. | MISSED |
| Founder-led PARTIAL | PASS. Ellison is Executive Chair and CTO (chief technology officer), per the proxy dated 9/25/26. | Upgraded |
| Skin in the game about 12,000x | 1,212x on FY26 total pay of $131.0M, which includes a $117.8M option grant. The triage used $13.2M of pay. | MISSED in size, KEPT as PASS |
| Contrarian PASS | PARTIAL. Contested, not capitulated (§11). | Downgraded |
| "$169B debt vs $37B cash" | Reconciles: $125.3B of bonds and loans plus $43.8B of leases = $169.1B. | KEPT definition clarified |
| Forward P/E 12.5 (share price ÷ expected earnings per share) | That uses FY28 earnings of $11.00. On FY27 guidance of $8.10 it is 16.9x. | KEPT with a caveat |
| "OpenAI concentration is the make-or-break question" | Confirmed. See §6. | KEPT |
| Date | Event | Why it matters |
|---|---|---|
| 2026-09-25 | Default insurance (5-year credit default swap) reported at a record 227 basis points, up 16% in a week UNVERIFIED: GuruFocus only. Proxy shows Ellison's pledged shares up from 346M to 413M (+67M, worth $9.2B). | Credit is getting worse while the stock sits 19% above its July low. |
| 2026-09-24 | Force majeure invoked on Project Jupiter (New Mexico, up to $165B, 2.45–2.5GW, developed by Blue Owl's Stack). The gas pipeline slipped from Aug 2026 to Feb 2027 after permit denials. Jupiter's bank loans are quoted at 89–91 cents on the dollar. Stock fell as much as 7% intraday. | The first delivery slip on a flagship site. Tests the "partners fund it" model. |
| 2026-09-21 | Rothschild Redburn reiterates the only Sell on the Street, target $110. | This bear has been directionally right for 12 months. |
| 2026-09-16 | Fed hikes 0.25 points; the 10-year crosses 5%. | Borrowing costs rise for a company that must borrow. |
| 2026-09-12 | Altman (Fortune): OpenAI IPO pushed to 2027. The stock fell five straight days from 9/9 to 9/15, −13.6%. | The market trades OpenAI headlines as Oracle headlines. |
| 2026-09-10 | Q1 FY27: revenue $19.35B (+30%), OCI +121%, backlog $664B, adjusted EPS (earnings per share) $1.92 vs $1.74 expected. The $20B stock sale is complete. The stock gapped up to $164–166, then closed $150.28 on 80M shares. | The earnings beat was sold, which is a bearish tell. |
| 2026-07-24 / 07-28 | 52-week closing low $114.99 (7/24); intraday low $114.50 (7/28). | This is the floor the trade plan is built on. |
| 2026-07-09 | S&P cuts Oracle to BBB- (one notch above junk), stable outlook. It names OpenAI concentration a "central credit risk" and forecasts FY27 cash after building of about −$42B. RT-FIX the Qual leg said 7/20; S&P via heise says 7/9. | Funding cost went up, and the room for error shrank. |
| 2026-06-26 | Worst week for the stock since 2001, on financing worries (CNBC). Stock at $148. | |
| 2026-06-10/11 | Q4 FY26: FY26 capex (building spend) $55.7B vs a ~$50B guide; FY26 cash after building −$23.7B; FY27 plan of about $40B of new debt and equity including a $20B stock sale. Stock −11%. | This began the second crash. |
Oracle sells the database software that banks, hospitals and governments run their most important records on. It also rents business apps by subscription. Its new and fastest-growing business is building giant warehouses of AI chips and renting them to AI labs like OpenAI.
Oracle Corporation was founded in 1977 by Larry Ellison and is headquartered in Austin, Texas. It has three businesses:
| Segment | Revenue | Growth vs a year ago | Share of the $19.3B total |
|---|---|---|---|
| Cloud infrastructure (OCI: renting computers and AI chips) | $7.39B | +121% | 38% |
| Cloud apps (SaaS: subscription business software) | $4.22B | +10% | 22% |
| Software license + support (old database and app fees; license $655M, support $4.9B) | $5.55B | −3% (license −15%, support −1%) | 28% |
| Services (consulting and implementation) | $1.4B | +5% | 7% |
| Hardware (servers and storage from the Sun acquisition) | $0.774B | +15% | 4% |
Oracle makes the giant filing cabinets that banks, hospitals and governments keep their most important records in. Once your records are inside, moving them out is scary and expensive. So customers keep paying Oracle every year to keep everything running. That's the old business, and it makes a lot of money.
The new business is building enormous warehouses full of AI computer chips and renting them out, mostly to companies that make AI chatbots like ChatGPT. Oracle borrows money to build the warehouses first. It gets paid back slowly, over about five years, while customers use them. The risk: Oracle signed rent on the warehouses for 15–19 years, but the customers only signed up for about 5. And the biggest customer is still losing money.
| Date | Event |
|---|---|
| 1977-06 | Founded by Larry Ellison, who was CEO until Sep 2014. |
| 1986 | IPO (initial public offering, first listing on the stock market). |
| 2005 / 2010 / 2016 / 2022 | Acquisitions: PeopleSoft (HR apps), Sun Microsystems (hardware), NetSuite (small-business apps), Cerner (health records). |
| 2014 | Clay Magouyrk joins from Amazon Web Services and builds OCI "Gen2." |
| 2014-09 | Ellison becomes Executive Chair + CTO; Hurd and Catz become co-CEOs. |
| 2019 | Catz becomes sole CEO after Hurd's death. |
| 2025-04-07 | Swing low $118.86 before the AI rally. |
| 2025-09-09/10 | Q1 FY26: backlog $455B (from $138B one quarter earlier). The stock rose 36% the next day to a $324.63 closing high; the intraday all-time high was $345.72. |
| 2025-09 | About $300B, 5-year OpenAI contract reported (WSJ/TechCrunch). Oracle has never named the amount in a filing. |
| 2025-09-22 | Magouyrk and Sicilia named co-CEOs; Catz moves to Executive Vice Chair. |
| 2025-10-16 | Analyst day FY30 targets: $225B revenue, $166B OCI, $21 adjusted EPS, 30–40% gross margin on AI infrastructure. |
| 2025-12-10 | Q2 FY26: backlog $523B; FY26 capex raised to ~$50B from $35B. Stock −11% after hours (from $220.9 to $188.2 over the next days). |
| 2025-12-17 | Blue Owl declines to fund the $10B Michigan data center (CNBC). |
| 2026-01-07 | "Death cross": the 50-day average falls below the 200-day. |
| 2026-02-02/05 | $45–50B calendar-2026 financing plan: $43.0B of bonds in FY26, $5.0B mandatory convertible preferred, and a $20B at-the-market (ATM) stock program. |
| 2026-03 | Abilene (Texas, OpenAI's flagship site) expansion beyond the initial build dropped UNVERIFIED: Bloomberg via IntuitionLabs. |
| 2026-03-10 | Q3 FY26: backlog $553B. "No more bonds in calendar 2026." |
| 2026-03-31 | Largest layoff in company history. Headcount ~162k to 141k by 5/31. |
| 2026-04-06 | Hilary Maxson (ex-Schneider Electric CFO) becomes CFO (chief financial officer). |
| 2026-04-28 | WSJ: OpenAI missed its revenue targets. Stock $172 to $163. |
| 2026-05 | Relief rally: +39.9% for the month, to $225.00 on 5/29. |
| 2026-06-10 | Q4 FY26: backlog $638B; FY27 capex $90–95B; $40B financing plan. Stock −11%. June −35%. |
| 2026-07-09 | S&P cuts to BBB- RT-FIX: not 7/20. |
| 2026-07-24 | 52-week closing low $114.99. |
| 2026-09-10 | Q1 FY27: backlog $664B; ATM sold out ($19.9B, 141M shares). |
| 2026-09-24 | Jupiter force majeure. |
| 2026-09-25 | Proxy: 413M pledged shares. Close $137.10. |
| 2026-10-28 | Analyst Day (new FY30 targets expected) UNVERIFIED. |
| 2026-11-18 | Annual meeting. |
| ~2026-12-10 | Q2 FY27 results UNVERIFIED. |
| 2029-01-15 | Mandatory preferred converts into 25.0–31.2M common shares. |
TAM (total addressable market) means the total yearly spending Oracle could compete for.
25% a year through 2029, 20% in 2030, then about 10% a year to 2036. That gives about $1.12T in 2029, $1.48T in 2031 and $2.39T in 2036.
The Qual leg's text said "20%, 20%, 15%" after 2029. That path would give $1.61T and $2.7T, which don't match its own table. The table's path is used.
Bear 4%, base 7%, bull 10%.
SaaS about $17B a year growing ~10%, plus about $31B of license, support, services and hardware, roughly flat. That totals about $53B in year 3, $57B in year 5 and $65B in year 10 EST.
| Horizon | OCI: bear / base / bull | Total revenue: base (bull) |
|---|---|---|
| 3-year (~FY30) | $45B / $78B / $112B | ~$131B ($165B) |
| 5-year (~FY32) | $59B / $104B / $148B | ~$161B ($205B) |
| 10-year (~FY37) | $96B / $167B / $239B | ~$232B ($304B) |
A flywheel is a loop where each step feeds the next and the business gets stronger as it spins. Oracle's loop works on the sales side. It does not yet work on the cash side, because the cash comes from lenders and prepaying customers, not from the business itself.
Ring colors: green = spinning, yellow = partially spinning, orange = unproven, red dashed = broken, grey = not graded in the report.
| Link | Status | Evidence |
|---|---|---|
| 1 to 2 | SPINNING | Multicloud database revenue +353%; 150M AI instances embedded in the apps; 70 multicloud regions. |
| 2 to 3 | SPINNING | Backlog $664B (+$209B a year); $30B+ of new AI contracts booked in Q1. |
| 3 to 4 | PARTIALLY SPINNING, stressed | $11.4B of financing-type prepayments in Q1, and $75B of the Q4 backlog was prepaid or customer-chip deals. But Oracle still needed the $20B stock sale, and Jupiter's partner financing stalled. |
| 4 to 5 | UNPROVEN | Renewals repriced +20% and utilization is 97.9% (both unaudited management claims), but gross margin fell from 70.9% to 60.0%. |
| 5 to 6 | BROKEN TODAY | Free cash flow (cash after building spend) negative in 7 of 8 quarters. |
Verdict: half-spinning. The revenue half spins; the cash half does not. By the framework, it isn't a long-term-hold flywheel until free cash flow turns positive. Management implies FY28–29 ("reasonably quick timeframe" after the ramp). No date has been guided, and none is verified.
