Financial Shenanigans
Financial Shenanigans — Nu Holdings Ltd. (NU)
Nu Holdings clears the high-conviction forensic tests but does not clear the high-conviction governance tests. There is no restatement, no admitted control weakness, no auditor resignation, no live regulatory enforcement, and earnings convert to cash at $1.22 of CFO per $1 of net income — a clean bill from the income-versus-cash test. The risk profile sits in three softer places: a founder-CEO who controls 74% of the voting power under the NYSE "controlled company" exemption, three parallel reporting frameworks (IFRS, FX Neutral, Adjusted Net Income, and now a new Managerial P&L that carries only KPMG "limited assurance"), and a 77% one-year jump in the loan book against an expected-credit-loss coverage ratio that management chose to hold flat at 15.4%. The single data point that would most change the grade is the 90+ delinquency reading and coverage ratio in the next two quarters: a coverage uplift would confirm provisioning discipline, a coverage drop would convert a yellow flag to red.
1. The Forensic Verdict
The forensic verdict is Watch (30/100). Earnings quality and cash conversion are strong, balance-sheet "soft asset" intensity is low for a bank scaling at 50%+, and there is no external evidence of accounting misconduct, restatement, or auditor concern. What keeps the score above the "Clean" band is a layered metric framework introduced at the same time growth started normalizing, plus founder control that removes the usual independent check on aggressive reporting.
Forensic Risk Score (0-100)
Red Flags
Yellow Flags
Clean Tests
3y CFO / Net Income
3y FCF / Net Income
3y Accrual Ratio
Non-IFRS Gap FY25
Scorecard — 13 shenanigan categories
Two reds, six yellows, eight clean tests. The two reds — concentrated control and a multi-layered non-IFRS framework — are governance and disclosure risks, not earnings-quality red flags. The clean tests cover the items institutional investors most often get burned on: revenue recognition, working-capital build versus revenue, accrual ratio, and acquisition-fueled CFO.
2. Breeding Ground
The structural conditions are mixed. The audit committee chair is a former PwC Brazil audit partner with a Brazilian CPA, the audit committee is fully independent, and the FY2025 20-F explicitly states "we have not been required to prepare an accounting restatement at any time." But the company is incorporated in the Cayman Islands, operates almost entirely in Brazil, files as a foreign private issuer (no Form 4 insider trading visibility), uses the NYSE "controlled company" exemption, and concentrates voting power and director-nomination rights in the founder-CEO.
The breeding ground is dampening, not amplifying. The architecture is what you would expect of a founder-controlled LatAm fintech that took the cleanest possible path through US capital markets: Cayman holding company, IFRS reporter, KPMG auditor, audit-committee chair with deep audit credentials, no restatement history, no compensation metric obviously tied to a manipulable non-IFRS number. The single concentrated-control risk is real but is a governance choice the market is already pricing — it is not, on its own, evidence that the numbers are stretched.
3. Earnings Quality
Earnings quality looks strong but has two pressure points: heavy capitalization of intangible assets and a balance-sheet expected-credit-loss coverage ratio that is unchanged through aggressive loan growth.
Income statement is matched by the balance sheet
Revenue grew 37% in FY2025 against a 50% jump in total assets, a 49% increase in credit card receivables, and a 77% jump in loans to customers. The receivables-to-revenue relationship is not a "DSO blowing out" story — both sides moved in step, and the gross loan portfolio remains short-duration (credit cards plus consumer loans). The forensic ratio that matters more for a bank is ECL coverage, which management held at exactly 15.4% in FY2024 and FY2025.
Reserve coverage held flat through a 56% portfolio jump
ECL coverage stable at 15.4% is the single most important forensic statement in this filing. Management is explicit that the stability "reflects that the increase in the ECL was primarily driven by the growth of our total portfolio and remained aligned with its underlying credit quality." The 90+ NPL ratio in Brazil disclosed at 6.6% is consistent with that. But under IFRS 9, a stable coverage ratio while the loan book grows 77% is only defensible if the new vintages carry the same expected-loss profile as the seasoned book. If the FY2026 90+ NPL ticks above 7% without a coverage uplift, the ratio will look mechanical, not analytical.
Capitalization of soft costs is rising
Intangibles jumped 73% YoY to $602M and "Other assets" 112% YoY to $1.40B. The MD&A explains the latter includes deferred card-issuance expenses amortized over time. Capex per cash-flow statement reads $7M (property, plant & equipment only); capex per MD&A reads $340.8M because it includes intangible-asset development. The line item disclosure is honest, but a reader who anchors on the cash-flow statement understates investment intensity by 47x. Combined goodwill plus intangibles is still only 1.4% of total assets, well below software peers, but the gap between depreciation/amortization ($98M FY25) and full capex ($341M) means there is a $243M annual book accumulation flowing through the balance sheet rather than the P&L.
Yellow flag: every dollar of intangible asset that is capitalized today is a dollar of operating cost that does not hit the income statement until later. Watch the intangible-amortization line over the next two years. If amortization does not catch up with the recent $254M intangible build, operating margins are partly funded by capitalization, not productivity.
