Oracle — $ORCL

The thesis

Oracle is no longer a sleepy mature software company. It has become a hybrid: a sticky, recurring software platform (more than 70% of revenue recurring or renewable) bolted to one of the most aggressive AI-infrastructure buildouts in the market. Remaining performance obligations stand at $638B, up 363% — roughly 9.5× annual revenue already under contract. That is real demand, not narrative. But the stock has been punished — a slide from roughly $251 to $140 — because the market has started asking what this growth costs. S&P cut the credit rating to BBB-, one notch above junk, and the backlog is concentrated in a handful of AI counterparties: if a customer renegotiates or can't find compute, a chunk of that $638B gets less bankable fast. This is now an execution story, a financing story, and a capital-allocation story all at once.

The growth the market is discounting

The gap between the two CAGRs is the Oracle story — revenue rips while FY27 earnings absorb the buildout's bill, then EPS catches up in FY28.

Margins — the part we don't like

Falling margins during a hype cycle are always more dangerous than they look. (Estimates per J.P. Morgan.)

Balance sheet

Strong business, stressed balance sheet — BBB- is the agencies saying the margin for error has narrowed. D/E improves in FY27 as the equity raise lands, then deteriorates again as debt outruns retained earnings. Interest expense estimates are ours, off Q4 FY26's $1.44B (+47% YoY) run-rate and rising debt.

Cash flow

CFO minus capex — the FCF walk. Investing cash flow is almost entirely capex during the buildout; the single biggest issue with the bull case is contracted future revenue with zero near-term FCF support. (Estimates per J.P. Morgan.)

Our valuation — scenario analysis

Skew at $141.50: roughly $22 of modeled downside vs $20 base / $79 bull upside.

Model risk

This framework is only as good as its two inputs, and both carry unusual uncertainty. The EPS anchor: our drift band may be too kind — consensus leans on backlog conversion happening on schedule; if a major AI counterparty renegotiates or defers, FY27 EPS doesn't miss by 4%, it misses by far more, and the bear floor moves below $119 with it. Non-GAAP earnings also flatter the picture while free cash flow runs deeply negative: when earnings and cash diverge this widely, the multiple the market will pay becomes less predictable, not more. And the multiple itself: every 1× of P/E is worth about $8 on the stock, so the honest range around any scenario is wide. These scenarios are brackets, not probabilities — treat the table as a map of the debate, not a prediction of the destination.

Where the Street sits

The debate is entirely about what Oracle pays to capture it. Analyst targets attributed as publicly reported.

Technical read — the level map

Potential double/triple bottom at 17-year support, below all major moving averages — we're watching for a false breakdown below $140 that scares out weak hands, then reverses.

Beta — the stock changed character

The AI pivot didn't just change the business — it changed the stock, from mature software (~1.7) to a leveraged AI trade (~2.8). R² of 0.35 means most of the movement is company-specific. OptionFlowTracker calculation, monthly total returns vs the S&P 500.

The tape is leaning bullish at the lows.

On July 9, clustered ask-side buying hit the ORCL July $143 calls — $2.39M in one block, more behind it — and the tape confirmed underneath, pushing $144 → $148 intraday. Confirmed the next morning, before the open, as genuine new positioning. July 10 continued the pattern: a block of July $140 calls at the ask, right at the money, opening well beyond anything already standing in the strike and confirmed the next morning — two for two. Short-dated and near the money — a bet on the move landing soon, not eventually. Ask-side, opening, near-the-money call blocks on back-to-back sessions at 17-year support is precisely the confluence our framework looks for. Research and education, not financial advice.

Insiders

Neutral to mildly negative. No clustered open-market buys signaling conviction, but no discretionary panic-selling to read as bearish either.

Key risks

Backlog concentration in a few AI counterparties — if a customer renegotiates or can't find compute, a chunk of the $638B gets less bankable fast. A funding plan with no room for a demand air-pocket (long-term debt toward ~$165B by FY28E, BBB-, negative free cash flow throughout). Margin compression to ~55% gross. Negative working capital. A rising beta in a market re-rating AI-infrastructure multiples. And a weekly close below $135–140 opens the door to $120.

Every Friday's weekly close vs the $135–140 line · Q1 FY27 earnings (September) · AI-counterparty headlines · morning-after confirmation on new flow prints.