Seven questions decide this stock. The first two, whether OpenAI can pay and whether Oracle can keep borrowing, matter more than everything else combined. Each question shows today's answer and what would change it.
| Question | Current answer | What would change it |
|---|---|---|
| 1. Can OpenAI pay for roughly half the backlog? | Unknown, and trending worse. S&P: OpenAI is "about half" of $638B. Press puts the contract at ~$300B over 5 years from 2027. OpenAI projects about $278B of cumulative cash burn in 2026–30, cut its compute plans from $1.4T to ~$600B, reportedly missed revenue targets (4/28), and moved its IPO to 2027. Its revenue is reported anywhere from $40B to $65B a year CONFLICT: both press; unverifiable. | Better: a $1T+ round or IPO filing; a public, prepaid payment schedule. Worse: contract "re-phasing," a failed round, missed payments. |
| 2. Can Oracle fund the build without falling to junk or a dilution spiral? | Barely. BBB- at S&P (7/9/26); Moody's Baa2 negative (affirmed 2/2/26, September status UNVERIFIED). About $20B of the $40B FY27 plan is still undone. Remaining FY27 funding cushion is $0–12B before new prepayments or debt RT-FIX. | Better: a non-August quarter with operating cash of $12B+ before prepayments; default insurance below 150 basis points. Worse: a new equity program, a junk cut, default insurance above 300 basis points. |
| 3. What does the AI rental business really earn after depreciation and rent? | Last disclosed 32% gross margin (Q3 FY26). The Quant leg's segment math says ~34% on extra revenue, but that may exclude depreciation on the $48.5B still under construction RT-FIX: snapshot, not steady state. Leaked documents said 14–16% on average (The Information via DCD). | Better: 30%+ disclosed after full depreciation, two quarters running. Worse: disclosure below 20%, or gross margin below 55%. |
| 4. Are the $288B of unstarted leases matched to customer contracts? | No evidence they are. Leases run 15–19 years; AI contracts about 5. Jupiter shows Oracle can delay payments, but by how much is unknown. | Better: disclosure of termination, delay or power pass-through terms. Worse: lease commitments growing faster than the backlog (now 43% of it). |
| 5. Is non-OpenAI demand durable? | Yes so far. The backlog excluding OpenAI is roughly $330–364B EST, up from $138B of total backlog in May 2025. Meta and NVIDIA are named customers. But backlog growth slowed to +$26B last quarter (from +$85B), and the book-to-bill ratio (new orders ÷ sales) fell to ~2.3x. | Better: book-to-bill back above 3x with more than half from non-OpenAI customers. Worse: backlog falls quarter over quarter. |
| 6. Is the database moat holding? | Slowly eroding. Support −1%, license −15%. Gartner ranks Oracle #3 in databases behind AWS and Microsoft. Multicloud database +353% extends the moat's reach. | Worse: support revenue falling more than 3% a year. |
| 7. Governance: pledged shares and succession. | 413M shares pledged (13.7% of the company); a $40.4B personal guarantee on the Paramount/Warner Bros deal. Ellison is 82; the lead independent director is 81. | Worse: any disclosed forced sale, or a pledge increase on a further price fall. |
These are the 16 pass/fail tests every stock in the framework has to face. Oracle passes 5, is borderline on 5 and fails 6. The fails cluster around cash and debt: the cost of building ahead of demand. The passes are about the founder, profits and the story being understandable.
Where the fails cluster: five of the six (free cash flow, dilution, current ratio, cash vs debt, and the moat on the AI half) are the cost of building three years ahead of revenue. The sixth, the beat record, is improving.
Sales and paper profits are growing fast. Cash is not. Oracle spends far more on data centers than its business brings in, and fills the gap with borrowed money, new stock and customers paying early. The balance sheet fails both of the framework's safety tests.
The next-12-month slice is about the size of the FY27 revenue guide ($90B). FY28's $131B consensus depends on far-off revenue arriving on time.
| Quarter | Capex (building spend) $B | Operating cash $B | OCI revenue $B | Backlog $B |
|---|---|---|---|---|
| Q2 FY25 | 3.97 | 1.30 | ~2.4 | ~97 |
| Q3 FY25 | 5.86 | 5.93 | 2.7 | 130 |
| Q4 FY25 | 9.08 | 6.16 | 3.0 | 138 |
| Q1 FY26 | 8.50 | 8.14 | 3.35 | 455 |
| Q2 FY26 | 12.03 | 2.07 | 4.1 | 523 |
| Q3 FY26 | 18.64 | 7.15 | 4.9 | 553 |
| Q4 FY26 | 16.49 | 14.62 | 5.8 | 638 |
| Q1 FY27 | 28.50 | 23.10 | 7.39 | 664 |
| Item | Value | Source |
|---|---|---|
| Price | $137.10 | close 9/25/26, yfinance |
| Shares outstanding | 3,023,736,000 | 10-Q cover, 9/7/26 |
| Market value of the stock | $414.6B (137.10 × 3.0237B; matches yfinance, so no split or data error) | calc |
| Cash + marketable securities | $36.37B + $0.71B = $37.08B (plus $2.6B restricted cash held in prepaid expenses) | 10-Q |
| Bonds and loans | $125.34B ($7.625B due within a year + $117.712B later). Market value of the notes $105.7B. | 10-Q |
| Leases on the balance sheet | Operating $34.62B + finance $9.19B = $43.81B | 10-Q Note 6 |
| Series D mandatory convertible preferred | $5.0B (carried at $4.954B) | 10-K |
| Enterprise value (stock value + debt − cash), excluding leases | $507.8B | calc |
| Enterprise value including leases | $551.6B (matches yfinance $552B) | calc |
| Off-balance-sheet leases signed but not started | $288B (was $248B at 11/30/25 and $260B at 5/31/26). Start Q2 FY27 to FY29; 15–19 year terms. Not in enterprise value. RT-FIX: the Macro leg's $260B is stale | 10-Q |
| Unconditional purchase obligations (committed buys of chips and equipment) | $34.15B (was $13.3B at 5/31/26) | 10-Q, 10-K |
| Short interest (shares borrowed and sold, betting on a fall) | 50.84M = 2.73% of float (yfinance, 9/15) CONFLICT: stockanalysis 44.6M = 2.40% | |
| Net debt (debt − cash) | Bonds and loans only: $88.3B. With leases: $132.1B. With leases + preferred: $137.1B. | calc |
CONFLICT resolved Sentiment cited net debt of "$88B (TradingKey) to $132B (TIKR)." Both are right on different definitions: $88.3B without leases, $132.1B with them. This report uses the $132–137B figure for valuation.
The leading indicators that matter for this kind of business are the backlog (RPO, remaining performance obligations: signed contracts not yet billed) and cloud-infrastructure (OCI) revenue. Scroll sideways on a phone; the first column stays put.
| Line item | Q2 FY25 Nov 2024 | Q3 FY25 Feb 2025 | Q4 FY25 May 2025 | Q1 FY26 Aug 2025 | Q2 FY26 Nov 2025 | Q3 FY26 Feb 2026 | Q4 FY26 May 2026 | Q1 FY27 Aug 2026 |
|---|---|---|---|---|---|---|---|---|
| Total revenue | 14,059 | 14,130 | 15,903 | 14,926 | 16,058 | 17,190 | 19,184 | 19,345 |
| Revenue growth vs a year ago | 8.6% | 6.4% | 11.3% | 12.2% | 14.2% | 21.7% | 20.6% | 29.6% |
| OCI revenue ($B) | ~2.4 EST | 2.7 | 3.0 | 3.347 | 4.1 | 4.9 | 5.8 | 7.388 |
| OCI growth | n/a | 49% | 52% | 55% | 68% | 84% | 93% | 121% |
| SaaS (cloud apps) revenue ($B) | ~3.5 EST | ~3.5 EST | ~3.7 EST | 3.839 | 3.9 | 4.0 | 4.1 | 4.219 |
| Software license + support | n/a | n/a | n/a | 5,721 | n/a | ~6,100 | n/a | 5,550 |
| Backlog (RPO), $B | ~97 EST | 130 | 138 | 455 | 523 | 553 | 638 | 664 |
| Book-to-bill ((backlog change + revenue) ÷ revenue) | ~0.9x | ~3.3x | ~1.5x | ~22x | ~5.2x | ~2.7x | ~5.4x | ~2.3x |
| Gross margin | 70.9% | 70.3% | 70.2% | 67.3% | 66.5% | 64.6% | 65.2% | 60.0% |
| Operating margin (stockanalysis basis) | 30.6% | 31.3% | 32.6% | 31.3% | 32.0% | 32.7% | 36.2% | 35.6% |
| GAAP operating income (company) | n/a | n/a | n/a | 4,277 (29%) | ~4,700 (29%) | ~5,500 (32%) | ~6,100 (32%) | 6,728 (35%) |
| Non-GAAP operating margin (company "adjusted") | n/a | n/a | n/a | 42% | 42% | 43% | 45% | 42% |
| Adjusted EBITDA (earnings before interest, tax, depreciation, amortization; yfinance) | n/a | n/a | n/a | 6,619 | 7,471 | 8,393 | 10,121 | 10,450 |
| GAAP net income | 3,151 | 2,936 | 3,427 | 2,927 | 6,135* | 3,721 | 4,304 | 4,760 |
| Non-GAAP EPS, actual / estimate | 1.47/1.48 | 1.47/1.49 | 1.70/1.64 | 1.47/1.48 | 2.26*/1.64 | 1.79/1.69 | 2.11/1.96 | 1.92/1.74 |
| Operating cash flow (OCF) | 1,304 | 5,933 | 6,157 | 8,140 | 2,066 | 7,151 | 14,620 | 23,103 |
| Capex (building and chip spend) | 3,970 | 5,862 | 9,080 | 8,502 | 12,033 | 18,635 | 16,493 | 28,499 |
| Free cash flow (FCF = OCF − capex) | −2,666 | +71 | −2,923 | −362 | −9,967 | −11,484 | −1,873 | −5,396 |
| Customer prepayment build (change in unearned revenue) | −1,851 | −419 | +119 | +2,550 | −2,083 | −217 | +4,392 | +15,360 |
| Diluted shares (M) | ~2,865 EST | ~2,878 EST | ~2,880 EST | 2,909 | 2,922 | 2,912 | 2,915 | 3,000 |
| Shares at quarter end (M) | 2,797 | 2,804 | 2,809 | 2,842 | 2,873 | 2,876 | 2,880 | 3,024 |
*Q2 FY26 includes a $2.7B pre-tax gain from selling the Ampere stake. It is a one-off. Sources: stockanalysis quarterly statements; Oracle 8-K press releases for Q2 FY26 to Q1 FY27; 10-Q/10-K; yfinance earnings history. Values marked EST were derived from growth rates, not read from filings.
