Financial Shenanigans

Financial Shenanigans — Oracle Corporation (ORCL)

Forensic verdict: ELEVATED (score 52 / 100). Oracle's reported earnings, revenue recognition and balance-sheet line items look like a faithful picture of a real, contracted demand surge — there is no restatement, no SEC accounting action, no auditor material weakness, no going-concern language, and related-party dealing is trivial (0.02% of revenue). The risk is not fabricated numbers; it is how the cash-flow story and the funding optics are being presented as the company pours capital into AI data centers. Operating cash flow is increasingly propped up by a tripling of payables days and large customer prepayments, free cash flow has gone deeply negative, a newly-coined non-GAAP "net cash outlay for CapEx" metric softens the funding need, and a disclosed $248B of future data-center lease commitments sits off the balance sheet. A securities-fraud class action (class period June–December 2025) now alleges management downplayed exactly these cash-flow and funding consequences.

The two findings that drive the grade: (1) a working-capital-fuelled operating cash flow — days payable outstanding ran from 43 days (FY2024) to 128 days (FY2026), the single largest reason CFO looks strong while FCF is −$23.7B; and (2) escalating off-balance-sheet and metric opacity — $248B of committed leases plus a new prepayment-netted CapEx metric that understates the cash the business actually consumes. The cleanest offsetting evidence: the balance-sheet expansion is genuine, depreciable hard property and equipment ($129.6B), not inflated goodwill or soft assets, and revenue quality looks clean (days sales outstanding flat at ~51 days, related-party sales negligible). The one data point that would most change the grade: whether FY2027 operating cash flow holds without a further payables stretch — if DPO normalizes and CFO still grows, the cash-flow flag eases; if CFO growth keeps leaning on payables and prepayments, it escalates toward High.

The numbers a PM needs in ten seconds

Source: derived from Oracle FY2026 consolidated financials (income, cash flow, balance sheet) and FY2026 earnings call; computed figures.

The headline tension: a 3-year CFO/Net-Income of 1.79x reads like elite cash conversion, but the same three years produced FCF/Net-Income of −0.31x because capital spending overwhelmed it. A negative accrual ratio (CFO above net income) would normally signal conservative accounting — here it is largely the mechanical result of a $9.3B depreciation load and a payables build, not a sign of clean earnings.

The core question: are the numbers faithful?

Largely yes at the line-item level — but the presentation is being stretched. Revenue, receivables, deferred balances and the income statement reconcile to a real, contracted business; the strain shows up one layer down, in cash-flow composition, off-balance-sheet commitments, and the metrics management chooses to headline. Keep the four buckets separate: the facts (capex $55.7B, FCF −$23.7B, debt +$46B, RPO $638B) are reported and not in dispute; the accounting judgments (data-center depreciation lives, lease-commencement timing, prepayment classification) are aggressive-but-defensible; the red flags (payables lifeline, off-balance-sheet leases, flattering CapEx metric, concentration) are real and linked; and confirmed misconduct is, to date, none.

Cash flow: strong on the surface, leaning on working capital

Operating cash flow grew to a record $32.0B in FY2026, up 54% — but free cash flow was −$23.7B, the second straight year of negative FCF, as capex hit $55.7B (a 6.0x multiple of depreciation).

Source: Oracle Consolidated Statements of Cash Flows, FY2021–FY2026.

The mechanism behind "strong" CFO is visible in the working-capital accounts. Days payable outstanding nearly tripled — from 43 days in FY2024 to 128 days in FY2026 — as accounts payable rose from $2.4B to $11.0B. Much of that is data-center construction and equipment financed through supplier credit and accruals that sit in operating cash flow. The cash-conversion cycle swung to −76 days, meaning Oracle is now funding a meaningful slice of operations on its suppliers' balance sheets.

Source: derived from Oracle FY2021–FY2026 financials (receivables, payables, revenue, cost of revenue).

