Financial Shenanigans

Financial Shenanigans

The Forensic Verdict

Nu Holdings sits in the Watch band on the forensic risk scale. Clean cash-conversion record, effective internal-control attestation, and improving credit metrics on one side; founder-controlled dual-class governance, opaquely disclosed compensation budget, in-flight "Managerial P&L" framework rollout, and visible intangible-asset capitalization on the other. Nothing in the public record points to revenue manipulation, restatement, or auditor concern. The two findings worth underwriting: (1) governance concentration that limits investor checks on judgment-laden accounting choices and (2) the Q4 2025 disclosure-framework reset — both raise the cost of trusting reported numbers without external corroboration. The single data point that would most change the grade is any future material weakness, qualification, or KPI definition change that breaks comparability with FY2024 financials.

Shenanigans scorecard — all 13 categories

Breeding Ground

The conditions that make accounting shenanigans more likely are present, but in a benign rather than predatory configuration. Concentration of power is the dominant signal: founder David Vélez Osorno is simultaneously CEO and Chairman, controls 88.3% of Class B shares (20 votes per share) for 74.3% of voting power, and the Shareholder's Agreement gives him a veto over related-party transactions, indebtedness above net equity, and other material actions. Nu is a "controlled company" under NYSE rules and is exempt from the majority-independent-board and fully-independent-comp-committee requirements. As a Cayman Islands foreign private issuer, individual NEO compensation is not disclosed; investors only see the aggregate $91.3M figure for FY2025 covering all directors and key management personnel. Form 4 insider filings are not required for FPI directors and officers.

The offsetting governance evidence is meaningful: an independent Lead Director (Anita Sands), an Audit & Risk Committee chaired by Rogério Calderón — a former PwC Brazil audit partner with a decade in that role — and seven of nine directors classified independent. The board added Roberto Campos Neto (former Banco Central do Brasil President) and CTO Eric Young in 2025. A compensation clawback policy was adopted in October 2023, and the FY2025 filing states no restatement-triggered recovery has been required.

The structure is not predatory — Vélez is a long-tenured founder with reputational skin in the game and a Sequoia-led board with operating history — but the machinery for catching aggressive accounting choices runs through fewer hands than it would at a U.S. domestic, single-class, fully-independent board. That is the right frame for breeding-ground risk on this name.

Earnings Quality

Reported earnings look earned, not manufactured — the qualifications are about classification, not creation. Net income reconciles cleanly to the cash flow statement, revenue is overwhelmingly interest income (79% of total in FY25) and interchange fees (recognized at point of transaction), and there is no material one-time-gain reliance. Revenue growth is decelerating naturally with scale: +144% (FY22) → +85% (FY23) → +45% (FY24) → +29% (FY25). Operating leverage is real: total non-interest expense as a share of revenue dropped from 67% (FY22) to 28% (FY25).

The one judgment-zone test that should make a forensic reader pause is the balance-sheet treatment of operating costs. Per the FY2025 MD&A, capital expenditures (cash payments for PP&E or intangible assets) were $340.8M in FY25 versus $175.0M in FY24 — nearly doubling. The cash-flow statement's "capex" line of $7.2M captures only property, plant and equipment; the difference is software, customer-onboarding, and product-development costs that get capitalized as intangibles. Intangibles on the balance sheet grew from $295.9M (FY23) to $601.7M (FY25), a 103% increase against 87% revenue growth. Depreciation and amortization is $98M FY25 — well below the gross capitalization pace.

Practical read: software and product-development costs that other digital banks might expense are sitting on the balance sheet and amortizing slowly. None of this is improper under IFRS, but the gap between cash capex ($340.8M) and D&A ($98M) means roughly $243M of FY25 development spend was capitalized rather than expensed — a benefit to current-period operating profit of similar order. If growth slows or product launches are written down, this becomes a future impairment risk.

Reserve and provisioning behavior is the other forensic-quality question. The company applies an expected-credit-loss (ECL) model that front-loads provisioning at loan origination. ECL as a share of total receivables held steady at 15.4% across FY24 and FY25, even as the loan book grew 58%. NPL 90+ in Brazil improved from 7.0% (FY24) to 6.6% (FY25); NPL 15-90 improved from 4.1% to 3.9%. Provisions grew +33% against +58% portfolio growth — the model's "front-loading" claim is doing some of the lifting here, but the credit data trend supports the lower marginal provisioning rate.

The effective tax rate trajectory is the third earnings-quality item to track. ETR fell from 33.0% (FY23) to 29.4% (FY24) to 25.8% (FY25), versus a Brazilian statutory rate of 40%. Q1 2026 came in at 8.7% — a $82.9M tax expense on $954.3M of pretax income. That single-quarter rate is almost certainly driven by Brazilian "juros sobre capital próprio" (interest on equity) deductions or deferred-tax-asset recognition rather than ongoing tax mix; it is not a sustainable run-rate. Investors valuing forward EPS off Q1 2026 risk anchoring on a transitory tax benefit.

Cash Flow Quality

CFO is durable but is essentially a deposit-funded operating book, and that single fact governs how to read the next four years of cash conversion. Three-year aggregate CFO of $7.17B versus net income of $5.87B gives a CFO/NI ratio of 1.22 — clean by industrial standards but mechanically driven by the bank business model: deposit growth shows up as an operating cash inflow, and loan growth shows up as an operating cash outflow. Both are large. In FY25 alone, deposits grew $13.0B, the loan book grew $4.1B, and credit card balances grew $7.1B.

