ORCL › analyze
ORCL · Analyze from before
Date: 2026-08-28 · Command: /analyze-from-before · Price: $151.94 (prev. close $148.87)
Baseline correction: The orchestrator named analyze-2026-07-09.md as the prior report. kb.py find ORCL surfaced a newer verdict-bearing report at analyze-2026-07-23.md (HOLD, conviction 7.0) that supersedes it. Per protocol ("read the newest before re-deriving"), this re-analysis is tested against the 2026-07-23 report, not 07-09. The 07-09 BUY verdict was already stale relative to 07-23 before this run started.
Headline finding: no new fundamentals since baseline — this is a pure re-rating
Oracle's fiscal year ends May 31; FY2026 results (revenue $67.36B, FCF −$23.7B, RPO $638B) were already fully reflected in the 07-23 baseline. Nothing has been reported since — Q1 FY2027 is due ~Sep 9-10, 2026, still two weeks out. Every fundamental line item below is unchanged from the baseline. What changed is price (+20.3% from $126.29 to $151.94) and the multiple the market is willing to pay for the same unchanged numbers.
Claim Ledger — every 2026-07-23 claim re-tested
| # | Baseline claim | Status | Fresh evidence |
|---|---|---|---|
| 1 | FY26 FCF −$23.7B, "the crux" | CARRIED | fin.py reconciles to −$23.69B (OCF $31.98B − capex $55.66B). Yahoo's snapshot freeCashflow field reads −$24.54B, a ~3.6% divergence — matches [[pitfall-yahoo-snapshot-fcf-field-diverges-from-quarter-sum]]; use the reconciled OCF-minus-capex figure, not the snapshot aggregate. |
| 2 | FY27 FCF likely deepens toward −$45B to −$50B | UNTESTED | No new quarter has printed. First read is the Sep print. |
| 3 | Total debt $156.2B, net debt $98.3B, net debt/EBITDA ~3.1x | CARRIED | roic.ai cross-check: net debt/EBITDA 3.29x, total debt/EBITDA 5.22x FY26 (up from 4.36x FY25) — same direction, confirms leverage still climbing on a gross basis even as EBITDA growth partly offsets net debt/EBITDA. |
| 4 | ROIC 12.1%, declining from 14.0% | REFRESHED | roic.ai's own basis: return_on_inv_capital 11.14% FY26 vs 13.72% FY25 — different denominator (methodology gap between Yahoo-based and roic.ai calcs) but same direction, same magnitude of decline. Moat-erosion signal confirmed independently, not an artifact of one vendor. |
| 5 | Gross margin 65.8%, down from 70.5% | CARRIED | Unchanged — same annual print, no new data. |
| 6 | RPO $638B (+363% YoY) | CARRIED | No new backlog disclosure until Sep print. |
| 7 | OpenAI/Stargate concentration is the central risk; BYO/prepaid mix ($75B) de-risks the capital model | CARRIED | No material new development found. Five new Stargate sites and the Abilene 600MW scrap were both already known at baseline; recent coverage recycles the same facts. Net: unresolved, unchanged, still the #1 named risk. |
| 8 | Dual-CEO governance (Magouyrk + Sicilia), overall governance risk score 9/10 | CARRIED | Unchanged; both still listed as co-CEOs in the current officer roster. |
| 9 | Active securities-fraud class action (Kessler Topaz), class period Jun–Dec 2025 | REFRESHED | Lead plaintiffs confirmed as Sparinvest S.A. and SEB Funds AB; amended complaint filed 2026-07-14. Case is progressing, not resolved or dismissed — still a live overhang, now with more procedural detail. |
| 10 | "Loss of investment-grade rating" listed as a forward thesis-break trigger | DRIFTED — under-stated, not fired | S&P downgraded Oracle to BBB- on 2026-07-09 (one notch above junk), citing the AI buildout and OpenAI concentration explicitly. Moody's holds a negative outlook. Neither the 07-09 nor the 07-23 report named this rating action, despite it predating both. The trigger has not fired — BBB- is still investment-grade — but the buffer to junk is now zero notches, materially tighter than either report characterized it. This is the most important correction in this re-analysis. |