GAAP means US generally accepted accounting principles, the official rulebook. This walk shows how $18.9B of reported profit becomes −$28.7B of cash after building.
| Step | $B |
|---|---|
| GAAP net income | 18.92 (includes the ~$2.7B Ampere gain) |
| + Depreciation and amortization (the yearly wear-and-tear cost of equipment) | +9.83 (quarterly 1.82, 2.28, 2.54, 3.19; more than doubled in 4 quarters) |
| + Stock-based compensation (employee pay in shares) | +4.81 |
| + Change in working capital | +16.43, of which +17.45 is customer prepayments |
| + Other non-cash items | −3.06 (mostly reverses the Ampere gain; deferred tax) |
| = Operating cash flow | 46.94 |
| − Capex | −75.66 |
| = Free cash flow | −28.72 |
| Free cash flow without the prepayment build (stress test) | −46.2 |
Operating leverage means profits growing faster than sales once fixed costs are covered.
| Measure | Q2 FY25 | Q3 FY25 | Q4 FY25 | Q1 FY26 | Q2 FY26 | Q3 FY26 | Q4 FY26 | Q1 FY27 |
|---|---|---|---|---|---|---|---|---|
| Operating cash ÷ capex | 0.33 | 1.01 | 0.68 | 0.96 | 0.17 | 0.38 | 0.89 | 0.81 |
| Free cash flow margin | −19.0% | 0.5% | −18.4% | −2.4% | −62.1% | −66.8% | −9.8% | −27.9% |
| Free cash flow margin ÷ operating margin | −0.62 | 0.02 | −0.56 | −0.08 | −1.94 | −2.04 | −0.27 | −0.78 |
| Gate | Result | Worked example |
|---|---|---|
| Current ratio 2:1 or better (short-term assets ÷ bills due within a year) | FAIL: 1.17 (up from 0.75 in FY25) | $55.63B ÷ $47.51B = 1.17 |
| Cash greater than debt | FAIL | $37.1B vs $125.3B (+$43.8B leases) |
| Net debt ÷ EBITDA (operating cash profit), bonds and loans only | 2.4x | $88.3B ÷ $36.4B 12-month EBITDA = 2.4. Oracle would need 2.4 years of cash profit to pay off net debt. |
| Net debt ÷ EBITDA including leases | 3.6x | $132.1B ÷ $36.4B. S&P uses "mid-4x" on its own adjusted definition UNVERIFIED definition. |
| Gross debt including leases ÷ EBITDA | 4.6x (stockanalysis 4.48x) | $169.1B ÷ $36.4B |
| Interest coverage (operating income ÷ interest) | 4.7x | Q1 FY27: $6.728B ÷ $1.428B. Interest expense rose +55% from $0.923B. |
| Altman Z-score (a bankruptcy-risk score; below 1.8 is distress, above 3 is safe) | 1.93, "grey zone" (stockanalysis) | |
| Piotroski F-score (0–9 financial-health checklist) | 5 (middling) |
| Uses | $B |
|---|---|
| Remaining net capex after prepayments (cap of ~$70B minus $18.0B spent) | ≤52.0 |
| Dividends (~$1.5–1.6B a quarter) EST | ~4.7 |
| Remaining FY27 principal ($7.21B minus ~$4.2B repaid in Q1) EST | ~3.0 |
| Total uses | ~59.7 |
| Sources | $B |
|---|---|
| Cash + marketable securities (8/31/26) | 37.1 |
| Operating cash before prepayments, 3 quarters: $23–35B range RT-FIX | 23–35 |
| Total | 60–72 |
| Untapped backstops (not counted above) | $10B revolving credit line (undrawn at 5/31/26) + ~$8.5B of commercial-paper room |
No convertible bonds. The only equity-linked security is the 6.50% Series D Mandatory Convertible Preferred (ticker ORCL-PRD):
| Amount | Coupon (yearly interest rate) | Maturity |
|---|---|---|
| $0.5B | floating | Feb 2029 |
| $3.0B | 4.55% | Feb 2029 |
| $3.0B | 4.45% | Sep 2030 |
| $3.5B | 4.95% | Feb 2031 |
| $3.0B | 4.80% | Sep 2032 |
| $3.0B | 5.35% | May 2033 |
| $4.0B | 5.20% | Sep 2035 |
| $5.0B | 5.70% | Feb 2036 |
| $2.5B | 5.875% | Sep 2045 |
| $2.25B | 6.55% | Feb 2046 |
| $3.5B | 5.95% | Sep 2055 |
| $5.0B | 6.70% | Feb 2056 |
| $2.0B | 6.10% | Sep 2065 |
| $2.75B | 6.85% | Feb 2066 |
The full ~50-tranche table is in 10-K Note 6.
Fallen-angel check RT-FIX: bond indexes use the middle of the three agency ratings. Moody's Baa2 is two notches above junk, so an S&P cut alone does not make Oracle a "fallen angel" (an issuer cut from investment grade to junk, which forces some funds to sell). The Macro leg's "largest fallen angel on record" is overstated and unsourced.
| Quarter end | Shares (millions) | Change | Driver |
|---|---|---|---|
| Nov-24 | 2,797 | ||
| Feb-25 | 2,804 | +7 | employee stock |
| May-25 | 2,809 | +5 | employee stock |
| Aug-25 | 2,842 | +33 | employee stock, option exercises |
| Nov-25 | 2,873 | +31 | employee stock, option exercises |
| Feb-26 | 2,876 | +3 | |
| May-26 | 2,880 | +4 | |
| Aug-26 | 3,024 | +144 | 141M ATM shares for $19.9B net (~$141 each, 10-Q Note 7) + ~3M employee |
CONFLICT resolved RT-FIX The Macro leg's "$138 average on 144M shares" wrongly included employee shares. The Sentiment leg's "~133M at ~$150 on a 2,825M base" used the wrong base. The correct figures are 141M shares at ~$141, which is 4.9% of the 2,880M starting base. The Red Team could not independently open Note 7, so this rests on the Quant leg's read UNVERIFIED second read.
The founder, Larry Ellison, owns about 38% of Oracle and still sets strategy. That is the best kind of alignment. The two new co-CEOs own almost nothing and are paid in huge option grants. The board is loyal to Ellison rather than a check on him, and Ellison has pledged a growing slice of his shares as collateral for personal loans.
| Person | Stake value at $137.10 | Pay used | Multiple | Verdict |
|---|---|---|---|---|
| Ellison | 1,158,375,174 shares = $158.8B | FY26 total $131.0M (incl. $117.8M option grant at $280.07) | 1,212x | PASS, ideal tier |
| Ellison | same | 3-year average $48.3M | 3,288x | PASS |
| Magouyrk | 250,148 shares = $34.3M | FY26 $627.5M ($621.7M one-time mega option grant) | 0.05x | FAIL (0.20x with exercisable options; ~0.3x with the grant spread over 5 years EST) |
| Sicilia | 182,929 shares = $25.1M | $256.6M | 0.10x | FAIL |
| Catz | 1,126,625 shares = $154.5M | $10.1M | 15x | Weak–moderate |
| Hura | 288,017 shares = $39.5M | $58.3M | 0.7x | FAIL |
Company verdict: ELITE via the founder, WEAK for the operating CEOs.
A Form 4 is the SEC filing an insider must make when buying or selling company stock.
Insiders (6): Ellison · Catz (also on the board of Paramount Skydance, an Ellison family company) · Henley · Magouyrk · Sicilia · Awo Ablo UNVERIFIED
| Independent director | Background | Board since | Role |
|---|---|---|---|
| Michael Boskin, 81 | Stanford/Hoover economist | 1994 | Lead independent director; Audit chair |
| Jeffrey Berg, 79 | Ex-ICM talent agency | 1997 | Independence Committee chair |
| Bruce Chizen, 71 | Ex-Adobe CEO | 2008 | Governance chair |
| Rona Fairhead, 65 | Ex-BBC Trust / FT | 2019 | |
| Charles Moorman, 74 | Ex-Norfolk Southern / Amtrak | 2018 | Compensation chair |
| Stephen Rusckowski, 68 | Ex-Quest Diagnostics | 2025 | |
| Tomislav Mihaljevic, 62 | Cleveland Clinic CEO | 2026 | Healthcare / Cerner tie |
Oracle is really two companies. The database is one of the hardest products in business software to switch away from: a strong moat. The AI rental business is mostly renting Nvidia chips in buildings, which anyone with money and power can do. The blend scores 6 out of 10, which fails the framework's Tier-1 bar of 9.
Oracle's real edges:
Oracle's disadvantages:
| Measure | Oracle OCI | AWS (Amazon) | Azure (Microsoft) | Google Cloud | CoreWeave |
|---|---|---|---|---|---|
| Quarterly revenue | $7.4B | $42.2B | n/a (growth only) | $24.8B | $2.575B |
| Growth vs a year ago | +121% | +37% | +43% | +82% | +112% |
| Backlog | $664B (whole company) | $496B | $678B (commercial) | $514B | $104B |
| Funds capex from own cash? | No | Nearly | Yes | Yes | No |
| In-house AI chip | No | Trainium | Maia | TPU | No |
| Note | 4% market share | Scale leader, re-accelerating | Biggest backlog | Momentum winner | 5% adjusted operating margin: the unit economics of pure GPU rental |
| Metric | ORCL | MSFT | AMZN | GOOGL | CRWV |
|---|---|---|---|---|---|
| Price | 137.10 | 516.17 | 249.67 | 343.92 | 87.59 |
| Market value $B | 415 | 3,833 | 2,693 | 4,206 | 48 |
| Enterprise value $B | 552 | 3,885 | 2,822 | 4,102 | 94 |
| Cash / debt $B | 37.1 / 169.1 | 76.8 / 128.8 | 123.0 / 251.6 | 242.5 / 120.8 | 5.5 / 51.6 |
| Revenue growth, last 12 months | 21.7% | 17.7% | 19.6% | 24.2% | 112.5% |
| Revenue growth, next fiscal year | +34% / +45% (FY28) | +17.8% | +15.6% | +23.8% | +151% |
| 4-quarter operating cash / capex $B | 46.9 / 75.7 | 182.9 / 115.9 | 148.5 / 151.0 | 185.7 / 132.4 | 6.9 / 20.6 |
| 4-quarter free cash flow $B | −28.7 | +67.0 | −2.5 | +53.3 | −13.7 |
| Gross / operating margin % | 64.0 / 35.6 | 67.9 / 45.1 | 50.8 / 13.7 | 60.9 / 34.0 | 67.4 / −1.9 |
| Price-to-sales, last 12 months | 5.78 | 11.55 | 3.47 | 9.43 | 6.36 |
| Enterprise value ÷ next-year sales | 6.10 | 9.94 | 3.41 | 8.23 | 7.33 |
| ...divided by forward growth | 0.18 | 0.56 | 0.22 | 0.35 | 0.05 |
| Forward P/E (price ÷ next year's expected earnings per share) | 16.9 (FY27) / 12.5 (FY28) | 21.8 | 23.8 | 22.8 | loss |
| PEG (P/E ÷ growth) | 0.81 | 1.62 | 1.48 | 1.25 | n/a |
| Share count change, a year | +6.4% | −0.1% | +1.8% | −0.2% | +16.5% |
| % off 52-week high | −57.5% | −6.8% | −13.1% | −15.8% | −42.8% |
| Short % of float | 2.7 | 0.9 | 0.8 | 1.5 | 17.6 |
| Earnings beats, last 8 | 5/8 | 8/8 | 8/8 | 8/8 | 2/6 |
| 11-factor grid score (culture excluded) | 7.5 | 7.5 | 8.5 | 7.0 | 5.5 |
Over the last four calls, Oracle hit or beat every revenue and earnings guide it gave. It also spent more on building than it promised, twice, and its need for outside money grew every quarter. On Wall Street's earnings estimates it beat 5 of the last 8 quarters, including the last 4 in a row.
