Financials
Financials — Oracle Corporation (ORCL)
Oracle is no longer the financial business it was three years ago. A 40-year-old, cash-gushing database and applications company — the kind that bought back a third of its stock and ran on capex below 2% of revenue — has, in the space of two fiscal years, turned itself into one of the most capital-intensive infrastructure builders on the planet. In FY2026 Oracle spent $55.7B of capex against $32.0B of operating cash flow, producing free cash flow of negative $23.7B, and funded the gap with $46B of new debt. It did so against a contracted backlog that exploded to $638B.
So the financials no longer answer "is this a high-quality software compounder?" — they plainly were. They now answer a harder question: is Oracle building a generational AI-infrastructure annuity, or pouring borrowed cash into a capital sink before the revenue and cash conversion arrive? Everything below is organized around that single tension.
FY2026 Revenue ($M)
Operating Margin
Net Income ($M)
Free Cash Flow ($M)
Source: FY2026 results (fiscal year ended 31 May 2026), reported 10 June 2026; figures in USD.
The 30-second read. Profitable and accelerating at the top line (revenue +17%, net income +37%), with a $638B contracted backlog that gives rare multi-year visibility — but free cash flow is deeply negative, gross debt is approaching $130B, both Moody's and S&P carry a negative outlook, and the stock has already round-tripped from $328 to $184. The whole investment case rests on whether the AI backlog converts to cash before the balance sheet runs out of room. This is a credit-and-conversion story wearing a software multiple.
1. The transformation in two charts
The first chart shows revenue and operating profit re-accelerating after a decade of single-digit growth; the second shows the capital being thrown at it.
Source: Consolidated statements of operations, FY2017–FY2026; derived from reported financials.
Revenue grew at barely 1–2% a year from FY2014 to FY2021 — Oracle was a mature, ex-growth franchise. The Cerner acquisition (FY2023) added a step-up, but the real inflection is FY2025–FY2026: revenue +17% to $67.4B, with management guiding FY2027 to roughly $90B (+34%). That re-acceleration is driven almost entirely by Oracle Cloud Infrastructure (OCI), where cloud-infrastructure revenue has been compounding 50–90%+ per quarter as Oracle rents GPU capacity to AI customers.
Source: Consolidated statements of cash flows, FY2017–FY2026; derived from reported financials.
For most of the past decade Oracle's capex was a rounding error — under $2.5B a year, well under operating cash flow, leaving a torrent of free cash. In FY2025 capex jumped to $21.2B and overtook the growth in cash flow; in FY2026 it reached $55.7B — 83% of revenue and 174% of operating cash flow. A software company does not normally do this. An electric utility or a telecom does. Oracle is now spending like infrastructure and being valued like software, and that gap is the central debate.
2. The backlog — why management is spending
The justification for the capex flood sits in one number: remaining performance obligations (RPO) — the dollar value of contracts signed but not yet recognized as revenue. It is the closest thing Oracle has to a forward order book.
Source: Company RPO disclosures, MD&A and quarterly releases through FY2026; per the latest filings.
RPO ended FY2026 at $638B, up 363% year-over-year — roughly 9.5 times annual revenue. For context, that is more contracted future business than most software companies will recognize in a decade. It is the bull case in a single figure: extraordinary visibility, and a logical reason to build data centers ahead of demand.
But the same number carries the bear case. A large share of the backlog — by external estimates $300B+ of a ~$500B+ pool — is tied to a single counterparty, OpenAI, with the balance concentrated in a handful of other large AI buyers. RPO is contracted revenue, not delivered or collected revenue; it assumes those customers take the capacity, pay on schedule, and remain solvent across a multi-year build. The backlog's quality depends entirely on the creditworthiness and staying power of a very small number of customers, and that is not visible in any of the statements.
Define it once — RPO (remaining performance obligation): the total value of contracted revenue a company has signed but not yet recognized. High RPO signals demand visibility; it is a booking, not cash, and converts to revenue only as the service is delivered.
3. Earnings quality — profits are real, cash is not (yet)
Oracle's reported profitability is genuine and improving: net income rose 37% to $17.0B in FY2026, and operating cash flow of $32.0B comfortably exceeds net income (a 1.9x cash-to-earnings ratio — healthy, helped by the deferred revenue that comes with prepaid cloud contracts). The earnings-quality problem is not the income statement. It is what happens after capex.
Source: Consolidated statements of cash flows; free cash flow = operating cash flow less capex.