| 11 | Analyst consensus ~Buy, mean PT $248-262 | DRIFTED | Current mean target $244.12 (median $240), little changed even as the stock rallied 20% — UBS cut its target $285→$245 on 2026-08-06 (still Buy); CLSA initiated Hold at $145 on 2026-07-20 (right near the trough — the most skeptical fresh voice); Citi placed ORCL on a 90-day catalyst watch on 2026-08-26, reiterating Buy/$330, calling the drawdown "four to five standard deviations" beyond normal and a buying opportunity. Net: the Street has not chased the rally — implied upside to mean target has compressed from the ~90%+ visible at the baseline's $126 to ~61% at $151.94, purely on price, not on estimate revisions. |
| 12 | Entry zone $110-115, do-nothing $115-170, trim/reassess $170-180 | DRIFTED — zone fired and passed | ORCL printed a weekly low of $114.50 the week of 2026-07-27 — four trading days after the baseline was written — squarely inside the prescribed add zone, then rallied +33% off that low to $151.94. The add zone worked exactly as designed. It will not be revisited without a new negative catalyst. |
| 13 | No insider-exodus red flag; insiders ~40.7%, Ellison the bulk | CARRIED, with a nuance | No new large insider selling recorded since the crash began (sensible — insiders don't sell into a rout), but also no insider buying at the $114-130 trough — nobody stepped up with size at the lows. Mild absence-of-golden-flag, not a red flag. |
| 14 | Vendor forward P/E $8.05 FY27 EPS at 15.7x (baseline, $126.29) | CONFIRMED SOUND METHOD — reused this pass | The baseline had already done the year-reconstruction correctly. This pass found fin.py/Yahoo now print forwardPE 13.91x off forwardEps $10.93 — that EPS is FY2028, one year further out than the FY2027 guide ($8.05-8.06) both reports have used throughout. See below — this is a live instance of [[pitfall-vendor-forward-eps-is-the-wrong-fiscal-year]], caught before it could flip the verdict. |
The valuation catch (headline)
get_stock_info returns both years on the same object:
epsCurrentYear : 8.05748 priceEpsCurrentYear : 18.857 <- FY2027, matches company's own $8.05 guide
epsForward : 10.92613 forwardPE : 13.906 <- FY2028, what fin.py's snapshot prints
If this session had quoted fin.py's printed PE(fwd) 13.91 at face value, the finding would have
been "ORCL got cheaper since baseline (15.7x → 13.9x)." That is false. Using the same
current-fiscal-year basis the baseline used throughout ($8.05-8.06 FY27 non-GAAP EPS), the true
multiple moved 15.7x → 18.85x — the stock got more expensive, not cheaper, because price
rose 20% on zero EPS revision. This is exactly the failure mode [[pitfall-vendor-forward-eps-is-the-wrong-fiscal-year]]
documents, and it would have inverted the verdict's direction had it not been cross-checked.
Valuation — re-derived on the same three-scenario logic as baseline (EPS unchanged at $8.06 FY27)
| Scenario | Multiple (current-FY basis) | Implied price | vs. 07-23 baseline band |
|---|---|---|---|
| Bear | 10-11x | $81-89 | ~unchanged ($75-95) |
| Base | 14-16x | $113-129 | ~unchanged ($113-135) |
| Bull (near-term) | 18.6-21x | $150-169 | ~unchanged ($150-170) |
Spot $151.94 now sits inside the bull-case bucket, not the base case it occupied at $126.29.
Nothing changed to justify that except price itself — Citi's catalyst call and a broader
AI-capex-cohort rebound, not new company evidence. Graham IV (√(22.5×5.96×13.04)) = $41.82,
essentially unchanged and still not a useful anchor for this name (negative tangible book, capex-heavy
profile) — carried at very-low weight, as in the baseline. Bogle expected return: ~1.3% yield + ~18%
guided EPS growth, entirely dependent on whether the multiple holds at 18.85x or reverts to the
14-16x base case — a ~20% air pocket either way. DYT/DDM remain non-anchors; the dividend is a
rounding error against the capital story.
