Nu Holdings (NU)
The Past, Present, and Future of Latin America’s Largest Digital Bank
Nu’s Past
In 2010, 28-year-old David Vélez faces a dilemma. After several years as an investment analyst at Morgan Stanley and General Atlantic, he is certain he wants to start his own business – yet lacks a compelling idea. Entrepreneurial spirits run deep in the Vélez family. His father, one of twelve siblings, runs a button factory in Medellín, and business conversations are common at the dinner table.[1]
Seeking inspiration for what should become his “life’s work”, David returns to his alma mater, Stanford University, to pursue an MBA. A classmate points him toward nearby VC firm Sequoia Capital, then considering opening an office in Brazil, and suggests he meet one of its managing partners: Doug Leone.
In Silicon Valley, Doug is known as an opiniated, uncompromising investor who seeks outliers driven to do extraordinary things for whatever reason – maybe, in his own words: “because daddy told them they weren’t good enough and they want to show daddy.”
Doug hires David as a part-time associate to research Brazilian startups. In practice, it becomes a full-time job on top of an MBA. David soon finds himself in the office by 04:30 am, using the four-hour time difference between São Paulo and Palo Alto to call founders, then attending class from 09:00 am before returning to Sequoia. His work takes him repeatedly to São Paulo, where he encounters what he later describes as the most painful customer experience of his life – opening a bank account in Brazil:
“I had to go into this banking branch that had bulletproof doors. There were a lot of armed guards that asked me to leave the branch and leave my backpack in a locker and then, walk back in and wait 45 minutes for a branch manager to throw a bunch of paperwork at me and then, start this process of five months trying to open up a simple bank account. And there was so much anxiety and frustration and pure rage about how hard it was to get a simple bank account, to then pay some of the highest fees and interest rates in the world, that I didn’t understand how it was possible that Brazilians were putting up with this. How isn’t anybody competing with these big banks and offering better solutions? And I started talking to my friends in the space and my Brazilian friends, and said, “You know what? Yeah, these banks are horrible. But you know what? All the other banks are the same. There are no ideal alternatives. If you complain to your bank they’re going to say, ‘Well, where else are you going to go?’”[2]
Two years in, Doug is impressed by David’s work ethic but underwhelmed by Brazil’s startup scene. He cancels Sequoia’s plans for a local office, giving David a shock yet also the final push to start his own business: a digital, customer-obsessed challenger bank in a country where 80% of deposits are controlled by five incumbents: Caixa, Bradesco, Itaú Unibanco, Banco do Brasil and Santander Brasil.
With $2m in seed funding and two co-founders – Christina Junqueira, a Brazilian banker at Itaú Unibanco, and Edward Wible, the company’s first software engineer – David launches Nubank in 2013. The team introduces its first product, a purple, no-fee Mastercard credit card, in 2014. After a slow start, it surpasses 1m customers by 2016. From there, the Brazilian customer base grows exponentially (2017: 3m, 2018: 6m, 2019: 20m, 2020: 33m, 2021: 52m, 2022: 71m, 2023: 88m, 2024: 102m, 2025: 113m).
Nu’s Present
Let’s jump to today: Nubank is the largest private financial institution in Brazil by number of customers, offering a full suite of banking products, including its savings account “NuConta”, personal loans and investments. It serves 131m customers – 113m in Brazil, 14m in Mexico, and 4m in Colombia – and generates $15.8/2.9bn in revenue/net income as of FY25. The total credit portfolio stands at $32.7bn, of which $21.8bn are credit card receivables, against $41.9bn in customer deposits.
Nubank’s rise rests on structural cost advantages – primarily in customer acquisition, cost to serve, cost of risk, and funding. These advantages enable reinvestment into lower product prices or superior customer service. The table below illustrates the magnitude of this edge in the form of a 21% efficiency ratio (non-interest expenses relative to revenue), a 30% return on equity (ROE) and revenue per employee five times that of peers.
Another observation: Itaú – widely regarded as by far the best-run incumbent bank – has preserved its historically strong economics (>20% ROE and >24% recurring ROE in FY25, see below) despite the rise of digital banks. Itaú launched its own digital offering “iti” in 2019 before integrating it into its core app, while Bradesco introduced “next” in 2017. Both have since reduced costs and halved their branch networks. David’s earlier painful customer experience would look different today: many incumbents have meaningfully improved.
Next, I outline the current state of play in Brazil and explain why, despite competitors attempting to retrofit customer-centricity onto legacy core systems and values, Nubank’s clean-slate architecture may confer competitive advantages that are difficult to replicate.