Non-GAAP EPS means adjusted earnings per share, excluding stock pay and some one-offs.
| Quarter (reported) | Revenue $M | Growth | Revenue vs estimate | Adjusted EPS actual / estimate | Result |
|---|---|---|---|---|---|
| Q2 FY25 (Dec-24) | 14,059 | 8.6% | not pulled | 1.47 / 1.48 | Miss |
| Q3 FY25 (Mar-25) | 14,130 | 6.4% | not pulled | 1.47 / 1.49 | Miss |
| Q4 FY25 (Jun-25) | 15,903 | 11.3% | not pulled | 1.70 / 1.64 | Beat |
| Q1 FY26 (Sep-25) | 14,926 | 12.2% | not pulled | 1.47 / 1.48 | Miss |
| Q2 FY26 (Dec-25) | 16,058 | 14.2% | not pulled | 2.26* / 1.64 | Beat *includes the Ampere gain; the gain-free figure is not verified |
| Q3 FY26 (Mar-26) | 17,190 | 21.7% | above own guide (+19–21% guided, +22% delivered) | 1.79 / 1.69 | Beat |
| Q4 FY26 (Jun-26) | 19,184 | 20.6% | not pulled | 2.11 / 1.96 | Beat |
| Q1 FY27 (Sep-26) | 19,345 | 29.6% | above own guide (+27–29% guided, +30% delivered) | 1.92 / 1.74 | Beat |
| Call | Promise | Status |
|---|---|---|
| Q1 FY26 call (2025-09-09) PM: from Oracle's Q1 FY26 release; not re-pulled by the legs | OCI revenue ~$18B in FY26 | KEPT ($18.1B) |
| Q2 FY26 call (2025-12-10) | FY26 revenue $67B | KEPT ($67.4B) |
| Q2 FY26 call | FY27 revenue raised ~$4B | KEPT (later guided at least $90B) |
| Q2 FY26 call | Q3 revenue +19–21%, cloud +40–44% | KEPT (+22%, +44%) |
| Q2 FY26 call | FY26 capex ~$50B | OVERSHOT ($55.7B) |
| Q2 FY26 call | AI gross margin 30–40% over contract life | PENDING (32% in Q3; not disclosed since) |
| Q2 FY26 call | Keep investment grade | KEPT, barely (BBB- on 7/9/26) |
| Q3 FY26 call (2026-03-10) | Up to $50B raised in calendar 2026, no further bonds | KEPT so far (bonds and loans fell from ~$128–129.5B to $125.3B) CONFLICT: 5/31 balance $128.1B per Qual vs $129.5B per Macro; the 8/31 figure is verified |
| Q3 FY26 call | 10GW+ secured over 3 years, over 90% partner-funded | AT RISK (Jupiter) |
| Q3 FY26 call | 90% of capacity delivered on or ahead of time | PARTIAL (Abilene 618MW / 75% built; Jupiter slipped) |
| Q4 FY26 call (2026-06-10) | FY27 revenue +34% ($90B), EPS $8.05 | RAISED to at least $90B and $8.10 |
| Q4 FY26 call | Q1 revenue +27–29%, cloud +58–64%, EPS $1.72–1.76 | BEAT (+30%; cloud +62% within range; EPS $1.92) |
| Q4 FY26 call | FY27 net capex ~$70B | pending (gross now $90–95B) |
| Q4 FY26 call | ~$40B FY27 financing incl. a $20B stock sale | ATM DONE ($19.9B, 141M shares) |
| Q4 FY26 call | FY30 revenue growth 31% a year, EPS growth 28% a year | PENDING |
At $137, Oracle trades at about 17 times this year's expected earnings and under 6 times last year's sales. That is cheap next to Microsoft and Google, and fair for a company with this much debt and customer risk. Our three scenarios run from $73 to $241. Weighted by probability, they land around $146, only about 7% above today. The bargain zone is lower.
Probability-weighted values: PM $146 (+6.6%) · Quant leg $154 (+13%) · Red Team $129 (−6%). Street mean target $238 sits at our bull case.
| Metric | Value | Worked example |
|---|---|---|
| Price-to-sales (P/S), last 12 months | 5.78x | $414.6B ÷ $71.78B |
| P/S on FY27 guide | 4.58x | $414.6B ÷ $90.48B consensus |
| P/S on FY28 consensus | 3.15x | $414.6B ÷ $131.4B |
| Enterprise value ÷ sales, FY27 / FY28 (incl. leases) | 6.10x / 4.20x | $551.6B ÷ $90.48B; ÷ $131.4B |
| Forward P/E, FY27 guide | 16.9x | $137.10 ÷ $8.10: you pay $16.90 for each $1 of expected yearly profit |
| Forward P/E, FY28 consensus | 12.5x | $137.10 ÷ $11.00 (this is yfinance's "forward P/E"; it is not FY27) |
| GAAP P/E, last 12 months excl. Ampere EST | ~24x | Red Team |
| PEG (P/E ÷ growth) | 0.62–0.81 | 16.9 ÷ 27.4% three-year EPS growth = 0.62; yfinance 0.81. Below 1 is conventionally "cheap for its growth." |
| Price ÷ operating cash | 8.85x | $414.6B ÷ $46.94B |
| Price ÷ free cash flow | not meaningful | free cash flow is negative |
| Enterprise value ÷ EBITDA, last 12 months | 15.2–16.0x | $551.6B ÷ $36.4B = 15.2 |
| Dividend yield | 1.46% | $2.00 ÷ $137.10 |
| Multiple | FY22 | FY23 | FY24 | FY25 | FY26 | Now |
|---|---|---|---|---|---|---|
| Price-to-sales | 4.52 | 5.73 | 6.08 | 8.09 | 9.64 | 5.79 |
| Enterprise value ÷ EBITDA | 13.1 | 18.7 | 18.3 | 24.9 | 25.5 | 15.9 |
| Price-to-earnings | 28.6 | 33.6 | 30.8 | 37.3 | 38.2 | 21.5 |
The entire fall is the multiple shrinking, not the business. On the framework's P/S bands, it moved from "optimistic" (above 10) to "aware" (4–8), and to "left for dead" (2–4) on FY28 sales.
| Line | Bull | Base | Bear |
|---|---|---|---|
| Story | Backlog converts on time; OpenAI funded | Build and power delays; ~1-year slip on part of OpenAI capacity | OpenAI stretches or cuts; FY27 ~$90B then flat |
| FY28 revenue | $135B | $120B | $95B |
| Price-to-sales band | 5.5x ("aware/attractive"; FY23–24 average) | 4.0x (bottom of "aware") | 2.5x ("left for dead"; below FY22's 4.5x low) |
| Market value | 5.5 × 135 = $742.5B | 4.0 × 120 = $480B | 2.5 × 95 = $237.5B |
| Shares | 3,083M (3,024 + 25M preferred + 34M employee) | 3,157M (+28M preferred + 34M employee + ~71M from $10B of stock at $140) | 3,271M (+31M preferred + 34M employee + ~182M from $20B of stock at $110) |
| Price per share | $241 (+76%) | $152 (+11%) | $73 (−47%) |
| Cross-check | 21.9x × $11.00 FY28 EPS = $241 | Enterprise value ≈ $237.5B + $137B net debt = $374.5B ÷ $95B = 3.9x sales RT-FIX: the Quant leg said 2.6x. That is above CoreWeave's 3.6x, so this "bear" is not a distressed price. A true 2.6x would be ~$34 a share. |
The Street's mean target of $238 sits at our bull case.
See the §9 valuation table on the Management & Moat tab. On enterprise value ÷ sales divided by growth, Oracle is 0.18 vs Microsoft 0.56, Amazon 0.22, Google 0.35. On PEG it is 0.81 vs 1.25–1.62.
Stock analysts still love Oracle and retail traders buy every dip. Bond investors are the ones panicking. The framework wants to buy when everyone has given up; that hasn't happened here, so the contrarian gate is not passed.
| Leg | What happened |
|---|---|
| Top | $324.63 close on 9/10/25 ($345.72 intraday). |
| Crash #1 | To ~$137.61 (week of 4/10/26), −58%. |
| Relief rally | To $225.00 (5/29/26), +64%. |
| Crash #2 | The Q4 print (6/11). The S&P cut (7/9). A closing low of $114.99 (7/24). |
| Bounce | To $170.70 (week high, 9/8). |
| Fade | To $137.10 after the earnings gap was sold, the OpenAI IPO delay and Jupiter. |
| Year to date | −29.0% vs the S&P 500's +13.1%. |
Bull; the sell side and retail on bounces. $664B backlog, OCI +121%, 97.9% utilization, non-OpenAI backlog more than doubled, new deals prepaid or customer-chip.
Bear; the credit market, Redburn, S&P. About half the backlog is OpenAI, OpenAI burns ~$278B through 2030, net debt is $88–132B depending on definition, $90–95B of capex, Jupiter force majeure.
Macro. Altman calls a 2026 IPO "ill-advised." Citizens warns the AI labs may spend more cautiously. Oracle is the most borrowed, highest-beta way to express this view (beta 2.04: it moves about twice as much as the market).
Which story the price is discounting: mostly B, with some C. $137.10 ÷ $8.14 FY27 consensus EPS = 16.8x, a mature-software multiple for a company growing 30%. But the stock has not priced B as hard as credit has.
Index funds and long-term institutions (sovereign funds, pensions, bank desks) plus dip-buying retail.
Growth and momentum hedge funds; credit hedgers who buy default insurance and short the stock; and Oracle itself, which sold $20B of stock last quarter.
An independent analyst was told to argue the short case as hard as possible and to audit every other report. The case is that Oracle has become a borrowed-money landlord for one money-losing customer. Most bad outcomes land 10–35% lower, a few 50–70% lower. The Red Team also caught 12 errors in the other legs; all are fixed in this report.