Fair-value range: $90-170, spot trading at the top of it. The market is now paying for execution and backlog conversion that has not yet been proven in a single reported quarter since the baseline.
Moat — unchanged assessment, re-confirmed independently
roic.ai's own ROIC series (11.14% FY26, down from 13.72% FY25) independently reproduces the baseline's moat-erosion read without relying on the same vendor Yahoo used. The core database/apps annuity remains the durable moat; OCI/AI infrastructure remains a capacity-and-execution moat, not a structural one — see [[pattern-ai-build-inflates-earnings-while-destroying-fcf]] for the mechanism (capex/COGS destroying cash while the income statement and backlog narrative stay upbeat). Nothing in this pass changes that read in either direction.
Sentiment — three developments worth naming
- S&P's BBB- downgrade (2026-07-09) is the most consequential fact this pass surfaces that neither prior report named explicitly. It is not new information to the market (priced in by now, three weeks stale relative to the 07-23 baseline) but it materially sharpens risk #2 in both reports ("debt-funded capex / leverage spiral") — Oracle now has zero rating notches between its current grade and junk, which would trip covenants, raise financing costs, and could force a disorderly equity raise if it fires.
- Citi's 90-day catalyst watch (2026-08-26) is a genuinely new bullish data point — a reputable desk calling the crash "4-5 standard deviations," reiterating Buy/$330, betting near-term catalysts (OCI momentum, capacity announcements) re-rate the stock inside 90 days. This is the proximate driver of the last week's rally, not a fundamental print.
- Litigation progressed, not resolved — an amended complaint with named lead plaintiffs (institutional: Sparinvest, SEB Funds) signals the case has institutional backing and is not a nuisance suit likely to be dismissed quickly. Ongoing overhang, unchanged in substance.
Synthesis — Weighted Verdict
Conviction: [6.5]/10 — HOLD (down from baseline's 7.0; verdict direction unchanged, size of conviction trimmed).
The company is exactly as good and exactly as risky as it was five weeks ago — every fundamentals and moat claim above carried unchanged, because nothing has been reported since. What moved is price, 20% higher, entirely on sentiment (a bullish catalyst call, a de-risking rally in the broader AI-capex cohort) rather than on delivered evidence. That converts a name that was genuinely "at fair value, coin-flip priced" at $126 into a name now priced for a good chunk of the bull case at $152, while the two hard risk facts that most threaten the bull case — the BBB- rating sitting one step above junk, and the still-unresolved OpenAI concentration — are unchanged. Trimming conviction reflects that the market closed the discount the baseline was willing to buy into, not that the business got worse.
Key risks (carried, re-ranked by what's closest to firing): 1. 🚩 Credit — zero buffer to junk. BBB- (S&P, 2026-07-09), negative outlook (Moody's). The single most actionable new fact this pass surfaces. 2. 🚩 OpenAI/Stargate concentration — unresolved, unchanged, still roughly half the RPO. 3. 🚩 FCF/leverage trajectory — FY27 print (Sep 9-10) is the first real test of whether the burn deepens toward the baseline's −$45-50B projection or surprises better. 4. ⚠️ Litigation — amended complaint, institutional lead plaintiffs, ongoing. 5. ⚠️ Valuation has richened without new proof — the stock now needs the bull case to be right, not merely possible, to hold this level.
Thesis-break signals (carried from baseline, unchanged): OCI growth <40% for two consecutive quarters (currently +93%, far from firing) · OpenAI materially restructures/defaults on Stargate · a distressed or larger-than-guided equity raise · further credit downgrade into junk (elevated in priority this pass, given the zero-notch buffer) · BYO/prepaid mix reversing.
A portfolio-specific passage was removed from the public build.