I built a free tool that values any stock 3 ways (DCF, owner earnings, Damodaran). Roast my analysis of Oracle.
I'm a solo founder and I built DeepValues.AI — a free tool that runs intrinsic-value analysis (DCF, owner earnings, and a Damodaran-style model, cross-checked against peers), reads the SEC filings, and tracks 13F smart-money positioning. Before I keep building, I want this community to tear apart a sample analysis. So: here's Oracle, the most argued-about stock in the market right now, worked through the same three-method framework by hand.
One honesty note up front: these are my own hand calculations with all my assumptions shown — not a screenshot of the app's output. Roast the assumptions, not just the answers.
**The setup (public facts — check my inputs):**
ORCL closed Friday at **$147.61** (Sep 18, 2026). Fiscal Q1 FY2027 (reported Sep 10): revenue **$19.34B** (+30% YoY, a record, beat consensus), cloud revenue **$11.61B** (+62%), OCI infrastructure **$7.4B** (+121% — ninth straight quarter of accelerating IaaS growth), non-GAAP EPS **$1.92** (+30%, vs $1.74 consensus). Backlog (RPO) hit **$664B**, up $209B YoY, with >$30B of new AI cloud contracts booked in the quarter. The stock is down **~55% from last September's record high** (~$330), trading at ~17.3x forward earnings vs ~24x for Microsoft. And on Friday it fell ~3% intraday after BI reported the co-CEO told a town hall the company has struggled to make generative AI useful for its own staff — despite spending billions building AI infrastructure for others. Next catalyst: Oracle AI World, Las Vegas, late October.
**What my three-method workup says (my numbers, my assumptions — all below):**
| Method | My fair-value estimate | vs. Friday close ($147.61) |
|---|---|---|
| DCF (2-stage) | ~$230 | ~56% above |
| Owner earnings | ~$180 | ~22% above |
| Damodaran-style relative check | ~$162 | ~10% above |
Key assumptions: FY27E adj. EPS ~$8.10 (company guidance floor). DCF: 15% EPS growth for 5 yrs, then 8% for 5 yrs, 3% terminal growth, 10% discount rate. Owner earnings: annualized Q1 non-GAAP operating income ($8.2B × 4 = $32.8B), less ~$7B estimated maintenance capex and ~21% cash taxes → ~$6.80/share, capitalized at a 10% discount rate with 6% growth. Relative check: 20x fair forward P/E on $8.10 EPS — a deliberate discount to MSFT/AMZN multiples for the funding risk.
Moat read (mine): real moat in enterprise database — switching costs are brutal — but OCI is a scale war against deeper-pocketed hyperscalers. The moat is the installed base, not the buildout.
13F: skipping the smart-money screen in this writeup — the app tracks 29 gurus' filings; it's free to run if you want that cut.
**Where I think my own analysis might be wrong:**
1. **The backlog is concentrated.** Roughly half the order book is tied to OpenAI — and this month OpenAI's CEO joined the call to *slow down* AI development, then ruled out a 2026 IPO. My model treats the $664B RPO as contracted demand. What if the demand side is the risk, not the funding side?
2. **The funding math is brutal and my DCF barely flinches.** FY27 capex guided at **$90–95B**, Q1 free cash flow was **negative $5B**, total debt ~$167B plus ~$260B in off-balance-sheet data-center leases, S&P just cut Oracle to BBB- (one notch above junk), and management plans ~$40B in raises via debt + equity. At what point does a serious model need an explicit dilution scenario instead of smooth assumptions?
3. **Terminal value on a buildout story.** OCI is growing 121% while the company burns cash to build it. Whatever terminal growth my model assumes when the capex cycle "normalizes" basically decides the verdict — move it a point and the answer flips. Should the analysis refuse to give a single fair value here at all?
4. **The Ellison signal.** In June he entered a 10b5-1 plan to sell up to 50M shares (~$7.5B); the next day Oracle announced the plan was canceled with zero shares sold and "no other plans to sell" — no reason given — while he's pledged $40B+ backing the Paramount Skydance bid for Warner Bros. My model doesn't price insider behavior or governance. Should it?
5. **Friday's town hall.** The co-CEO reportedly admitted Oracle can't make generative AI useful for its own employees, and the CIO flagged sticker shock on OpenAI model costs — while the company spends ~$90B+/yr building AI infra for others. If the builder doesn't get ROI internally, what does that say about the durability of the $664B backlog?
**The ask:** tell me what's dumb about this. Wrong comps? Bad assumptions? Missing the actual risk? I'm not here to pitch — the tool is free, no card, and I'd rather hear "your DCF is garbage because X" now than after launch.
And if you want: drop any US ticker in the comments and I'll reply with the same three-method workup. Genuinely useful for me to see where the framework breaks.