Thesis: Oracle has turned itself into a borrowed-money landlord for one customer that loses money. Wall Street still values it like a software company.
Shared inputs: shares today 3,024M; FY28 consensus sales $131.4B and EPS $11.00; net debt $137B incl. leases, rising to ~$177B after ~$40B more borrowing; each case adds ~34M employee shares plus up to 31M from the preferred. The scenarios overlap, so the probabilities don't sum to 100%.
| Scenario and trigger | Probability (24 months) | Price math | Price | PM response |
|---|---|---|---|---|
| 1. OpenAI stretches or cuts. No $1T+ round, IPO past 2027, contract "re-phasing." FY28 sales $105B; ~40% of lost sales is lost profit (−$8.3B after tax); $15B of stock at $100 (+150M shares) to 3,239M. | 30% | EPS ~$7.90 × 12 | $95 (−31%) | ACCEPT. This is the central risk. Handled by the ≤5% cap, a breaker on any contract restructure, and Zone 3 only if the contract is intact. |
| 2. Junk rating at S&P (BB+), Moody's to Baa3. Borrowing reaches 5x cash profit, or FY27 free cash flow worse than −$42B. +$0.6B a year of interest; $10B of stock at $115 (+87M). | 30% | EPS $10.55 × 11 | $116 (−15%) | ACCEPT the probability, MITIGATE the tail. Moody's is two notches above junk, so index forced-selling needs two agencies RT-FIX. A cut to junk is a breaker (§16). The price impact is moderate because much is priced. |
| 3. Depreciation catch-up. Server life cut from 6 to 5 years, or new Nvidia chips obsolete the fleet. ~$150B server base EST; +$5B a year of cost = −$1.30 EPS. | 20% | EPS $9.70 × 12 | $116 (−15%) | MITIGATE. It's non-cash and doesn't change the funding math. But it is real economic cost, and non-GAAP does not exclude it. Tracked via KPI #4 (margin after depreciation). |
| 4. AI margin ~15%, not 30–40%. Disclosure at Analyst Day, or two quarters of gross margin below 58%. ~$61B FY28 AI sales × a 17-point shortfall = −$10.4B gross profit = −$2.70 EPS. | 25% | EPS $8.30 × 13 | $108 (−21%) | MITIGATE. Segment math shows ~34% on extra revenue and non-GAAP operating margin flat at 42%. But that math may exclude depreciation on unfinished buildings RT-FIX, and leaks say 14–16%. KPI #4 is the tell. Unresolved until a disclosure. |
| 5. Jupiter-style delays spread. A second force majeure, Texas grid curtailment, pipeline slips. FY28 sales $118B with rent and interest still running; −$4.2B profit; $10B of stock at $125 (+80M). | 40% | EPS $9.40 × 13 | $122 (−11%) | ACCEPT. The most likely of the bad scenarios and already live. The price impact is modest and partly priced. A second force majeure is a validator failure, not a breaker. |
| 6. Funding spiral: more stock sales. Prepayments dry up or agencies cap borrowing. $25B of stock at $115 = +217M shares (+7%) to 3,306M. | 45% | $11 × (3,058 ÷ 3,306) = $10.20 × 12.5 | $127 (−7%) | ACCEPT. The single most likely scenario. This is why the plan buys below $124 only and makes new equity above 3% of shares a breaker. |
| 7. Re-priced as a GPU landlord, valued in pieces. Microsoft, Meta or Amazon cut 2027 AI spend. Old business $22B operating profit × 12 = $264B; AI arm $76B × 2.5 = $190B; minus $177B net debt = $277B ÷ 3.09B shares. | 25% | Sum of the parts | $90 (−35%) | MITIGATE: accept the method, reject the multiple. 2.5x is below the weakest peer (CoreWeave 3.6x, with 5% margins). At 3.6x: (264 + 273.6 − 177) ÷ 3.09 = $117. At 2.5x the market would be saying Oracle's AI margin is worse than CoreWeave's. |
| 8. Recession + credit crunch. Junk spreads above 5 points, S&P 500 −25%. FY28 sales $110B; legacy holds; AI start-ups can't raise. | 15% | EPS $8.00 × 9; cross-check: beta 2 × −25% = −50% | $70 (−49%) | ACCEPT. Sizing handles it: −49% on a 5% position is −2.5% of the portfolio. Rate cuts would help refinancing but not the credit spread. |
| 9. OpenAI restructures or goes insolvent. Failed round, missed payments. FY28 ~$90B; capacity re-rented at half price; stranded leases; emergency stock sale. | 5% | 1.5x sales: $135B ÷ 3.4B; pieces $30–45 | $40 (−71%) | ACCEPT as the tail. On a ≤5% position that is −3.5% of the portfolio: survivable. This is why Tier-1 sizing is off the table. |
| 10. Founder's pledged shares sold. Price below lenders' loan-to-value limits (undisclosed) plus Paramount/Warner stress. Up to 413M shares (13.7%). | 5% (only after another scenario drives the price near $100) | −20% on top of a $105 scenario | $85 (−38%) | MITIGATE. The board says these aren't margin loans, and it is conditional on scenarios 1, 8 or 9. Monitor pledge filings; the loan-to-value is a known-unknown. |
| Up. Everything converts on time. FY28 $135B, EPS $11. | 20% | 5.5x sales or 21.9x EPS | $241 (+76%) | Consistent with the Quant bull case. |
Weighted ≈ $129 (−6%).
PM's read: the bad outcomes outnumber the good. Most land at $90–127, 7–35% below spot. Four (scenarios 1, 7, 8, 9) cost 30–70%. That doesn't make Oracle a short at $137. It makes it a stock you only buy where the base case is already in the price: the $116–124 zone.
Also flagged (see ledger): default-insurance level conflict (13); Moody's September status (14); 5/31 bond balance (15); prepayment-strip definition (16); OpenAI revenue $40B vs $65B (17); share growth +6.4% vs +7.0% (18); moving averages adjusted vs unadjusted (19; this report uses unadjusted technicals levels for trading, 200-day $164.54 and 200-week $150.26, vs the Macro leg's $163.8 and $147.6); P/S median (20); bond discount mostly rates (21); retail "$54B" (22); Analyst Day date (23); 34% margin snapshot (24); the $4.9B netting gap (25); the next-12-month vs FY27 window (26); unaudited management claims (27); OpenAI round and ~7% cost (28); prepayment interest immaterial (29).
These are the questions we could not answer from public information, ranked by how much each could move the stock. The top two are worth more than everything else combined. Oracle discloses neither.
| Unknown | Why it matters | How it could surface |
|---|---|---|
| 1. OpenAI's contract terms: exact share of the $664B (S&P "about half" on $638B; press ~$300B; others 54–58%), cancellation or delay rights, take-or-pay (must pay for unused capacity?), prepayment, parent guarantee, payment schedule | Worth ±$100 a share (Red Team) | OpenAI IPO filing (S-1); a 10-Q concentration disclosure; press |
| 2. Lease flexibility on the $288B: termination, delay and power-cost pass-through terms; match to contract length | Decides whether a slip becomes stranded rent | Jupiter resolution; 10-K lease notes; Analyst Day |
| 3. True AI gross margin after full depreciation and rent (last disclosed 32% in Q3 FY26; leaks 14–16%; Q4 and Q1 undisclosed) | Separates a CoreWeave-like value from a software-like one | Analyst Day (10/28 UNVERIFIED); Q2 call |
| 4. Rest of the FY27–28 funding: ~$20B of the $40B plan undone; debt vs equity; any new stock program; FY28 capex plan | Dilution and rating path | Q2 print; 8-K bond filings |
| 5. How much of the $664B is prepaid or customer-chip, and how much capacity isn't built or powered (only $75B disclosed, once) | Oracle's real cash need | Management disclosure |
| 6. Rating triggers: S&P and Moody's exact downgrade thresholds; Moody's September status; Fitch rating | Junk risk | Agency reports |
| 7. Chip useful life and resale value: earnings sensitivity to a 5-year life; value of 4–6-year-old GPUs | Depreciation cliff; re-rent value | 10-K policy changes; secondary GPU markets |
| 8. The $11.4B of prepayments: refundable? Which customers (did OpenAI pay)? Rank in distress? | Quality of 37% of operating cash | 10-Q notes |
| 9. OpenAI's real finances: revenue $40B vs $65B CONFLICT; burn; runway | Customer credit | OpenAI round or IPO documents |
| 10. Partner structures (Crusoe, Vantage, Blue Owl/Stack): guarantees, residual-value promises, risk of being consolidated; status of the $3.3B landlord-loan guarantee that matured Sep 2026 | Hidden liabilities | Q2 10-Q |
| 11. Ellison pledge: loan-to-value, lenders, forced-sale triggers, Paramount/Warner exposure; his exact total share count (proxy table not parsed directly) | Forced-supply tail | Proxy, 13D/Form 4, press |
| 12. Dated default-insurance level: 203 vs 227 basis points CONFLICT; single sources | Credit signal quality | Bloomberg/Markit data |
| 13. Database erosion speed to AI-native data platforms (license −15%) | The cash cow's durability | Segment trend |
| 14. The $4.9B gap between prepayments ($15.36B) and management's capex netting ($10.5B) | Accuracy of "net capex ≤$70B" | Q2 call Q&A |
| 15. When free cash flow turns positive: no date guided | The framework's entry trigger | Analyst Day |
| 16. Revolver and commercial-paper balances at 8/31/26 (last verified 5/31/26) | Liquidity backstop | Q2 10-Q |
| 17. Event dates: Q2 FY27 earnings (~12/10) and Analyst Day (10/28) unconfirmed by Oracle investor relations | Trade timing | Oracle IR |
| 18. Q3 13F flows through the July low (due ~11/14); holder-level verification (Fintel/WhaleWisdom); identities of "Jupiter Topco" and "CTC" | Marginal-buyer read | 13F filings |
| 19. Options data: implied-volatility rank (no history; ORCL not in wheel-radar); current put/call and skew | Wheel pricing | Add ORCL to wheel-radar |
| 20. Earlier backlog next-12-month percentages (Q4 FY26 and before); FY25 quarterly OCI/SaaS and backlog for Q1–Q2 FY25 (derived, not filed); revenue vs consensus by quarter (not pulled) | Trend precision | Older 10-Qs |
| 21. Unaudited management claims: 97.9% utilization, +20% renewal pricing | Pricing-power evidence | None audited |
| 22. Oracle's 2008–09 recession revenue record (not verified, so not used) | Recession scenario | Historical filings |
| 23. Glassdoor figures from a search summary, not a page read; low-view technical podcasts not reviewed; Reddit/Substack posts not retrieved | Talent and sentiment depth | Direct reads |
| 24. Co-CEO option strikes and vesting | Correctly annualizing their pay | Proxy detail |
| 25. The cause of the 9/24 volume spike (1.9x average) is presumed to be Jupiter; not confirmed by the Technicals leg | Minor | News |
Don't buy the stock at $137. The macro gate is yellow, the crowd hasn't given up, and the price only reflects the middle scenario. Get paid to wait by selling puts at $120 or below. Buy more only if the story holds. Take profits in three steps between $165 and $240. Leave the last 30% to run.
