History

How the Story Changed

Nu Holdings has been telling essentially the same headline story since David Vélez co-founded the company in 2013 and rang the IPO bell in December 2021 — disrupt Latin American banking, compound a low-cost deposit franchise, monetize the same customer over time. Across the most recent seven filings, the language shifted considerably: "scale efficiently with discipline" (mid-2025) gave way to "AI-first" and "rebuilding banking around AI" (late-2025/2026), Mexico moved from launch-mode to break-even, and a brand-new US national bank charter appeared in the narrative for the first time. Bad news has been rare and pre-empted (PIX Financing slowdown, Nucoins repositioning, Q1'26 ECL spike were each labelled "expected seasonality" or "prudence"), and credibility has improved — but the deck is now stretched across three geographies, two consumer segments, an AI rebuild, and a US optionality bet, all at once.

1. The Narrative Arc

The arc has three distinct chapters. 2013–2020 was a single-product disruptor scaling on word-of-mouth. 2021–2024 was the public-company "platform" phase — broadening the product mix from credit cards into deposits, payments, lending, investments, insurance. 2025 onward is the AI-and-geographic-optionality phase: foundation models in production, a US charter, a strategic investment in Tyme (African digital bank), and Mexico finally turning the corner. The same CEO has narrated all three chapters, which is unusual for a company that has gone from ~3M customers to ~135M in twelve years.

2. What Management Emphasized — and Then Stopped Emphasizing

Three patterns jump out of the heatmap:

• Three themes have only intensified — Mexico expansion, secured lending, ARPAC monetization. These are the load-bearing pillars of the current bull case.
• Two themes appeared from nowhere — "AI-first" (Q3 2025) and the US bank charter (Q4 2025 onward). Both are now top-of-page in management's framing, but they were not mentioned at all twelve months earlier. The AI-first rhetoric quieted slightly in Q4 2025 (the year-end report) then re-emphasized in Q1 2026 ("not adding AI to banking, we are rebuilding banking around AI").
• Two themes were quietly dropped — PIX Financing expansion (after Q3 2024's "intentionally slowed pace of eligibility expansion") and the Nucoins loyalty program (which absorbed a $40M one-off marketing expense in September 2024 and was then never trumpeted again). Neither was framed as a failure; both simply stopped being mentioned.

3. Risk Evolution

Three risks have risen in the FY2025 risk-factor disclosure:

• AI risk is newly material. Last year's 20-F barely mentioned generative AI; the FY2025 filing now lists "use and provision of solutions powered by artificial intelligence could lead to operational or reputational damage, competitive harm, legal and regulatory risk and additional costs" inside the top-tier risk summary. That's a direct concession that the "AI-first" pitch from the call carries real downside.
• Brazil political risk has been promoted ahead of the October 2026 presidential elections. The FY2025 filing explicitly flags election uncertainty as something that "may further increase volatility in the market price of securities."
• International expansion now carries the US too. The "international expansion may not be successful" risk used to mean Mexico and Colombia; in the FY2025 filing, it now covers the OCC-approved US bank too.

Notably, two risks that should be discussed more visibly remain at constant intensity: Vélez's 74.4% voting power (which structurally limits any activist or board response if execution slips), and the IT-systems-at-scale risk (now arguably understated given the AI rebuild's dependency on third-party foundation models and cloud capacity).

4. How They Handled Bad News

Nu has not had a crisis in the period covered — no scandal, no SEC enforcement, no founder resignation, no asset write-down. Instead, the test of management's communication style comes from three soft setbacks:

The pattern is consistent: pre-emptively frame the bad news as discipline, then never re-litigate it. This works when the underlying business is compounding — and it has — but it leaves no audit trail for investors who want to know whether the original concern was right. PIX Financing's "more data to ensure credit models remain resilient" caveat was never explicitly answered; Nucoins was never tested against a revenue counterfactual.

"In Q1'26, we delivered our first quarter of IFRS profitability ahead of internal plan, ARPAC has nearly doubled, even as we have onboarded millions of newer, less mature customers." — David Vélez, Q1 2026

This is the template: lead with the milestone, follow with the volume metric that supports it, and never directly engage the EPS-miss-versus-consensus framing that the sell-side actually moved on. It is effective communication. It is also non-falsifiable in the short term.

5. Guidance Track Record

Nu rarely issues hard numerical guidance; the "promises" worth scoring are the directional commitments that the market priced and capital allocators acted on.

Credibility score: 8 / 10. Of eleven valuation-relevant commitments, seven have been delivered or are tracking ahead of plan. Two — PIX Financing re-acceleration and the Nucoins loyalty payoff — were quietly dropped rather than failed-and-acknowledged. Two more (AI-first and US optionality) are aspirational and unscoreable today. The half-point markdowns are for the style of how soft setbacks were communicated (pre-empted, then never re-litigated), the broadening of new commitments without first closing out old ones, and the structural fact that Vélez owns 74.4% of the voting power, which means external accountability for any future drift will be limited.

6. What the Story Is Now

The current story is the cleanest version Nu has ever told: the largest private bank in Brazil by customer count is now a profitable, AI-rebuilding, three-country platform with a US optionality bet attached. The core compounding engine — customer acquisition, deposit franchise, ARPAC monetization, falling cost-to-serve — has done what management said it would do across every public quarter, though the marquee external long anchor changed in 2025 (Berkshire fully exited Q1 2025, disclosed via 13F/A Aug 14, 2025).

What has been de-risked since 2021: Brazil scale (largest private bank), GAAP profitability (achieved 2023, $2.9B in 2025), Mexico monetization (break-even Q1 2026), secured-lending diversification (33% of portfolio now non-credit-card), and the efficiency-ratio thesis (~27% headed below 20%). What remains stretched: the AI-first claim is real enough to be in production but cannot yet be cross-checked against margin expansion, and US-charter optionality is being scored as "bounded downside" by a team that has never run a US national bank. The Tyme strategic-investment call-options that appeared in the Q1 2026 financials are not yet a story management is publicly building on — worth tracking.