The "free cash flow" line on the cash flow statement ($3,493M in FY25) is misleading on its own because it subtracts only PP&E capex of $7.2M, ignoring the $340M of intangible-asset additions that the MD&A includes in its capital expenditure definition. Re-cast FCF using the MD&A capex definition gives roughly $3,160M for FY25 — still strong, but $333M lower than the headline.

The accrual ratio is negative at -1.0% of average assets in FY25 — net income is smaller than CFO by roughly 1% of the asset base, which is the opposite signal from a typical earnings-management concern. Working-capital quality is also clean in the sense that there is no factoring, no off-balance-sheet receivables sale, and no supplier finance program disclosed. The cash flow statement reconciliation discloses ECL of $4,701.1M as an add-back (versus $4,204.9M on the income statement) — the $496M difference reflects FX translation and other movements through the credit-loss line, not a hidden expense.

Acquisitions are immaterial: $1.5M in FY25 and $5.6M in FY24. Hyperplane (Jul 2024) and the Tyme minority stake (Dec 2024) do not move the cash flow story. Goodwill held flat at roughly $410M across FY23-FY25, which is what you want from a non-acquisitive compounder — no purchase-accounting tailwind to revenue or working capital.

Metric Hygiene

The metric-hygiene risks are the most concrete forensic issues for this name, and they cluster around three items.

The "Managerial P&L" change is the most consequential single disclosure event of FY2025 from a forensic perspective. Per the FY2025 20-F, "in the beginning of the fourth quarter of 2025, we introduced the Managerial P&L, representing an evolution in our disclosure framework… a structural, complementary reorganization of line items designed to enhance comparability as the business scales." Management states the change "preserves net income, cash flow, and capital." A separate Q1 2026 6-K confirms an independent firm provided limited assurance under ISAE 3000 over the Managerial P&L reconciliation report — not a full audit. Limited assurance is appropriate for management-defined metrics, but investors should not equate it with audited financial statements.

The persistent $80-135M gap between SBC as reported in the cash flow statement and SBC added back in the Adjusted Net Income reconciliation deserves a diligence question to investor relations. It does not appear to be improper — the gap likely reflects different scopes (e.g., inclusion of social-charge accruals or different vesting treatment) — but the company's filings do not reconcile the two numbers, and a reader cannot derive the difference from the extracted disclosures.

What to Underwrite Next

The forensic risk here is a margin-of-safety adjustment, not a thesis breaker. The accounting is conservative on most tests, governance is concentrated but not yet litigious, and there are no smoking guns. Position sizing should respect the limits of disclosure; valuation should not anchor on Adjusted NI or Q1 2026 tax-rate normalization.

The five items to track next quarter and into the FY2026 20-F:

1. Managerial P&L re-presentation. Confirm that FY2024 and FY2023 financials are re-presented to the new framework in the FY2026 20-F so analysts can build like-for-like trend lines below revenue. If only forward periods are presented, comparability is broken.
2. Effective tax rate normalization. Watch FY2026 quarterly tax expense relative to pretax. A run-rate of 22-27% is plausible given Brazilian interest-on-equity deductions and international mix; a sustained sub-15% rate would imply a structural shift that needs explanation in the next 20-F notes.
3. Intangible-asset capitalization vs amortization. Track quarterly intangible additions, D&A, and any impairment charges. The MD&A capex figure ($340.8M FY25) should grow no faster than revenue; sustained outpacing implies a thicker future amortization wedge or impairment risk.
4. ECL coverage if loan growth slows. With the model front-loading provisioning at origination, a sharp deceleration in new originations (the FGTS regulatory cliff from November 2025 is a near-term test) should mechanically reduce provisioning expense even if late-stage NPLs deteriorate. Watch the NPL 90+ trend and Stage 3 coverage in the FY2026 audited financials.
5. Any change to non-IFRS definitions. Adjusted NI currently adds back SBC after tax. Adding new line items (e.g., excluding "transformation costs", "AI investment", or international-expansion losses) is a classic earnings-management pattern. Watch the FY2026 reconciliation for new add-backs.

The grade would downgrade to Elevated (41-60) if: (a) FY2024/2023 financials are not re-presented on the Managerial P&L framework, breaking comparability; (b) any new non-IFRS add-back is introduced that excludes recurring costs; (c) auditor changes or any material weakness is disclosed; (d) intangibles growth continues to outpace revenue by 15+ points; or (e) ECL coverage as a share of receivables declines while NPL 90+ moves higher.

The grade would upgrade to Clean (0-20) if: (a) the Managerial P&L is reconciled to fully-audited historical financials for three prior years; (b) individual NEO compensation is voluntarily disclosed despite the FPI exemption; (c) the SBC gap between cash flow statement and Adjusted NI is reconciled in plain English; and (d) ECL coverage and NPL 90+ continue their FY2025 improvement.

Bottom line. The accounting risk on Nu Holdings is the kind a long-only manager should respect through position sizing — call it a 100-200 bps haircut to the multiple investors would otherwise pay for a digital-bank with this growth and ROE. It is not a short thesis on its own, and there is no evidence in the public record of revenue manipulation, classification gaming for headline earnings, or auditor concern. The risk is concentrated governance, not concentrated fraud — and the difference matters.