$NU: is Nubank Capital One in 1994 or Capital One in 2006? | Vanshap Capital

Summary

• Evan Vanderveer, who says they have owned and studied Nubank for almost four years, lays out $NU as a 140M-customer digital bank with mid-30s ROE and substantial runway after Brazil’s penetration. Andrew says the stock was trading around $15, or high-teens to roughly 20× LTM earnings. Evan’s core claim: bears see 60% of Brazil’s adult population already on the platform and call it mature, but “while they’ve spent the last say 5 years getting those customers, they’re going to spend the next five sort of deepening those relationships” — ARPAC rising from ~17 toward the high-20s, where mature cohorts already sit and incumbents earn in the 40s.
• Andrew Walker’s central framing question — is this Capital One at its November 1994 IPO (13× over 12 years) or Capital One in 2006 (3× in 20 years vs. 8× for the market)? The lineage is direct: QED’s Nigel Morris backed Nubank early and Capital One alumni staffed its data science, board and operations. Evan’s rebuttal is the profit pool: Brazil’s banking system generates ~$100B of gross profit and Nubank has ~7% of it — and just ~1% of payroll loans, the largest credit vertical — so “it’s hard for us to see how they don’t grow materially from here.”
• The efficiency gap is the moat claim: ~20% efficiency ratio vs. 40-60% for incumbents, and 13,000 customers per employee vs. ~1,300 at legacy banks. Andrew’s commoditization pushback — borrowers “kind of just want the lowest rate” — is met head-on: in a commodity business the lowest-cost producer wins, and “no one is going to be more efficient than Nubank,” letting it underprice the private legacy banks rather than get undercut.
• On valuation, Evan says nobody knows the right cost of equity for an emerging-market fintech and frames NU as “more of a tech company than a bank”: ~$2 of EPS by 2029/2030 at a ~16× multiple, with no multiple expansion assumed and US upside excluded entirely. MELI and Kaspi — Andrew’s cautionary comps for EM compounders that had not generated much alpha over five years — are also Evan’s two other large holdings. Evan defends them as multiple-compression stories (Kaspi from 19× to ~6×) around businesses whose long-term models remain intact.
• Part of the recent stock weakness came from senior departures including the CFO — which Evan frames as Velez deliberately hiring for a global push: new CFO Rob Livingston was CFO of Visa North America after about 18 years at Capital One, the chief product officer came from Shopify, and international investment is capped at 100 bps of the efficiency ratio. Evan says this means “it’s not like they’re betting the company.”
• Risks Evan concedes as real: Brazil’s 90-day-plus NPLs at 15-year highs, a heavily indebted consumer stretched by a mid-teens Selic rate, and an October presidential election the market is pricing as a Lula win. Andrew adds structural worries — regulators could eventually constrain a bank becoming systemically significant, he cannot recall a banking fintech that successfully expanded cross-border, and the US expansion faces precedents such as Santander’s lack of a screaming success and Barclays’ exit.
• The AI debate cuts both ways: 60%+ of customer inquiries are AI-handled, Cognition’s Devin agents drove a claimed 12× engineering-efficiency gain and 20× cost savings, and credit-model updates fell from months to days — Evan’s justification for Velez joining OpenAI’s board. Andrew’s longer-term worry stands unresolved: if AI lets consumers constantly re-optimize commodity financial products, “the profit pool is going to zero and that 30% ROE… everybody just competes it away.”
• Ultimately Evan calls it “particularly a jockey bet” on Velez, who with his co-founder still controls the company and owns about 20%, and once voluntarily scrapped a compensation package worth a few percent of the company. Invoking Munger — “you really need to trust the banker” — his close is that, at this valuation, “you’re buying into a situation with a customer franchise at a very attractive price,” though “time will tell.”