A Developed Financial System and the PIX Revolution since 2020
If you have traveled to Latin America, you have noticed cash remains king in some countries, while financial inclusion and formal employment can be limited. None of this applies to Brazil. The share of informal employment is low around 35% compared to e. g. 55% in Mexico. There is virtually no unbanked population – in part due to Nubank enabling more than 25 million citizens to access their first credit card or bank account. 85% of all retail transactions in Brazil are conducted digitally: thereof 50PP via Brazil’s instant account-to-account (A2A) transfer system Pix, and 35PP via debit and credit cards.
Pix launched in 2020 and was mandated for all major banks by the Central Bank of Brazil (BCB). It has since become a runaway success, used by more than 170 million Brazilians. What began as a peer-to-peer tool now replaces cash and debit cards at the point of sale. Today, it is common for consumers to unlock their phones, open their banking app, tap Pix, scan a QR code, and confirm payment.[3] On its fifth anniversary, the BCB summarized Pix’s impact on payment behavior: “Many people didn’t use the accounts they had. Or they only received their salary, withdrew everything, and used cash exclusively. After Pix, people realized the convenience of paying bills via mobile and changed this behavior, starting to actually use their accounts.”
Beyond physical retail, merchants such as Magalu often offer a 10% online discount for Pix payments, reflecting lower acceptance costs (0.3% vs. 2.3% for credit cards). This raises a natural question: if Pix is so dominant, why do credit cards remain Nubank’s most popular product in Brazil, and why does credit card payment volume continue to grow at a mid-teens rate?
First, Pix primarily transfers funds you already have, whereas credit cards generate funds you do not. That is why credit cards are complementary to Pix, while debit card and cash are substitutes. Second, a deeply ingrained Brazilian habit is to pay for most higher-ticket purchases in up to 12 equal monthly installments (“parcelado”) to preserve liquidity. With credit cards, these installments are interest-free to the consumer (“parcelado sem juros” or PSJ), while the issuer earns its return from the merchant through a high merchant discount rate (MDR).[4]
Pix initially offered no installment feature. Banks such as Nubank therefore created a product called “Pix no creditó”. Because the transaction remains a Pix, the consumer still qualifies for the 10% discount, and the merchant avoids the card MDR. However, there is a catch for the consumer: the bank must be compensated for providing financing of up to twelve months. As a result, “Pix no creditó” is not interest-free for the consumer who typically pays a 2-7% monthly interest rate on the installments. In many cases, this can prove more expensive than paying the undiscounted price with a credit card under “parcelado sem juros”.
In summary, interest-free installments are a key reason why cards remain both popular and necessary in Brazil. Additional benefits included fraud protection, dispute management and rewards. Pix has recently introduced a new protocol, “Pix Parcelado”, which mirrors “Pix no creditó” but draws on a dedicated bank credit line rather than the existing card limit. If adoption increases, banks are likely economically indifferent, as they continue to earn attractive returns through interest charges. As with “Pix no creditó”, compensation for credit risk shifts from the merchant to the consumer.
The prevalence of PSJ also explains why, of around $100bn in total credit card receivables outstanding in Brazil, only 24% earned interest in 2024. At Nubank, that share is somewhat higher: $7.6bn of $21.8bn in card receivables earned interest in 2025 (35%). In its oldest and most deeply penetrated product, Nubank therefore holds 25-30% market share. Relative to total financial retail industry revenue of around $150bn in Brazil and above $3tn globally, Nubank’s market share stands at 10% and 0.5%, respectively.
Cultural and Financial Distinctions
Cardholders who consistently pay their balances in full are called transactors. They incur no interest charges and pose no default risk. In the U.S., commercial banks hold more than $1.2tn in credit card balances, 70% of which revolve and earn interest. In Brazil, the proportions are reversed. Largely due to the prevalence of PSJ, only 24% of industry-wide credit card balances earn interest, and revolving credit is typically a last resort.