Free cash flow — the cash left for debt holders and shareholders after the business reinvests — went from a dependable $12–14B a year through FY2021, to negative $0.4B in FY2025, to negative $23.7B in FY2026. That is the single most important change in these financials. It is not a sign of deteriorating operations; it is a deliberate choice to pull forward years of data-center spending. But the distinction matters less to a lender than to a storyteller: for now, Oracle's growth consumes cash rather than generating it.
Two earnings-quality nuances a careful reader should hold:
Capitalized interest flatters operating profit. FY2026 pre-tax income ($19.6B) sits just below operating income ($20.6B), implying net non-operating expense of only ~$1B despite ~$130B of debt. With gross interest on that debt likely in the $5–6B range, a large portion is being capitalized into the cost of data-center assets under construction rather than expensed — and is offset by interest earned on cash. Reported interest cost will rise materially as those assets are placed in service and capitalization stops.
Depreciation is about to scale fast. PP&E quintupled to $129.6B in two years; depreciation has only begun to catch up (D&A $9.3B in FY2026). As that asset base depreciates, it will weigh on gross margin and reported earnings — gross margin already fell from 80% (pre-cloud) to 65.8% as infrastructure costs entered cost of revenue. The margin profile of "Oracle the infrastructure landlord" is structurally lower than "Oracle the software licensor."
4. Balance sheet — the binding constraint
For a software company, the balance sheet is usually an afterthought. For Oracle today it is the governing variable, because the capex is funded with debt, not cash.
Source: Consolidated balance sheets, FY2019–FY2026; total debt = short-term plus long-term borrowings.
Gross debt jumped to $129.5B in FY2026 (from $92.6B), and net debt to $97.6B. Against EBITDA of ~$29.9B, that is net debt / EBITDA of ~3.3x (gross ~4.3x) — elevated for a technology company and high enough that Moody's and S&P have both placed Oracle's credit on negative outlook, signaling possible downgrades. Interest coverage (EBIT / interest) has compressed from double digits a decade ago to roughly 4–5x today, and will tighten further as capitalized interest rolls into the P&L.
A second feature worth understanding: Oracle's equity is artificially thin because of history, not weakness. A decade of aggressive buybacks (including $36.6B in FY2019 alone) drove retained earnings deeply negative and shareholders' equity to below zero in FY2022. Equity has since rebuilt to $43.1B as buybacks stopped and profits accumulated. The takeaway: conventional ROE and book-value ratios are distorted and near-meaningless for Oracle — judge returns on invested capital and leverage on EBITDA, not on the equity line.
The balance-sheet swing risk. Oracle is funding a multi-year, $50B+/year build largely with debt, into a backlog concentrated in a few AI customers, while two rating agencies signal downgrade risk. If demand or financing tightens, there is — in one analyst's words — "very little room for error." This is the risk a software multiple does not normally carry.
5. Margins and segment mix — software economics meeting infrastructure costs
Source: Derived from reported income statements, FY2019–FY2026.
The mix shift is visible in the margin lines. Gross margin has fallen from ~80% to 65.8% as low-margin infrastructure revenue (and the depreciation of GPUs and data centers) replaces high-margin license revenue in the cost of sales. Operating margin has held near 30% only because Oracle has tightly controlled operating expense — but the gravitational pull is downward as the infrastructure business scales. Revenue mix tells the same story: cloud services and license support is now ~77% of revenue ($44B of $57B in FY2025), and within it the fast-growing piece is OCI, the lowest-margin, most capital-hungry part of the portfolio.
This is the trade Oracle has accepted: lower-margin, capital-intensive revenue today in exchange for very high growth and a vast backlog. For an investor, the question is whether OCI eventually achieves the scale and utilization to earn an acceptable return on the data centers — or whether margins keep grinding lower while the asset base balloons.
6. Returns on capital and capital allocation
Source: Derived from reported financials; NOPAT = operating income × (1 − effective tax rate).
Define it once — ROIC (return on invested capital): after-tax operating profit divided by the debt-plus-equity put into the business. It measures how efficiently a company turns capital into profit, independent of how it is financed. A ROIC durably above the cost of capital (~8–10% here) creates value; below it destroys value.
Oracle's ROIC ran a healthy 13–18% through the cloud build-out — but fell to ~10.4% in FY2026 and is mechanically headed lower in the near term, because invested capital (debt + equity) is ballooning years ahead of the revenue and profit those assets will eventually generate. This is the normal optics of any front-loaded capex cycle; whether it is value-creating or value-destroying depends entirely on the incremental return those data centers earn once filled. That number is not yet observable — which is exactly why the backlog and cash-conversion metrics matter so much.