Close $137.10 (prior $139.54). All-time high $345.72 intraday 2025-09-10 (OpenAI-deal gap day, $241.51 to $328.33, +36% on 131.6M shares). 52-week high $322.54 (2025-10-16); 52-week low $114.50 (2026-07-28). −60.3% from the all-time high, −57.5% from the 52-week high, +19.7% off the low. SMA20 $148.63, SMA50 $142.25 (flat), SMA200 $164.54 (falling); death cross since 2026-01-07. 200-week MA $150.26 (price BELOW); 50-week MA $176.03. RSI(14) 40.6 daily / 42.0 weekly: neutral-weak, not oversold. Daily MACD −1.23 vs signal 0.33, histogram negative and widening (−0.79, −0.63, −0.81, −1.21, −1.56). Weekly MACD −9.06 vs −9.73, histogram slightly positive. Bollinger (20, 2): $162.29 / $134.97; price on the lower band. ATR(14) $7.10 = 5.18%/day. HV20 51.2% (13.9th percentile of the 39–126% 1-year range). Higher lows $114.50, then $137.43, then $133.48; highs stalled at $159.26, then $170.70. Q1 FY27 gap to $164.43/$166 sold to $150.28 on 80.4M. 09-24 low $133.48 on 56.6M (~1.9x average): distribution (heavy selling by large holders). Verdict: early, unconfirmed base, not a Stage-2 uptrend. Confirmation needs (1) a weekly close above the 200-week MA ($150.26) that holds, (2) a weekly close above $171, (3) SMA50 turning up above SMA200. Failure: a weekly close below $133 retargets $121.76, then $114.50.
Glossary for this block: SMA = simple moving average, the average close over N days. RSI = a 0–100 momentum gauge; below 30 is oversold. MACD = a trend-momentum indicator. ATR = average daily price swing. HV20 = realized volatility over 20 days. Bollinger Bands = a price envelope two standard deviations around the 20-day average. MA = moving average.
Fib retracements to the $345.72 top. "Fib" = Fibonacci retracement checkpoints measuring how far a move has pulled back.
| Swing low | 0.382 level | 0.5 level | 0.618 level | 0.786 level |
|---|---|---|---|---|
| $60.78 (2022-09-29) | $236.87 | $203.25 | $169.63 | $121.76 |
| $82.04 (2023-03-13) | $244.99 | $213.88 | $182.77 | $138.47 |
| $118.86 (2025-04-07) | n/a | $232.29 | $205.52 | $167.41 (broken) |
Glossary: "Fib" = Fibonacci retracement checkpoints measuring how far a move has pulled back. HVN = high-volume node, a price zone where many shares changed hands. P/S = price-to-sales. TTM = trailing twelve months.
| Trigger | Action | Why this level |
|---|---|---|
| Trim 1: $165–172 | Trim 20% | Falling SMA200 ($164.54), 0.618 fib ($169.63), 0.236 bounce ($169.07), earnings-week high ($170.70). |
| Trim 2: $200–206 | Trim 25% | 0.5 fib of the 2022 swing ($203.25), 0.382 bounce ($202.83), 0.618 fib of the 2025 swing ($205.52), $190–195 post-top HVN just below. |
| Trim 3: $228–240 | Trim 25% | 0.5 bounce ($230.11), 0.5 fib of the 2025 swing ($232.29), TTM P/S 10x ($237.38). |
| Runner | Keep the remaining 30%, uncapped | Framework layer 4: room for a big move. |
| Valuation rule | No adds above 8x TTM P/S (~$190); mandatory trim above 10x (~$237) | Protects against paying up in euphoria. |
| Time rule | No weekly close above the 200-week MA by the March 2027 print (Q3 FY27): cap the position at Zone 1 size. No sequential free-cash-flow improvement and no decelerating capex guide by the June 2027 FY27 results: re-underwrite from scratch. | Stops dead money from sitting forever. |
Re-quote live, because after-hours spreads mislead.
Glossary: CSP = cash-secured put, where you are paid now to agree to buy at the strike. CC = covered call, where you are paid now to agree to sell shares you own at the strike. DTE = days to expiry. Delta ≈ rough odds of finishing in the money. IV = implied volatility, the move size options are pricing.
| Zone | Expiry (days left) | Strike | Premium | Yield | Annualized | Basis if assigned | Delta |
|---|---|---|---|---|---|---|---|
| 1 | 10-30 (34) | $135 | $7.62 | 5.65% | 60.6% | $127.38 (−7.1%) | −0.42 |
| 1 | 10-30 (34) | $130 | $5.38 | 4.13% | 44.4% | $124.62 | −0.33 |
| 1 | 11-20 (55) | $130 | $7.35 | 5.65% | 37.5% | $122.65 | −0.35 |
| 2 | 11-20 (55) | $125 | $5.42 | 4.34% | 28.8% | $119.58 | −0.28 |
| 2 | 11-20 (55) | $120 | $3.88 | 3.23% | 21.4% | $116.12 | −0.21 |
| 2 | 10-30 (34) | $120 | $2.32 | 1.93% | 20.8% | $117.68 | −0.17 |
| 3 | 11-20 (55) | $115 | $2.65 | 2.30% | 15.3% | $112.36 | −0.16 |
| 3 | 11-20 (55) | $110 | $1.80 | 1.64% | 10.9% | $108.20 | −0.11 |
| 3 | 01-15-27 (111, through earnings) | $110 | $4.80 | 4.36% | 14.3% | $105.20 | −0.18 |
| 3 | 01-15-27 (111, through earnings) | $100 | $2.73 | 2.73% | 9.0% | $97.27 | −0.11 |
| Band | Expiry (days left) | Strike | Premium | Yield | Annualized | Delta |
|---|---|---|---|---|---|---|
| Income only | 10-30 (34) | $150 | $4.38 | 3.19% | 34.3% | 0.33 |
| Trim 1 | 11-20 (55) | $165 | $3.38 | 2.46% | 16.3% | 0.23 |
| Trim 1 | 11-20 (55) | $170 | $2.73 | 1.99% | 13.2% | 0.19 |
| Trim 1 | 01-15-27 (111) | $170 | $7.05 | 5.14% | 16.9% | 0.31 |
| Trim 2 | 01-15-27 (111) | $200 | $3.25 | 2.37% | 7.8% | 0.17 |
| Trim 2 | 11-20 (55) | $200 | $0.84 | 0.61% | 4.1% | 0.07 |
| Trim 3 | 01-15-27 (111) | $230 | $1.69 | 1.23% | 4.0% | 0.09 |
| Trim 3 | 01-15-27 (111) | $240 | $1.26 | 0.92% | 3.0% | 0.07 |
Below are the risks ranked by what would actually kill the investment case, the dated checkpoints that would prove it right or wrong, and five numbers we will track that Wall Street mostly doesn't.
| Risk | Mechanism |
|---|---|
| 1. OpenAI's ability to pay | ~Half the backlog; $278B of projected burn; IPO slipped; compute plans cut by more than half. Kills via contract "re-phasing," which strands capacity against fixed rent. |
| 2. Funding access and price | BBB-; ~$20B of the FY27 plan undone; a $0–12B cushion; default insurance at a reported record; new 10-year money ~7% EST. Kills via a junk cut plus a forced equity raise at lower prices (at $100, $20B = 200M shares, +6.6%). |
| 3. Duration mismatch | $288B of 15–19-year leases vs ~5-year contracts. Kills slowly: rent continues after contracts end. |
| 4. True AI margin | 32% claimed; 14–16% leaked; depreciation on $48.5B not yet flowing. Kills the "better than CoreWeave" premium. |
| 5. Macro / cost of capital | Fed hiking; 10-year 5.18%; beta 2.04; liquidity past peak. Amplifies 1 and 2 together. |
| 6. Construction and power | Jupiter slip; PJM capacity prices at the cap ($329.17/MW-day); an ERCOT (Texas grid) queue of 410–438GW; 27 states with data-center bills. Delays revenue, not rent. |
| 7. Dilution machinery | ATM precedent; 25–31M preferred shares; $4.8B a year of stock pay. |
| 8. Governance | 413M pledged shares; a $40.4B personal guarantee; a friendly board; untested succession (82/81/81). |
| 9. Depreciation-policy risk | A 6-year server life vs peers' 5. |
| 10. Database erosion | License −15%; #3 in Gartner's ranking. |
| 11. Political and headline risk | The 15% TikTok USDS stake; Stargate politics; the Section 232 chip tariff (currently exempt for data centers of 100MW+). |
| Validator | Check by | Passes if |
|---|---|---|
| Analyst Day disclosure | 2026-10-28 UNVERIFIED | AI gross margin 30%+ after depreciation disclosed; FY30 targets held; a date for positive free cash flow |
| Q2 FY27 print | ~2026-12-10 UNVERIFIED | Revenue +30–34% USD; cloud +65%+; EPS $1.85–1.93; backlog ≥ $664B |
| Organic cash in a non-August quarter | Q2 FY27 10-Q | Operating cash before prepayments ≥ $8B (Red Team cover bar: $12B); net capex on track for ≤$70B |
| No new equity | Q2 and Q3 FY27 10-Qs | Share count ≤ 3,045M |
| Q3 13F filings | ~2026-11-14 | Quality long funds adding at $115–130 |
| Jupiter resolved | 2027-03-31 (pipeline due Feb 2027) | Force majeure lifted, or partner financing re-closed without new lender restrictions |
| OpenAI funding | 2027-06-30 | $1T+ round or IPO filing; payment schedule confirmed |
| Rating stable | 2027-06-30 | No S&P cut; Moody's outlook back to stable |
| Default insurance | Rolling | Below 150 basis points |
| Free-cash-flow path | FY27 results, ~June 2027 | Sequential improvement; FY28 capex guide below FY27 |
| Breaker | Check by |
|---|---|
| OpenAI contract restructured, delayed or unpaid, or OpenAI fails to raise | Continuous; each print |
| Backlog down quarter over quarter | Each print (next ~12/10) |
| Junk at S&P or Moody's, BBB- on negative watch, or default insurance above 300 basis points | Continuous |
| New equity beyond the completed $20B, or shares up more than 3% in 12 months excluding the preferred | Each 10-Q |
| Gross margin below 55%, or two quarters below 58% without a disclosed 30%+ AI margin | Each print |
| Two consecutive weekly closes below $114.50 | Weekly |
| A second force majeure on another major site, or any enforced sale of pledged shares | Continuous |
| KPI | How to compute | Current | Target | Next check |
|---|---|---|---|---|
| 1. Self-funding ratio: does the business pay for its own build? | (Operating cash − all prepayment build) ÷ capex | 0.27 in Q1 FY27 ((23.10 − 15.36) ÷ 28.50) | ≥0.5 in a non-August quarter by Q3 FY27; ≥1.0 by FY29 | Q2 FY27 10-Q, ~Dec 2026 |
| 2. Non-OpenAI backlog: demand that doesn't depend on one customer | Backlog − estimated OpenAI share | ~$330–364B EST | Grows ≥$15B a quarter; >50% of new bookings non-OpenAI | Q2 FY27 print |
| 3. Rent-to-contract cover: is Oracle signing rent faster than customers sign contracts? | Unstarted leases ÷ backlog | 0.43 ($288B ÷ $664B); was 0.41 at Q4 FY26 ($260B ÷ $638B) and 0.47 at Q2 FY26 ($248B ÷ $523B) | ≤0.45 and flat or falling | Each 10-Q |
| 4. AI gross margin after depreciation | Disclosed, or derived from segment costs incl. depreciation | Last disclosed 32% (Q3 FY26); ~34% on extra revenue (derived, likely before full depreciation) | ≥30% disclosed, two quarters running | Analyst Day / Q2 FY27 |
| 5. Share count discipline | Quarter-end shares | 3,024M (8/31/26) | ≤3,045M at the Q3 FY27 10-Q (employee dilution only) | Q2 FY27 10-Q |
v1.0 — 2026-09-26 (Liquid Wheel Research · deep-dive team)
First full deep dive on Oracle. Five research legs merged (Quant, Qualitative/management/moat, Macro, Sentiment, Technicals), then audited by an independent Red Team; 12 corrections applied. 16-gate checklist (5 PASS / 5 WATCH / 6 FAIL), capital-structure six-pack, eight-quarter table, cash bridge and FY27 funding test, debt and dilution schedules, management and board read, moat and competitor tables, earnings tracker, sentiment map, bull/base/bear scenarios, ten downside scenarios, known-unknowns ledger, entry ladder, exits, wheel strikes, validators, breakers and five first-principles KPIs. Data as of the 2026-09-25 close; filings through the 10-Q for the quarter ended 2026-08-31 and the proxy filed 2026-09-25. Score 6.0 / 10. Status: ON WATCH. Written plain-English first: every section opens with a plain summary, every technical point ends with "What this means," and every term is in the Glossary tab.