Brazilian revolvers, however, face some of the highest borrowing costs globally – currently averaging 15.2% per month, or 450% annually. Bradesco (14.2% p.m.) and Itaú (13.0% p.m.) price below the industry average, while Nubank positions itself a notch more customer-friendly at 12.3% per month.[5] These rates reflect elevated default risk. In Brazil, more than 60% of revolving balances ultimately default, compared with NCO rates of around 4% in the U.S. (or 6% when measured against revolving card balances only). Please see the lead indicator for later defaults, 90 days past due loans relative to total revolving balances in Brazil, below.[6]
Turning to deposit yields Nubank offers its Brazilian customers, the company again positioned itself more competitively and accumulated $33bn in local deposits. Since launching “NuConta” in 2018, it has raised expectations for returns on account balances. NuConta initially passed through 100% of the interbank rate (CDI, currently 14.9% p. a.), prompting incumbents to follow. However, differences remain in ease of access. Incumbent banks typically require customers to manually place funds into a certificate of deposit, reflecting the limits of fully embracing a customer-centric model.
Nubank, by contrast, originally paid 100% of CDI automatically from day one, with daily liquidity and no action required. In 2022, automatic accrual was adjusted to begin after 30 days. To earn 100% of CDI immediately, customers must make three clicks in the app and move funds into a separate “money box” (“Caixinha”). To offset an otherwise worse customer experience, Nubank introduced Turbo Money Boxes (“Caixinha Turbos”), offering 115-120% of CDI to more engaged customers or those depositing over R$900 per month. Such yields are typically unavailable to retail customers at incumbent banks.
By consistently positioning itself more attractively than peers, Nubank reduces churn and increases the likelihood customers recommend the app to friends. In Brazil, Nubank’s Net Promoter Score has long been around 90, well ahead of competitors. Active customers represent 83% of total accounts and more than 60% use Nubank as their primary bank. The company reports the lowest complaint index published by the BCB and service agents are incentivized to create “WoW” moments (e. g. personalized letters or gifts to customers once a complex issue has been resolved).
More than 80% of customers in Brazil (and 55% in Mexico) were acquired through unpaid word-of-mouth referrals. As a result, marketing spend remains exceptionally low for a retail credit card bank at $303m, or 0.4% of total assets. Average customer acquisition cost (CAC) is $7.0, compared with monthly ARPAC of $13.3 and annual ARPAC of $159.6. With a net income margin of 18%, Nubank should recoup its customer acquisition spend in almost one quarter – implying an unusually attractive LTV/CAC.
The company built its core banking system in-house and runs it entirely in Amazon’s cloud, enabling scalability alongside rapid growth. Cloud infrastructure costs are included in its “cost to serve” metric (monthly operations, transactional and support expenses per active customer) which has remained stable at $0.8 since 2021, while monthly ARPAC has nearly tripled from $4.7 to $13.3 in FY25.
Driven by higher revenues and customers per employee, Nubank’s cost to serve is up to 85% lower than that of incumbents, while incumbent retail ARPAC is 3x higher. Given Nubank’s largely fee-free model and incumbents’ older, more affluent customer base, absolute levels will not converge, but cross-sell headroom remains – reflected in older cohorts already trending above $25 ARPAC (see below). Co-founder Cristina recently expressed her view that Nu could be 2-3 times larger in Brazil with its existing customer base.
Asset Quality and Underwriting Philosophy
As with any company, for everything that can go right, something can go wrong. For a bank, the primary risk is always asset quality: loans that deteriorate and must be charged off beyond prior provisions. The most prominent recent example of an asset-quality-driven crisis was the GFC in 2008, triggered by a surge in defaults in mortgages and mortgage-backed securities. Nubank is predominantly a credit card bank, and credit card loans are the riskiest form of unsecured consumer credit, with default rates up to ten times higher than most other lending categories.[7]
As of late 2025, 9.0% of industrywide credit card receivables in Brazil are classified as non-performing loans (NPLs) from borrowers at least 90 days past due (see below). Notably, this 9.0% 90+ NPL ratio exceeds levels observed in Brazil during the GFC, when the country was less severely affected than many others. In the U.S., the GFC temporarily pushed credit card NCO rates above 10% of balances – about 2.5 times the long-term average of 4%.
Therefore, the question is not whether credit card lending leads to elevated default losses – it does. As an example, Nubank’s most recent 6.7% 90+ NPL ratio for its Brazilian consumer credit portfolio (cards and personal loans) compares with only 3.6% at Itaú, whose consumer credit mix includes a larger share of secured lending, such as mortgages.[8] The more relevant questions are: a) how well are credit card lenders compensated for assuming this risk, and b) whether they underwrite conservatively and maintain loss reserves in excess of NPLs throughout the cycle.