Non-operating "Other income" is small and stable
"Other income" is small (3.4% of pretax FY25) and is disclosed as FX gains on foreign transactions and monetary adjustments on recoverable taxes — these are mechanical, not one-time gain manufacturing. Clean.
4. Cash Flow Quality
Cash flow quality is the single strongest part of the forensic picture. Earnings convert to cash, but the conversion mechanism deserves to be named: this is a bank, not a SaaS company, and CFO strength is the natural by-product of deposit funding.
CFO consistently exceeds net income
Three years in a row CFO and FCF have landed within $30M of each other at roughly 1.22x net income. The non-cash adds bridging the two are dominated by ECL provisions ($4.7B added back in FY25 reconciliation), interest accruals, and deferred tax movements — exactly what you would expect from a fast-growing IFRS bank.
Accrual ratio is clean
The accrual ratio has been negative — CFO exceeds NI — for four consecutive years. The FY2021 spike is the pre-IPO ramp where loan-book growth ($-$2.9B of operating cash use) outran net income; that pattern reversed once deposits matched receivables.
Source of CFO: deposit growth, not working-capital tricks
Deposits grew $13.0B in FY2025 — that is the dominant inflow inside the $3.5B CFO line. Per IFRS, deposit growth is correctly classified as an operating cash inflow for a bank. There is no securitization, no recourse-factoring program, no supplier finance, and no "off-balance-sheet" receivable sale disclosed. The CFO line is real, but it is structurally dependent on deposit growth keeping pace with loan growth. Borrowings and financing also tripled to $4.4B, suggesting management is starting to lean on wholesale funding to top up retail deposits.
Yellow flag: CFO of $3.5B in FY2025 is supported by $13.0B of net deposit inflows and $2.3B of net wholesale borrowings. The day deposit beta rises or wholesale spread widens, the CFO/NI ratio will compress mechanically. This is not a manipulation — it is the bank-model cyclicality investors should price.
Capex picture depends on which line you read
The cash-flow statement reports capex as $7M FY25 because it tracks only property, plant & equipment additions. The MD&A reports total capex (including capitalized intangibles) of $340.8M — a 47x gap with what most data feeds will surface. Depreciation/amortization is $98M, so the full-capex / D&A ratio is roughly 3.5x. This is a disclosed convention, not deception, but it shows up in the data files in a way that materially understates investment intensity. Any forensic check that uses "capex per cash-flow statement" as a productivity proxy on this name will be wrong by an order of magnitude.
5. Metric Hygiene
This is where the forensic picture earns its yellow flags. Nu reports four distinct earnings frameworks side by side: GAAP/IFRS, FX Neutral, Adjusted Net Income, and now the new Managerial P&L. Each is internally consistent. The risk is that the headline number an investor anchors on changes from quarter to quarter, and that the framework drift is doing the work of accounting choices that would otherwise be debated more openly.
Non-IFRS gap is closing — that helps
The Adjusted-to-GAAP gap has compressed from 16.1% in FY2023 to 7.1% in FY2025 as the SBC add-back has stayed flat in dollars while GAAP earnings tripled. This is a positive forensic signal — management's preferred number is converging on the audited number, not diverging.
The Managerial P&L is the new red flag
The Q1 FY2026 6-K is the first reporting period in which the new "Managerial P&L" framework is used as the primary disclosure. The release states KPMG performed a limited assurance engagement on the compilation process — meaning KPMG concluded "nothing came to its attention" suggesting impropriety, which is a deliberately lower bar than the reasonable assurance of a full audit opinion. Limited assurance is a recognized professional standard, not a flaw, but it is the right framework to use to bridge audited IFRS results into a presentational re-cut — not the right framework to use as the headline number an investor sees first.
Red flag: management introduced a presentation framework in Q4 FY2025 that re-classifies P&L lines and applies tax-equivalency adjustments, and the auditor signs off only at the "limited assurance" level. If the Managerial P&L becomes the headline framework the IR deck uses, the reader will need to learn the reconciliation bridge to the audited IFRS statements every quarter. That is exactly the disclosure shape that historically precedes definition drift.
6. What to Underwrite Next
This work belongs in the diligence-overlay bucket, not the thesis-breaker bucket. The earnings, cash flow, and balance sheet pass the high-conviction tests. The governance, presentation, and reserve-policy items are real but priced. A position-size haircut of 5-10% versus what the income-statement quality alone would suggest is reasonable; a valuation haircut is not.
Five specific items to monitor
Practical use
Treat this as a position-sizing limiter, not a thesis breaker. The accounting plumbing is sound, earnings convert to cash, and there is no external evidence of misconduct, restatement, or auditor concern. The risk is structural: a founder who controls the vote, a new presentational framework that carries only limited assurance, and a stable headline ECL coverage ratio that the next two quarters of credit data will validate or invalidate. Underwrite with a 5-10% sizing discount versus what the earnings quality alone would dictate, monitor the Q2 and Q3 FY2026 90+ NPL print and the IFRS-vs-Managerial-P&L bridge, and revisit the grade when the Managerial P&L has either three quarters of consistent definitions or a definition change that invites questions.