Baseline before v1.0: the 12-factor triage of 2026-09-26 (score 8.5/12, "worth the deep dive"). No earlier ORCL deep dive or tracker exists.
Next scheduled review: Q2 FY27 earnings, about 2026-12-10 UNVERIFIED. Earlier checkpoint: Analyst Day, 2026-10-28 UNVERIFIED.
Liquid Wheel Research, Deep-Dive Framework. Data as of the 2026-09-25 close, compiled 2026-09-26. Price $137.10. Shares 3,023,736,000 (10-Q cover, as of 2026-09-07). Market value of the stock $414.6B. Enterprise value (stock value plus debt minus cash) $551.6B including leases.
Oracle's fiscal year ends May 31. "FY27" runs June 1, 2026 to May 31, 2027. "Q1 FY27" is June to August 2026, reported 2026-09-10. Filings used run through the 10-Q for the quarter ended 2026-08-31 and the proxy filed 2026-09-25.
Five research legs were merged: Quant (framework Stages 2, 4, 8), Qualitative/management/moat (3, 5, 6), Macro (1), Sentiment (7) and Technicals (9–10). An independent Red Team then audited all five. The framework is investing-framework.md, Stages 0–12. The 16 gates are the same checklist used for IREN v3, so grades can be compared across names. There is no earlier ORCL deep dive. The "prior grade" column uses the 12-factor triage run on 2026-09-26 (score 8.5/12).
| Tag | Meaning |
|---|---|
| CONFLICT | Two legs disagree. The better-sourced number is used and the other is shown. |
| UNVERIFIED | Comes from a single secondary source, or no source. |
| RT-FIX | A Red Team correction that has been applied. |
| EST | An estimate or assumption, not a reported figure. |
| PM | A portfolio-manager judgment or an addition not found in the agent legs. |
Every number above traces to one of these. Company filings come first because they are the most reliable. Press and aggregator sources are marked, and single-source claims are flagged in the text.
Every term, abbreviation and ratio used in this report, each defined in one sentence. Where it applies, Oracle's actual number is given. Anywhere on the page, a word with a dotted underline is in this list: hover over it, or tap it on a phone, to see the definition.
| Term | Plain-English definition | Oracle's value |
|---|---|---|
| 0.786 / 0.618 / 0.5 / 0.382 / 0.236 fib | Standard "how far did it pull back" checkpoints on a big price move, from Fibonacci ratios; traders watch them as support or resistance. | 0.786 of the 2022 swing = $121.76 |
| 10-K / 10-Q / 8-K | Annual report / quarterly report / event filing a company must file with the SEC. | 10-Q for the quarter ended 8/31/26 |
| 12-factor triage | The framework's 5-minute screen deciding whether a stock deserves deep research. | 8.5/12 |
| 13F | Quarterly filing where big investors disclose their stock holdings. | Q2 2026: 1,831 buyers vs 1,562 sellers |
| 16 gates | The framework's pass/fail checklist used across all deep dives. | 5 PASS / 5 WATCH / 6 FAIL |
| 200-week moving average | The average price over roughly four years; long-term buyers often defend it. | $150.26 (price below) |
| Abilene | Oracle's flagship Texas data-center campus built for OpenAI. | 618MW, 6 of 8 buildings |
| Adjusted / non-GAAP | Company-defined figures that exclude items like stock pay and one-offs. | Non-GAAP EPS $1.92 vs GAAP $1.56 (Q1 FY27) |
| Altman Z-score | A formula estimating bankruptcy risk; below 1.8 is distress, above 3 is safe. | 1.93 (grey zone) |
| Amortization | Writing down the cost of intangible assets (like acquired customer lists) over time. | $202M added back in Q1 non-GAAP |
| Analyst Day | A company event where management gives long-term targets. | 2025-10-16; next 2026-10-28 UNVERIFIED |
| Annualized | A short-period number scaled up to a full year. | Q1 interest $1.428B × 4 ≈ $5.7B |
| Assignment (options) | When a put seller is required to buy the shares, or a call seller to sell them. | |
| ATM (at-the-market) program | Selling new shares gradually into the open market to raise cash; dilutes owners. | 141M shares, $19.9B, ~$141 each |
| ATR (average true range) | The average daily price swing, in dollars. | $7.10 = 5.18% a day |
| Backlog | See RPO. | $664B |
| Bare metal / hypervisor | Renting the raw machine vs renting through a virtualization layer (the hypervisor) that splits it up. | OCI Gen2 is bare metal |
| Basis (options) | Your effective purchase price if a put is assigned: strike minus premium. | 11-20 $120 put: $116.12 |
| Basis point (bp) | One hundredth of a percent (0.01%). | 227 bp = 2.27% |
| BBB- / Baa3 | The lowest investment-grade credit rating at S&P / Moody's; one notch lower is junk. | S&P BBB- (7/9/26); Moody's Baa2 |
| Bear / base / bull case | Pessimistic / middle / optimistic scenarios. | $73 / $152 / $241 |
| Beta | How much a stock tends to move vs the market; 2 means twice as much. | 2.04 |
| Bid/ask midpoint | Halfway between what buyers pay and sellers ask; used to price options. | |
| Black-Scholes | The standard formula for estimating option prices and deltas. | |
| Bollinger Bands | A price envelope two standard deviations around the 20-day average. | $162.29 / $134.97 |
| Bonds / notes | IOUs a company sells to investors, paying interest (the coupon). | $125.3B bonds and loans |
| Book-to-bill | New orders ÷ revenue; above 1 means the order book is growing. | ~2.3x (Q1 FY27) |
| BYO hardware | "Bring your own chips": the customer buys the GPUs and Oracle runs them. | Part of $75B self-funded backlog |
| CAGR | Compound annual growth rate: the steady yearly growth that gets you from A to B. | 36.2% needed for $125B FY28 |
| Capex (capital expenditures) | Money spent on buildings, chips and equipment. | $75.7B over 12 months; FY27 guide $90–95B |
| Capitulation | When investors give up and sell in despair, often marking a bottom. | Not yet seen |
| Cash-secured put (CSP) | Being paid now to agree to buy a stock at a set price, with cash set aside. | 11-20 $120 put, $3.88 |
| CDS (credit default swap) | Insurance against a company defaulting; the price rises when lenders worry. | ~203–227 bp CONFLICT |
| Commercial paper | Very short-term company IOUs. | $1.468B (5/31/26) |
| Constant currency | Growth with exchange-rate moves stripped out. | Q1 revenue +30% both ways |
| Construction in progress | Buildings and equipment not yet in service, so not yet depreciated. | $48.5B |
| Contrarian gate | The framework test: buy when others have given up, avoid when everyone is bullish. | Not passed |
| Conviction-spec | The framework's sizing tier for names without a moat of 9+; smaller than Tier-1. | Oracle's tier; ≤5% |
| Covenant | A condition lenders attach to loans. | In compliance (5/31/26) |
| Covered call (CC) | Being paid now to agree to sell shares you own at a set price. | Trim 1 line: 11-20 $165 |
| CoWoS | TSMC's advanced chip packaging, a bottleneck for AI chips. | Sold out through Q4 2026 |
| CPI | Consumer price index, the main inflation measure. | 3.4% headline |
| cRPO | The part of the backlog expected to become revenue within 12 months. | ~$86B (13%) |
| Current ratio | Short-term assets ÷ bills due within a year; the framework wants 2 or more. | 1.17 |
| Days-to-cover | Days of normal volume shorts would need to buy back their positions. | 1.82 |
| Death cross | The 50-day average falling below the 200-day; a bearish trend signal. | Since 2026-01-07 |
| DEF 14A (proxy) | The annual shareholder-meeting filing with pay, ownership and board details. | Filed 2026-09-25 |
| Delta | Roughly the odds an option finishes in the money. | 11-20 $120 put: −0.21 |
| Depreciation | The yearly cost of equipment wearing out, spread over its useful life. | $3.19B in Q1 FY27; 6-year server life |
| Diluted shares | Share count including options and awards likely to become shares. | 3,000M average (Q1 FY27) |
| Dilution | New shares shrinking each owner's slice. | +6.4% a year |
| Distribution (technicals) | Heavy selling by large holders, visible as high-volume down days. | 9/24: 56.6M shares |
| Dividend / ex-dividend | Cash paid to shareholders / the date you must own the stock by to get it. | $0.50 a quarter; ex-date 10/8 |
| Dot plot | The Fed officials' chart of where they expect rates to go. | 4.00–4.25% end-2026 |
| DTE | Days until an option expires. | |
| Duration mismatch | Long-term obligations funded by shorter-term income. | 15–19-year leases vs ~5-year contracts |
| EBITDA | Earnings before interest, taxes, depreciation and amortization; a rough stand-in for cash profit. | $36.4B over 12 months |
| Enterprise value (EV) | The price to buy the whole company: stock value + debt − cash. | $551.6B incl. leases |
| EPS | Earnings per share: profit ÷ shares. | FY27 guide $8.10 |
| ERCOT | The Texas power grid operator. | 410–438GW queue |
| EV/EBITDA | Enterprise value ÷ EBITDA. | 15.2x |
| EV/S | Enterprise value ÷ sales. | 6.10x FY27 |
| Executive Chair | A chairman who also has an executive role. | Ellison |
| Face value / fair value | What a bond pays back at maturity / what it would sell for today. | $125.0B / $105.7B |