Deep dive

1. A bank born from armed guards and bulletproof doors — now 60% of Brazilian adults

• Evan’s origin story: David Velez, a Colombian entrepreneur whose father started or co-owned a button factory where David worked in quality control, bought one cow and grew it into a herd of six before selling the cows and going to Stanford. He tried to open a São Paulo bank account in 2012 and found “armed guards and bulletproof doors and all kinds of months of paperwork.” Out of that came Nubank in 2013: branchless, mobile-first, zero-fee credit card, and “a fanatical customer culture focused on NPS.” Sequoia seeded it with $2M, Evan said he believed.
• The scale today, per the earnings report the day before taping: ~140M customers, ~60% of Brazil’s adult population, 80-90% of referrals organic (so marketing costs are extremely low versus incumbent banks), an efficiency ratio around 20 vs. 40-60% at incumbent banks, and ROE in the mid-30s — up from barely profitable when Evan’s team started studying it almost four years ago.
• Why now: macro concerns, credit concerns, and management transitions drove “a very significant decline in the stock price” earlier in the year — the stock traded ~$15 at recording, up ~10% on the morning’s earnings.

2. The variant view: the market sees saturation, Evan sees deepening

• Evan’s answer to “what is everyone missing” in a name with 87 Fiscal AI write-ups (Andrew later referred to about 70 fund-letter mentions): the market treats 60% penetration as maturity, but the next five years are about deepening relationships as cohorts mature — run those numbers out in Brazil alone and “you come up with a far more profitable enterprise.”
• Mexico, which the market has been “correctly so… extremely skeptical of,” has substantial runway. Evan said it could become “a similar size, if not bigger,” though the transcript does not make clear what comparison he intended; management benchmarked it to Brazil circa 2020 and monthly ARPAC is running at double where Brazil was at the same stage.
• The meta-point: “we’re taking the sort of 10-year view where maybe many of these other folks are not taking the same time horizon.”

3. Capital One in 1994 or Capital One in 2006?

• The lineage is direct, not merely analogical: QED’s Nigel Morris was an early backer, Capital One people joined on data science, the board, and operations, and the playbook — start with low-limit cards, rely on constantly updating models, raise limits as customers prove out — is Capital One’s.
• Andrew’s history lesson as pushback: Capital One crushed it from the November 1994 IPO — up ~13× in 12 years, approaching 4× price-to-book by 2001 — then competitors copied balance transfers, poached its executives, and it grew into a mature fintech: up only ~3× since 2006 vs. ~8× for the market, now trading around tangible book. With NU at 60% penetration and in the teens for market share, “they’re Capital One in 2006 at this point, right?”
• Evan’s rebuttal runs through the profit pool: Brazilian banking generates roughly $100B of gross profit and Nubank has about 7%; in payroll loans — the largest credit business in Brazil, which they just entered — they’re at ~1%. Despite what Evan estimated were a few thousand Brazilian fintechs, the brand is strong enough that new verticals come with a distinct advantage: “it’s hard for us to see how they don’t grow materially from here, frankly.”

4. What’s the right cost of equity for a Brazilian bank? “I’m not sure anyone has the right answer”

• Andrew’s valuation challenge: 30% ROE at 18-20× earnings would justify “infinite price to book” in the US — but if a Brazilian bank’s cost of equity is 15% they’re still earning a ton, and at 45% “they’d actually be destroying capital.” Which is it?
• Evan’s honest non-answer, worth keeping: “at the core maybe some of it is philosophical… I’m not sure anyone has the right answer.” His resolution: treat it as “more of a tech company than a bank,” underwrite ~$2 of earnings by 2029/2030 at a ~16× multiple, and note Itaú — with roughly half or less of Nubank’s ROE — trades at ~7× 2029 earnings. “I don’t think you really need multiple help here.”
• On Andrew’s cautionary comps — MELI and Kaspi, EM darlings that had not generated much alpha over five years — Evan says they are also his two other large holdings. His defense: five-year underperformance reflects 2021 starting valuations (Kaspi’s multiple went from 19× peak to ~6× trough around the Ukraine war), while the three businesses’ long-term models remain intact. MELI, Evan said, has compounded in the 20s since its IPO.

