Deck
Oracle Corporation · ORCL · NYSE
Oracle is a 48-year-old enterprise-software company that sells database licenses and support, cloud applications, and — now its growth engine — rented AI compute through Oracle Cloud Infrastructure.
$184
Price
Jun 18, 2026
~$517B
Market cap
$67.4B
FY2026 revenue
+17% YoY
~162,000
Employees
Public since 1986. Riding the AI-infrastructure boom, ORCL peaked at $345.72 in September 2025, then de-rated roughly 46% to $184.29 by June 2026 — a violent round trip that swung from $138 to $250 in the spring of 2026 alone.
2 · The case
Oracle is two companies sharing one balance sheet.
- The floor — a 48-year database annuity. Roughly $19.5B of near-100%-margin license support, where "substantially all" customers renew on expiration, sits inside a 63%-margin cloud-and-license business — a wide-moat cash machine that funds everything above it.
- The bet — rented AI compute. Management is routing all incremental capital into Oracle Cloud Infrastructure, a layer Oracle's own 10-K calls a business with "low barriers to entry," now funded by debt and equity after free cash flow turned negative.
- The bridge — a $638B backlog. Contracted remaining performance obligations ended FY2026 at $638B, up 363% year-over-year, roughly nine times annual revenue. The whole case reduces to one question: does that backlog convert to cash at a return above its ~$70–90B-a-year capex?
The annuity is proven; the bet is not. Almost everything the market pays a premium for sits in the unproven half.
3 · The money picture
The cash machine inverted into a construction hole.
$32B
Operating cash flow
record, +54% YoY
$48B
Net capex outlay
~7x FY2024
-$23.7B
Free cash flow
from +$11.8B in FY2024
~$125B
Total debt
from $92.6B; net-cash a few years ago
Record $32B of operating cash flow was dwarfed by $48B of net capex, swinging free cash flow deeply negative for the first time in the modern era. FY2027 capex is guided to ~$70B, funded by ~$40B of fresh debt and equity — including a $20B at-the-market stock issuance — from a company that spent the prior decade retiring a third of its shares.
4 · The central tension
Is the $638B backlog durable revenue — or concentrated credit?
- Read one: contracted visibility. The backlog is audited, signed demand anchored by "the who's who of AI" — OpenAI, xAI, Meta, NVIDIA, AMD — forward visibility no enterprise-IT peer can match, with a multicloud database that grew 404% as the lock-in travels into rival clouds.
- Read two: a few cash-burning counterparties. Analysts estimate roughly half the backlog sits with one customer, OpenAI, whose ~$25B of revenue is dwarfed by a ~$60B-a-year compute commitment — and Oracle carries no disclosed credit reserve against it.
- The clock runs slow. Only 12% of the $638B converts to revenue within twelve months, so most of the capital must be spent years before the cash arrives. One renegotiation, delay, or first reserve re-prices the whole story.
Contracted is not collected. The same fact is the bull case and the bear case, read in opposite directions.
5 · Where the valuation is decided
The OCI margin — not revenue — is the whole valuation.
- Consensus capitalizes 30–40%. The Street's ~$253 mean target and ~$11 FY2028 EPS are built on management's target of a 30-to-40% OCI margin "over the life of a customer contract" — a contractual goal, not a reported result.
- The margin is moving the other way. Consolidated gross margin stepped down about 5 points during the FY2026 ramp, and management has pre-committed to a further step-down in FY2027 — while leaked figures put early Nvidia-rental margins in the mid-teens.
- Reported profit has a tailwind to unwind. Extending server useful lives from five to six years lifted FY2025 net income by $573M — Oracle is lengthening depreciation precisely as AI peers shorten it for faster obsolescence.
If steady-state OCI margin settles in the mid-20s rather than the low-30s, FY2028 EPS lands ~10–15% below consensus — and the price is capitalizing the wrong earnings.
6 · What the tape says
Records keep printing; the multiple keeps falling.
- Beats now meet selling. A year ago a record-backlog print produced Oracle's best day since 1992; today the same RPO records draw sellers. The market stopped paying for bookings and started demanding proof the backlog converts to cash.
- Not a crowded short — a de-risking long. No reported short interest is staged, ~27.5M shares trade daily, and Ellison's 40.6% stake leaves a ~1.65B float. The amplifier isn't a squeeze; it's a long base still cutting, so a downside surprise lands on holders already selling.
- Two scheduled reads break the standoff. The Q1 FY2027 print on September 10, 2026 and the October 28 Analyst Day are the first two looks at the OCI margin path — the single variable on which the cash drain, the leverage, and the multiple all resolve at once.
Against zero Sell ratings, a clean low-30s margin print could snap the de-rate shut as fast as it opened.
7 · The two-sided picture
A great franchise making a great bet — not yet a proven one.
- What supports it. A wide-moat database annuity where "substantially all" customers renew funds the build from below; the $638B backlog is booked and audited, not forecast; and founder Larry Ellison's 40.6% stake — one share, one vote — keeps incentives with owners.
- What cuts against it. The cash that proves conversion hasn't arrived: FY2026 free cash flow was -$23.7B against ~$125B of debt, OCI margin is guided lower two years running, the backlog leans on a few cash-burning AI labs with no reserve, and Ellison has pledged 346M shares against personal loans — a forced-selling tail risk in exactly the drawdown the bears fear.
- The call on the desk. The report lands on lean long, wait for confirmation — the contracted backlog and durable annuity tilt the case long, but conviction is earned only once the cash and margin actually print.
Watchlist to re-rate: OCI gross margin at the September 10 print (does it trough in the low-30s or slide toward the mid-teens); whether RPO keeps rising with no credit reserve at OpenAI or xAI; and operating cash flow visibly narrowing the gap to the ~$70B FY2027 capex with the investment-grade rating intact.