Two things stand out. DSO is flat at ~51 days across the period — a genuinely clean signal that revenue is not being pulled forward into receivables. But DPO is the outlier: a payables stretch of this magnitude is a non-recurring lift to CFO. On top of it, management confirmed roughly $8B of customer prepayments flattered the FY2026 cash picture and guided to $20–25B of prepayments in FY2027 — real customer cash (contract liabilities), but lumpy, deal-driven, and not a repeatable annual cash engine. Strip the payables expansion and prepayments and the underlying recurring cash generation is materially lower than the $32B headline.

Earnings quality: income statement vs balance sheet

The income statement looks resilient — operating margin actually rose to 30.6% in FY2026 — yet the cash statement and balance sheet show a business consuming cash and adding leverage. The reconciling items are depreciation (which lags the capex wave) and the payables build.

Source: derived from Oracle FY2021–FY2026 consolidated statements of operations.

Gross margin has fallen ~15 points since FY2021 (80.6% → 65.8%) as low-margin, depreciation-heavy cloud-infrastructure revenue scales — an honest, disclosed consequence of the strategy. The forensic nuance is that operating margin held up only because sales-and-marketing costs were cut ("efficiency actions") to offset the gross-margin step-down. That is a legitimate lever, but it means reported operating profitability is being defended by discretionary cost reductions while the structural gross-margin trend deteriorates.

The item to watch is depreciation policy on $129.6B of property and equipment (capex/depreciation of 6.0x). The capitalize-and-depreciate model has enormous operating leverage: small changes in assumed useful lives for servers and data-center equipment would swing reported margins materially. There is no evidence today of life-extension or under-depreciation — depreciation rose to $9.3B — but with property and equipment now the dominant asset, this is the single largest future earnings-quality sensitivity. Importantly, the asset growth is tangible — goodwill was flat at $62.3B and intangibles ran off, so soft-asset growth was negative against +17% revenue. The balance-sheet expansion is real data centers, not capitalized softness.

Source: Oracle Consolidated Balance Sheets, FY2021–FY2026.

Leverage and the off-balance-sheet question

The buildout is debt-funded. FY2026 saw $46.1B of debt issuance; total debt (including finance leases) reached $156.2B, net debt $124.3B, and net debt/EBITDA sits at ~4.2x with EBIT interest coverage of ~4.5x. None of this is hidden — but the metric management does not foreground is the $248B of additional future lease commitments disclosed in the December 11, 2025 quarterly filing (substantially all data-center and cloud-capacity arrangements, commencing FY2026–FY2028 on 15-to-19-year terms). Because those leases have not yet commenced, they are off the balance sheet today; reported leverage understates the committed obligation.

Source: Oracle FY2021–FY2026 balance sheets and ratio file; lease-commitment figure per Q2 FY2026 10-Q (December 2025).

Metric hygiene: a new yardstick and a giant backlog

Two presentation choices warrant scrutiny. First, in FY2026 management introduced "net cash outlay for capital expenditures" — reported capex minus customer prepayments and timing. It pegged FY2026 at $48B versus the $55.7B cash-flow-statement figure, and guided FY2027 to "around $70B net cash outlay" while acknowledging reported capex will be $20–25B higher (i.e., ~$90–95B). A new, non-standard metric that nets customer cash against capex to shrink the apparent funding need is exactly the kind of definition to monitor.

Second, remaining performance obligations of $638B (+363%) are headlined as "exceptional visibility." The backlog is contractually real, but the company disclosed that only 12% converts within 12 months and 34% in 13–36 months — so the bulk is long-dated, and Standard and Poor’s flagged concentration risk tied to OpenAI in September 2025. A backlog metric this large, this concentrated, and this back-end-loaded is a genuine visibility signal and a metric whose headline growth rate outruns its near-term cash realization.

Source: Oracle Q4 FY2026 earnings call and GAAP-to-non-GAAP reconciliation commentary.