5. Commodity products, hungry competitors, and a stretched Brazilian consumer

• Andrew’s SoFi-era pushback: financial products are near-perfect commodities — “they kind of just want the lowest rate” — so a 30% ROE invites undercutting. Evan flips the logic: in a commodity business you want to be lowest-cost, and at 13,000 customers per employee vs. ~1,300 at legacy banks, Nubank can underprice the private incumbents rather than be underpriced. Competitors are “more likely… eating away at the legacy banks.”
• On regulatory risk, Evan leans on the social license: Nubank banked much of the previously unbanked population using its data to underwrite people with no credit score, earning “a very sort of fanatical following” — regulators or the population turning on them “seems unlikely, but anything is possible.”
• The risks he volunteers unprompted: 90-day-plus NPLs in the Brazilian system at 15-year highs, a mid-teens Selic rate only just starting to fall, a heavily indebted consumer getting “credit card offers in the mail every 5 seconds,” and an October presidential election where the market is pricing another Lula win — “it’s an emerging market at the end of the day.”
• Housekeeping on the model: the vast majority of the forecast value still comes from core Brazil — ARPAC moving from ~17 to the high-20s, where mature cohorts already sit and incumbents, with broader product sets, earn in the 40s. Mexico just crossed break-even; the US is excluded entirely.

6. The executive exodus, reread as a global hire-up

• Evan surfaces the bear point himself: part of the stock’s recent weakness came from departures of highly regarded executives including the CFO — because, in his telling, Velez is deliberately taking the company global. Rob Livingston, the new CFO, was CFO of Visa North America after about 18 years at Capital One; the new chief product officer came from Shopify. And discipline is explicit: international investment is capped at 100 bps of the efficiency ratio for the next few years — Evan says they are not betting the company.
• Andrew’s counter: swapping local banking expertise for global fintech résumés is a classic prelude to “oops, we lost control of our core business.” Evan’s rebuttal — the departures were at the very top, not through the operating ranks; he believes each country now has a local CEO under a LatAm CEO; Velez’s direct reports have actually shrunk; and the Brazilian credit underwriters are still underwriting Brazil — “it’s not like they’ve hired some team from Iceland.” He liked Livingston’s stated priority: “continue to focus on Brazil… not the next 10 markets.”
• Andrew’s Citibank precedent stands as the structural worry: he cannot recall a banking fintech that has successfully crossed borders — “write off, write off, write off.” Evan’s hedge: “if it’s not working they will stop doing it is our hope,” but digital banks remain a small minority of global banking and “our opinion is that over time they should be the whole thing.”

7. Jockey bet, the OpenAI board seat, and whether AI destroys the profit pool

• Evan’s framework: they look for “customer fanatics,” and NU is “particularly a jockey bet” — Velez and his co-founder still control the company, own about 20%, and Velez once scrapped a compensation package that would have added a few percentage points to his stake “because he didn’t think it was fair to shareholders.” Munger’s rule applies: “you really need to trust the banker.”
• Andrew’s needle on the OpenAI board seat: he believed Zuckerberg and Bezos did not sit on outside boards — “a weird thing for a fanatic with all of his net worth tied up in this business.” Evan concedes the risk but cites the AI payoff already visible: 60%+ of customer inquiries handled by AI on first pass; Cognition’s Devin agents delivering a claimed 12× engineering-efficiency gain and 20× cost savings; credit-model refresh cycles down from months to a handful of days. Passing on that opportunity “would be silly.”
• Andrew’s bigger unresolved worry: AI-optimized personal finance — constantly sweeping deposits to the highest yield, refinancing every loan — could mean supernormal profits for a year or three, then “the profit pool is going to zero.” Evan’s answer: lowest-cost wins any commodity endgame, and even the roughly five largest US banks pay “almost nothing” on deposits “and yet still the people keep, including myself, keep money at these banks.” Security plus service plus low cost keeps the customer, “probably even in our generation.”

8. The US expansion: god-kings or a genuine niche?

• Andrew’s sharpest skepticism: he might believe Mexico and Colombia, but a US push suggests “maybe these guys just think they’re god kings” — Santander’s US move was not, in his view, a screaming success, Barclays entered and exited, and existing apps including Robinhood, Chase and Capital One “are pretty damn good,” so the branch-line critique misses where the fight actually is.
• Evan’s defense is the niche, not the whole market: roughly one in three Hispanic people in the US is unbanked, co-founder Cristina moved to Miami in the past year or so, the push probably targets Texas and California immigrant populations already familiar with Nu, and the stated ambition is a modest 5-10M customers — not 10% share. It’s capped, excluded from the investment valuation, and “a very interesting experiment… we shall see.”
• Evan’s close: with the valuation reset, “you’re buying into a situation with a customer franchise at a very attractive price” that over the next five to ten years “could and should produce very attractive returns going forward — but time will tell.”