With respect to a), despite higher defaults, credit card banks are attractively compensated, generating a ROA that is multiples of that of a typical commercial bank. On b), I look for management teams that maintain consistently high NPL coverage ratios (at least 125%) across the credit cycle. Anything less risks under-reserving and overstating current earnings at the expense of future results. By that standard, Nubank’s 90+ NPL coverage ratio (loan loss reserves divided by NPLs) has consistently been at or above 200% (Q4/25: 231%, Q4/24: 227%, Q4/23: 240%, Q4/22: 242%, see below).
Credit loss allowance expenses totaled $4.2bn in FY25, or 28% of the average interest-earning portfolio (IEP), resulting in a risk-adjusted net interest margin of 10%. Actual net charge-offs (NCOs) totaled $2.8bn, or 19% of the average IEP. Thus, credit loss allowance expenses relative to NCOs amounted to 153% in FY25, well above 100%, indicating that reserves are being built rather than depleted. While Brazil’s macro backdrop remains top of mind for investors, David and his team have demonstrated the ability to develop the bank under persistent headwinds: high interest rates and inflation, an 8% GDP contraction between 2014-16, impeachment and corruption scandals.
Mexico as the Potential Next S-curve and a Wild Card in the U.S.
For Nubank to grow its earnings over the coming years, management does not count solely on Brazil. Since launching in Mexico in 2020, Nubank has gained 14m customers, adding 1m per quarter, or +41% year-over-year as of FY25. A comparison of Mexico with Brazil five years post-launch suggests that several key metrics are tracking ahead of Nubank’s home market at a comparable stage.
Although Brazil’s population of 213m exceeds Mexico’s 132m, the opportunity might be comparable. More than 40% of Mexico’s population remains unbanked and only 25% hold a credit card. Mexico’s GDP per capita of $14.2k also compares favorably with $10.3k in Brazil. Mexican revenues are just beginning to inflect, contributing 7% to total FY25 revenue. The country briefly reached break-even in Q3/25 before returning to a loss in Q4/25, reflecting continued investment.
Beyond Mexico, Nubank has received approval to establish a de novo bank in the United States. The entity must be capitalized within 12 months and commence operations within 18 months.[9] Co-founder Cristina Junqueira has relocated to Miami to lead the effort, targeting the country’s 68m Hispanic population.
With expansion into more sophisticated markets, investors question whether Nubank’s current ROE of 30% is sustainable. The argument is that for example in the U.S., the average commercial bank earned a 1.2% ROA in 2025 and requires 9-10x leverage to generate a 10%+ ROE (see below).[10]
However, one could also argue that entering additional countries and greater scale might ultimately increase – not reduce – ROE once the investment phase subsides. In its scaled home market, excluding the startup losses in other markets, the business already generates a ROE above 40%. In the U.S., a more relevant benchmark may be a subset of the banking industry aligned with Nubank’s positioning rather than the entire universe.
Nine monoline credit card banks operate in the U.S., with more than half of their assets in consumer lending and more than 90% of that in credit cards. Revolving card balances carry an average annual interest rate of 22%. Net of funding costs, credit card banks earn a net interest margin of nearly 10% compared with 3% for the average U.S. commercial bank. Despite long-term NCO rates around 4% in card lending, credit card banks generated an average 3% ROA and 23% ROE over the past decade, according to FDIC data.
Higher returns are possible but depend materially on marketing intensity. Credit card banks typically spend 1-2% of total assets on marketing. Capital One, the largest U.S. issuer., spent $5.9bn in FY25 – more than Nike (4.7bn) or Coca-Cola ($5.4bn). Nubank spent $303m over same period.
Whether Nubank has a chance to exceed the subset’s average ROE or not, will depend on customer acquisition costs, success in unpaid referrals and customer churn rates. It won’t be straightforward. However, Cristina often notes that there is room left for providers who credibly aim to treat customers better (see quote below). She may have identified an opening here given that large U.S. incumbent banks typically do not offer attractive, automatically accruing yields on savings accounts. Even the digital challenger bank SoFi passes through only 90% of the interbank rate.
“There’s nothing about our business model that we feel that is specific to Brazil or Latin America even. If you think about this using Jeff Bezos’ hat – People are always going to want to pay lower fees. They’re always going to want to have better service. They’re always going to want to have better options. They’re always going to want to have a better experience. They’re always going to want to deal with things in a faster, more convenient way than going to a branch. They’re always going to want to get better yield on their deposits, on their savings, right? So those are all problems that we know how to solve, that we’ve been working on for a while now. So, we have a hypothesis that even beyond the people that already know and trust us and are here in the U.S. – and there could be a few million of them – there’s a wider audience with whom this message can also resonate.”
Thanks for reading!