The non-GAAP EPS gap is wide: $7.63 reported vs $5.83 GAAP (+30.9%). Management itself disclosed that ~$0.80 of that is one-time investment gains (Ampere and Bloom Energy) — to its credit, quantified openly — leaving a cleaner $6.83. The honest read of "underlying" non-GAAP EPS growth is therefore mid-teens, not the +27% headline. Non-GAAP also excludes $4.8B of stock-based compensation (7% of revenue), standard for the sector but material. The Ampere gain is also worth flagging because Ampere is a related party (a former Oracle director chairs it; Oracle holds ~29% and $1.6B of equity/convertible debt).

Breeding ground: amplifies the flags

The structural conditions tilt toward amplifying, not dampening, the accounting risk. Larry Ellison remains the controlling shareholder (~40%+, founder, Executive Chair and CTO); the board is majority non-independent (five of ten directors are insiders); and financial leadership turned over at the worst possible moment of accounting complexity — Safra Catz moved from CEO to Vice Chair in September 2025, two co-CEOs were installed, a new Chief Accounting Officer is in place, and the CFO presenting the FY2026 results stated she had "only been here for two weeks." Auditor tenure (Ernst and Young) is multi-decade. Oracle also carries a compliance history (SEC FCPA settlements in 2012 and 2023) and now the live securities-fraud class action. Related-party transactions themselves are immaterial (0.02% of revenue, Independence-Committee-governed, with an Ellison price-protection agreement) — that specific test is clean — but the combination of concentrated control, insider-heavy board, and fresh finance leadership during a debt-funded, prepayment-reliant, off-balance-sheet-heavy expansion raises the bar for trust in management's presentation choices.

The 13-category shenanigans scorecard

Source: Oracle FY2021–FY2026 financials, Q4 FY2026 earnings call, Q2 FY2026 10-Q, and FY2025 proxy; severities assigned by this analysis.

Two reds (CF4 working-capital lifeline; EM5 off-balance-sheet leases), six yellows, five clean. The reds and the KM/CF yellows are linked — they all trace to one root cause: a debt- and prepayment-funded capex wave that the cash statement and headline metrics present more favorably than the underlying recurring economics support.

What to underwrite next

Five things to track, in priority order:

1. DPO and operating cash-flow composition (red, CF4). Watch days-payable in each FY2027 quarter. Downgrades the grade: DPO stabilizes near ~100 days or lower and CFO still grows — proof the cash engine is durable. Upgrades it: DPO keeps climbing or CFO growth visibly depends on payables/prepayment timing.

2. The $248B lease commitments as they commence (red, EM5). Track how much moves on-balance-sheet each quarter and whether matched cloud revenue arrives on schedule. Disproves the concern: leases capitalize on commencement against ramped, contracted revenue. Confirms it: commitments grow faster than revenue conversion, or capacity sits idle.

3. Depreciation useful-life disclosures (yellow, EM4). Read the property and equipment note in the FY2026 10-K for any extension of server/data-center lives. Any lengthening that flatters gross margin is a red-flag escalation.

4. RPO conversion and customer concentration (yellow, KM1). Monitor the "12% within 12 months" conversion rate and any disclosure on top-customer concentration (OpenAI/AI-infrastructure counterparties) and their ability to pay over 15-to-19-year terms.

5. The "net cash outlay for CapEx" metric and reported capex gap (yellow, KM1/KM2). Confirm FY2027 reported capex lands near the ~$90–95B implied figure, not the $70B "net" headline, and that prepayments actually arrive.

Bottom line for position sizing. This accounting risk is a valuation haircut and a position-sizing limiter, not (yet) a thesis breaker. The reported earnings and revenue appear faithful, the demand is contracted, and the asset base is real and tangible — but a buy-side underwriter should discount the headline $32B operating cash flow for its payables/prepayment content, add the $248B of committed leases to the leverage picture, treat non-GAAP EPS growth as mid-teens rather than +27%, and size the position to survive the scenario the class action describes: capex and committed obligations that compound faster than cash returns if the AI-infrastructure ramp slips. Demand the proof in FY2027 cash flow before paying for the backlog.