Capital allocation has flipped 180 degrees. Oracle was historically a buyback machine — it retired roughly 40% of its shares between FY2013 and FY2022, shrinking the diluted count from ~5.0B to ~2.7B. In FY2026 buybacks essentially stopped ($206M, versus $17–36B in peak years); every spare dollar — and many borrowed ones — now goes to capex. The dividend ($5.8B, ~$2.02/share, ~0.9% yield) continues, but with free cash flow negative, the dividend is currently funded by debt, not cash flow. Share count has even ticked up slightly as stock compensation dilution is no longer offset by repurchases.
Source: Consolidated statements of cash flows; capex shown as cash outflow magnitude.
7. Valuation — a software multiple on an infrastructure balance sheet
At ~$184 (mid-June 2026, down sharply from a $328 peak), Oracle carries a market capitalization near $530B and an enterprise value around $628B once net debt is added. On trailing numbers that is roughly 31x earnings, 7.9x sales, and ~21x EV/EBITDA — a clear premium to mature peers like IBM, SAP and Salesforce, and even rich versus Microsoft on a sales basis.
Source: peer market data as of mid-June 2026; fundamentals from latest reported fiscal years. ORCL revenue growth shown trailing; management guides FY2027 to ~+34%.
The valuation tension is best framed this way. On trailing metrics, Oracle looks expensive for a business with lower margins, far higher leverage, and negative free cash flow than the peer set. On forward expectations, it can look reasonable: if FY2027 revenue reaches the guided ~$90B and the backlog converts, today's enterprise value falls to roughly 7x forward sales on a business growing 30%+. The multiple is not pricing the present — it is pricing the successful conversion of a $638B backlog into high-margin, cash-generative revenue. That is a coherent bet, but it is a bet, and the ~44% drawdown from the peak shows the market repricing the risk, not just the growth.
There is no standard "Fair Value" or Quality Score available in the provided data for Oracle, so this page relies on peer-relative and history-relative anchors rather than a single model price. Sell-side consensus sits around a $252–268 target (implying ~35–45% upside from spot), but the dispersion is unusually wide — from sub-$240 cuts citing balance-sheet risk to $400 bull cases — which itself signals how binary the outcome is.
8. The year-wise financials
Percentages are stored as decimals.
Source: Consolidated financial statements FY2017–FY2026 (figures in $M except EPS and ratios); ROIC and net debt derived from reported financials. Note: FY2018 net income depressed by a one-time ~$8.8B tax-reform charge; FY2022 equity negative due to cumulative buybacks.
The table compresses the whole arc: a slow-growth, ultra-high-margin, cash-rich, share-shrinking franchise through FY2021 — then the Cerner step-up (FY2023) — then the AI capex inflection (FY2025–FY2026) where free cash flow turns sharply negative and debt steps up by $37B in a single year even as profits hit a record.
9. What the financials confirm, contradict, and what to watch
What they confirm. Demand is real and the franchise is accelerating: record revenue, record net income, OCI compounding at extraordinary rates, and a $638B backlog that few companies in history have matched. Oracle's core software economics remain excellent, and operating cash flow of $32B shows the underlying engine still converts profit to cash before reinvestment.
What they contradict. The "safe, cash-rich software compounder" framing is gone. Free cash flow is negative $23.7B, gross debt is ~$130B with a negative credit outlook, gross margin is sliding toward infrastructure levels, ROIC is falling, buybacks have stopped, and the dividend is now debt-funded. The valuation still prices software-grade certainty onto a business that has taken on utility-grade capital intensity and single-customer concentration.
The one fact the case turns on. Oracle has already booked the demand (RPO) and already committed the capital (capex). The only thing not yet proven is the bridge between them — whether that backlog converts into delivered revenue and, crucially, into cash, fast enough to service the debt before the rating agencies or the AI customers force the issue. Bulls see a $638B annuity being built ahead of the curve; bears see $50B+ a year of borrowed capex chasing a backlog that leans on one or two counterparties.
The first financial metric to watch is free cash flow conversion — specifically the trajectory of capex as a share of operating cash flow, and the point at which it inflects. FY2026 ran capex at 174% of operating cash flow. The bull thesis requires that ratio to peak and fall as OCI capacity fills and revenue catches up; if capex stays above operating cash flow for several more years while debt and depreciation climb, the AI build becomes a balance-sheet problem regardless of how large the backlog looks. RPO tells you the demand is contracted; free cash flow conversion tells you whether Oracle can actually afford to deliver it. Watch it every quarter.