If you’d like to dive deeper, below are a few Substack posts from great authors covering additional aspects of NU Holdings.
If you don’t want to miss anything, you can follow me on Twitter: patient_capital
Further Reading
Wolf of Harcourt Street ⟶ Nu Holdings Investment Thesis
Kris ⟶ The NU Analysis (The Perfect Weekend Read)
Ian Bezek ⟶ Why I, A Former Nubank Skeptic, Am Now Looking For Shares To Triple
Tony Cueva Bravo ⟶ Nubank dismantled Brazil’s banking oligopoly. Here’s how they do it.
Compound & Fire ⟶ Nu Holdings: Revolutionizing Banking in Latin America
[1] The factory is later rebuilt in San José, Costa Rica when the family relocates during Colombia’s 1980s cartel violence
[2] David Vélez in January 2025 on Sequoia’s Crucible Moments podcast: https://sequoiacap.com/podcast/crucible-moments-nubank/
[3] Nubank generates by far the most Pix transactions in the country with an estimated volume share above 25%
[4] The merchant can either receive the installments over time, net of the MDR or the full price instantly, net of the MDR plus an anticipation fee as the bank assumes the risk of not receiving all 12 installments over the next year from the cardholder
[5] https://www.bcb.gov.br/estatisticas/reporttxjuros/?codigoSegmento=1&codigoModalidade=204101
[6] It is possible to track the lead indicator for later defaults, 90 days past due loans relative to total revolving balances here: https://www3.bcb.gov.br/sgspub/consultarvalores/consultarValoresSeries.do?method=consultarGraficoPorId&hdOidSeriesSelecionadas=21127
[7] An example: in the U.S. credit card default losses are almost comparable in absolute numbers to those on U.S. corporate bonds despite there being ten times as much U.S. corporate bonds outstanding ($11.5tn as of 3Q25 according to SIFMA) than credit card balances ($1.2tn)
[8] Historically, Nubank’s 90+ NPL ratio for its credit card segment in Brazil has often been 100bps below the market average
[9] With $3bn in cash at the holding level, there should be ample liquidity to incubate the entity without raising new equity
[10] To track the state of the overall progression of the U.S. commercial banking industry, I recommend the FDIC’s Quarterly Banking Profile: https://www.fdic.gov/quarterly-banking-profile
Disclaimer: This article is for informational purposes only. It is no investment advice. The Imprint applies.
Holding Disclosure: At the time of writing, the fund holds a position in Nu Holdings.
NOTICE: This is a marketing communication. Please read the prospectus of the UCITS and the PRIIP-KID before making any final investment decision. Covesto Asset Management GmbH acts as a tied agent within the meaning of § 3 para. 2 WpIG exclusively on behalf and under the liability of NFS Netfonds Financial Service GmbH, Heidenkampsweg 73, 20097 Hamburg, insofar as it provides services attributable to investment advice and investment brokerage (§ 2 para. 2 nos. 3 and 4 WpIG) relating to financial instruments within the meaning of § 2 para. 5 WpIG. Covesto Asset Management GmbH is registered as a tied agent of NFS Netfonds Financial Service GmbH in the public register maintained by the Federal Financial Supervisory Authority (Bundesanstalt für Finanzdienstleistungsaufsicht – BaFin). The register can be accessed under www.bafin.de.
NU's explosive earnings power is driven by an efficient flywheel (more users, more products per user, more revenue per product), pushing average revenue per user to a record $16 while cost to serve a user remained flat at $0.80 per user, allowing incremental revenues to flow directly into profits. Obviously no legacy bank can serve a customer for less than $1 because they have to pay for the physical locations of their branches which is a huge advantage especially as younger mobile native population becomes wealthier over time. A 32% market correction pushed the forward price-to-earnings ratio down to 13x, making it incredibly cheap. See my DCF calculation with an Expected Annual Return over 40% assuming they can keep up their existing growth rates of ~35% and trade at 25x P/E which is below historical average multiples:
https://freecashflowcompounder.substack.com/p/portfolio-valuation-as-of-may-31
Thanks to Patient Capital Fund for highlighting Nu Holdings Ltd. (NU; NU US)
(1) At first glance, NU looks attractive. High growth but NTM P/E ~15x. However, with Brazil 10y government bond yield ~14%, NU's ~7% earnings yield no longer looks that attractive.
What do you think?
(2) There are complaints that NU is taking advantage of borrowers: https://www.washingtonpost.com/world/2023/12/24/brazil-credit-card-debt-crisis/.
What are your thoughts on this risk?