| Fallen angel | A company cut from investment grade to junk, forcing some funds to sell. | Not yet |
| FCF margin | Free cash flow ÷ revenue. | −40.0% |
| Fed funds rate | The Fed's main policy interest rate. | 3.75–4.00% |
| Finance lease | A lease that is effectively a purchase; counted like debt. | $9.19B |
| Fiscal year (FY) | A company's accounting year. | Ends May 31; FY27 = Jun 2026–May 2027 |
| Float | Shares available to trade (excluding insiders). | ~61% of shares |
| Flywheel | A self-reinforcing business loop. | Half-spinning |
| FOMC | The Fed committee that sets rates. | Next 10/27–28 |
| Force majeure | A contract clause allowing delay for events outside a party's control. | Jupiter, 9/24/26 |
| Form 4 | The filing insiders make when they buy or sell. | $0 bought, $119.7M sold |
| Forward P/E | Price ÷ next year's expected EPS. | 16.9x FY27; 12.5x FY28 |
| Free cash flow (FCF) | Operating cash flow minus capex: cash left after paying for buildings and equipment. | −$28.7B over 12 months |
| GAAP | US generally accepted accounting principles, the official rules. | GAAP EPS $1.56 (Q1) |
| GDPNow | The Atlanta Fed's live estimate of current-quarter growth. | 5.0% |
| GPU | Graphics processing unit, the chip used for AI. | 300k+ delivered since Q4 |
| Gross margin | Share of revenue left after the direct cost of delivering it. | 60.0% (Q1 FY27) |
| Guidance | Management's own forecast. | FY27 revenue ≥$90B |
| GW / MW | Gigawatt / megawatt, units of data-center power (1GW = 1,000MW). | 850MW delivered in Q1 |
| High-yield (HY) spread / OAS | Extra yield junk bonds pay over Treasuries. | ~2.80 points |
| HV20 (realized volatility) | How much the stock actually moved over 20 days, annualized. | 51.2% |
| HVN (high-volume node) | A price zone where many shares traded; acts like a floor or ceiling. | $140–155 |
| Hyperscaler | A giant cloud company (AWS, Azure, Google Cloud). | |
| IaaS | Infrastructure-as-a-service: renting raw computing power. | OCI |
| Implied volatility (IV) | The size of move the options market is pricing. | 50–54% |
| Interest coverage | Operating income ÷ interest expense. | 4.7x |
| Investment grade | Credit ratings of BBB-/Baa3 or better. | Barely |
| IPO | Initial public offering, a first stock listing. | Oracle 1986; OpenAI 2027? |
| ISM | Monthly business surveys; above 50 means growth. | 54.6 / 55.4 |
| IV rank | Where today's implied volatility sits vs its own past year. | n/a |
| Junk | Credit ratings below BBB-/Baa3; more expensive borrowing. | One notch away (S&P) |
| Known-unknowns ledger | The ranked list of questions we can't answer yet. | §14 |
| Lease (operating) | Renting buildings long-term; now counted as a liability. | $34.62B |
| Liquidity cycle | The rise and fall of money available in markets. | Past peak |
| Loan-to-value (LTV) | Loan size ÷ collateral value; a margin call comes if it gets too high. | Undisclosed for Ellison |
| MA (moving average) | The average closing price over a set number of days or weeks. | 200-week MA $150.26 |
| MACD | A trend-momentum indicator; a negative histogram means sellers are winning. | Daily −1.23 |
| Macro stamp | The framework's Stage 1 verdict on the economy. | YELLOW |
| Mandatory convertible preferred | Stock-like security paying a fixed dividend that automatically becomes common shares. | $5B, 6.5%, 25–31M shares 1/15/29 |
| Margin call | A lender's demand for more collateral, which can force sales. | Board says the pledges aren't margin loans |
| Market cap | Share price × shares outstanding. | $414.6B |
| Moat | A durable competitive advantage. | 6/10 |
| MOVE | The bond market's fear gauge. | ~104 |
| Multicloud | Oracle's database running inside rivals' clouds. | +353% |
| Neocloud | A newer company that only rents AI chips (e.g. CoreWeave). | |
| Net debt | Debt minus cash. | $88.3B (bonds) / $132.1B (incl. leases) |
| Net margin | Net income ÷ revenue. | 24.6% (Q1 FY27) |
| No-chase line | The price above which the plan forbids buying. | $152 |
| OCF ÷ capex | How much of the build operations pay for. | 0.81 (0.27 without prepayments) |
| OCI | Oracle Cloud Infrastructure, the computer and AI-chip rental business. | $7.39B a quarter, +121% |
| Open interest | Number of open option contracts. | 01-15-27: 243k puts |
| Operating cash flow (OCF) | Cash the business brings in from operations. | $46.94B over 12 months |
| Operating leverage | Profits growing faster than sales once fixed costs are covered. | Not yet visible in cash |
| Operating margin | Operating income ÷ revenue. | 35% GAAP; 42% non-GAAP |
| Organic growth | Growth not bought through acquisitions. | 100% |
| P/E | Share price ÷ earnings per share. | 16.9x FY27 |
| P/FCF / P/OCF | Price ÷ free cash flow / price ÷ operating cash flow. | n/m / 8.85x |
| P/S (price-to-sales) | Market value ÷ yearly revenue. | 5.78x trailing |
| P/S bands (framework) | >30 euphoric, >20 excited, >10 optimistic, 4–8 aware, ~5 attractive, 2–4 left for dead. | Aware (trailing) |
| Par | 100 cents on the dollar, a bond's face value. | Jupiter loans at 89–91 |
| PEG | P/E ÷ expected growth rate; below 1 is conventionally cheap. | 0.62–0.81 |
| Performance stock options (PSOs) | Options that vest only if a target is hit. | FY28 revenue $100B / $125B / $250B |
| Piotroski F-score | A 0–9 financial-health checklist. | 5 |
| PJM | The Mid-Atlantic power grid operator. | Capacity $329.17/MW-day |
| Pledged shares | Stock put up as collateral for a personal loan. | 413M (13.7%) |
| PM (portfolio manager) | The person making the final sizing and trade calls; "PM" tags mark that judgment. | Probability-weighted value $146 |
| Prepayment | A customer paying upfront, often so Oracle can buy chips. | $15.36B in Q1 FY27 |
| Put/call ratio | Put volume ÷ call volume; a fear gauge. | Stale (0.45) |
| Q1–Q4 | Fiscal quarters. | Q1 FY27 = Jun–Aug 2026 |
| Related party | A company tied to an insider doing business with the firm. | Skydance ~$6.5M |
| Restricted cash | Cash set aside for a specific purpose, not freely usable. | $2.6B |
| Revolver (revolving credit line) | A standby bank loan a company can draw on. | $10B, undrawn |
| RPO (remaining performance obligations) | Signed contracts not yet delivered or billed: the backlog. | $664B |
| RSI | A 0–100 momentum gauge; below 30 oversold, above 70 overbought. | 40.6 |
| Rule of 40 | Revenue growth % + FCF margin % should be 40 or more. | −18.3 |
| Runner | The uncapped share of a position kept for a monster move. | 30% |
| SaaS | Software-as-a-service: subscription apps. | $4.22B a quarter, +10% |
| SBC (stock-based compensation) | Employee pay in shares; a real cost that dilutes owners. | $4.81B over 12 months |
| Section 232 tariff | A US national-security tariff; 25% on advanced chips with a data-center exemption. | Oracle exempt |
| Secular correction | A long, deep decline (60–85%) that resets a stock. | −60.3% (edge) |
| SEP | The Fed's Summary of Economic Projections. | GDP 2.3% (2026) |
| Short interest | Shares borrowed and sold, betting on a fall. | 2.7% of float |
| Skin in the game | Insider stake value ÷ annual pay. | Ellison 1,212x |
| SMA | Simple moving average. | SMA200 $164.54 |
| SOFR | The benchmark overnight lending rate used for floating loans. | Term loan at SOFR+1.35% |
| Stage-2 uptrend | A confirmed rising trend after a base. | Not yet |
| Stargate | The OpenAI-led US AI data-center program Oracle builds for. | |
| Stock split / offering check | Checking that market cap matches to catch errors. | Reconciles |
| Strike | The price at which an option lets you buy or sell. | |
| Superclusters | Very large linked groups of GPUs for AI training. | |
| Take-or-pay | A contract where the customer pays for capacity whether it uses it or not. | Unknown for OpenAI |
| TAM | Total addressable market. | ~$574B cloud infrastructure (annualized) |
| TGA / reserves | Treasury's cash account at the Fed / banks' cash at the Fed. | Reserves $2,930B |
| Thesis breaker / validator | An event that kills / confirms the investment case. | §16 |
| Tier-1 | The framework's top sizing tier; requires a moat of 9+. | Not eligible |
| TPU / Trainium / Maia | Google / Amazon / Microsoft in-house AI chips. | Oracle has none |
| Trim | Selling part of a position at a target. | $165–172 first |
| TTM | Trailing twelve months, i.e. the last four quarters. | Revenue $71.78B |
| Unconditional purchase obligations | Committed future purchases, like chips. | $34.15B |
| Unearned revenue | Cash received for service not yet delivered. | $30.8B |
| Utilization | Share of built capacity in use. | 97.9% (unaudited) |
| VIX | The stock market's fear gauge. | 14.87 |
| Wheel | Selling puts to buy, then calls to sell, repeatedly, around a stock you'd own. | §15 |
| Working capital | Short-term assets minus short-term liabilities; changes affect cash flow. | +$16.43B (12 months) |
| Yield (option) | Premium ÷ strike or price. | 11-20 $120 put 3.23% |
| YoY / YTD | Year over year / year to date. | Revenue +29.6% YoY; stock −29.0% YTD |