The most successful fintech in Brazil is the central bank

Three. Full. Weeks.

That’s how much time an HR team can save every year by managing employee benefits with iFood Benefícios.

For context, that’s 110 hours.

Not spent recruiting. Not improving culture. Not retaining your best people.

Spent adding employees, managing balances and cards, processing payments, preparing reports and dealing with bureaucracy.

Employee benefits are supposed to help your people. They probably shouldn’t consume three weeks of the HR team’s time.

iFood Benefícios puts all of that administrative plumbing into one management platform—with zero fees for the company.

Your HR team has better things to do.

Request a demo here.

The most consequential fintech of the decade isn’t Nubank or Revolut, it’s PIX. Brazil’s central bank built it as public infrastructure, made payments nearly free, and shifted the profit pool in Brazilian finance toward lending, software and data. Within a year, 62% of Brazilian adults had used it. Six years in, it accounts for 55% of the country’s payments by count. Along the way, it became a growth engine for a generation of fintechs, including Nubank, one of the world’s most valuable digital banks.

The wildest part is what it cost. A team of about 60 inside the central bank built PIX in-house in about two and a half years for the price of a Series A in some mediocre AI company. My back-of-the-envelope estimate puts the build at US$15–20 million, including staff time, and operating costs at about US$10 million a year. Call it US$70 million over five years, against an estimated US$22 billion in savings for users. That’s about 300 times what it cost. To put things in perspective, that is the multiple Kleiner Perkins or Sequoia made on their investment in Google in 1999.

Before PIX, moving money in Brazil was either free and physical (cash) or digital and costly (cards, TED, DOC, boleto). Five banks held about 81% of banking assets and more than 80% of credit, making a ton of money in fees. Those banks earned up to ~$2 billion a year from TED fees alone. The system’s return on equity was 16.5% in 2019, meaning every $100 of shareholder money earned $16.50 a year. That is the average. Itaú, Bradesco and Santander Brasil each earned 20–23%, against 11–12% for US banks and about 6% for European ones. It was a textbook case of “your margin is my opportunity”, and both Nubank and the central bank went after it.

You’d think that making payments nearly free would have gutted those returns. Far from that, Itaú still earns above 20% on equity. Nubank earns 33%. That’s the part I find mind-blowing. PIX could save users an estimated US$22 billion in five years and leave Brazilian finance with a profit pool rich enough to support a generation of challengers. As transfer fees fell, more of the opportunity was in credit. And now Nubank earns a higher return on equity than Brazil’s biggest banks did before PIX. What a wild ride.

So why did PIX take off? There were six critical strategic and design decisions. Let’s break them down.

Decision 1. The central bank is both operator and rule-setter. Brazil’s central bank wrote the rules, ran the system that settles each payment, and operated the directory that tells the system which account belongs to a phone number or another PIX key. That let it set common standards all the way down to what you see in your banking app. The central bank handled the plumbing. Banks and fintechs handled the customer experience. This is where PIX differs from one of its closest peers, India’s UPI. India’s UPI was built and is operated by NPCI, a bank-backed nonprofit created at the initiative of the Reserve Bank of India and the Indian Banks’ Association. Brazil’s central bank took on that job itself.

Decision 2. Make the big players join. The central bank required every regulated institution with more than 500,000 active customer accounts to let customers both send and receive PIX payments through their banking apps. Think about what it would have been asking the big banks to do voluntarily. They were making a ton of money from transfers, and PIX was about to make those transfers free for individuals. If even one of the biggest banks had sat it out, millions of customers would have had a reason to stick with the old ways of paying. Instead, PIX launched with 734 participating institutions. The first time you tried it, there was already a good chance the person you wanted to pay could receive it.

I think that’s the biggest reason PIX took off in Brazil and CoDi didn’t in Mexico. Mexico’s central bank launched CoDi to make instant payments part of everyday life. But it only made one side of the transaction mandatory. Banks covered by the rules had to let customers pay a CoDi request. Whether customers could use the same banking app to ask for money was up to each bank.

Decision 3. Open PIX to non-banks. The central bank let eligible payment institutions and fintechs join the same network as banks and offer PIX under the same rules. So if you had a Nubank account, you could send money to someone at Itaú or get paid by someone at Bradesco, all from the Nubank app.

Think about what that meant for Nubank. You didn’t have to choose right away between a hundred-year-old bank you already trusted and a digital newcomer you weren’t sure you trusted yet. You could keep your Itaú account and use Nubank for everyday payments because the two accounts could send money to each other. That gave Nubank a shot at becoming your main bank before you’d even decided whether you wanted to switch.

Decision 4. Make PIX free for individuals and cheap for merchants. The central bank wrote free transfers between individuals into the PIX rulebook. Before PIX, a bank transfer could cost the equivalent of about US$3–5. Think about paying someone back for lunch. You could owe them about US$6 and pay another US$3 just to send it. With PIX, you sent US$6 and they got US$6. That made bank transfers useful for all the little payments you’d never have paid a fee to make.

For merchants, banks and payment providers could still charge fees. In the last quarter of 2022, on a sale worth US$100, a Brazilian merchant paid an average of 33 cents with PIX, compared with US$1.13 for debit and US$2.34 for credit. That’s about one-seventh of the credit card fee, or US$2 back in the merchant’s pocket on every US$100 sale that moved from credit cards to PIX. And the money arrived in seconds.

Decision 5. Replace branch and account numbers with aliases and QR codes. Before PIX, sending someone a bank transfer meant asking for their bank details and typing them in. With PIX, you could ask for their phone number. If they’d registered it as their PIX key, the system would look up the account for you. A tax ID, an email address or a randomly generated key worked the same way. Your app showed you who you were about to pay before you hit send.

QR codes made it even simpler at checkout. The merchant could show you a code with the payment details already in it. You scanned it in your banking app, checked the amount and confirmed. Think about how different that felt from making a bank transfer. You could pay a friend with a number you already had in your contacts, or pay for lunch by pointing your phone at a code. PIX made both things easy enough to do every day.

Decision 6. Any use case, one rail. From day one, person-to-person, person-to-business, business-to-person, business-to-business and government payments all ran on the same rail.

Think about how critical that was to driving a major change in consumer behavior. Brazilians who were used to paying in cash were transitioning into the digital payments universe. With PIX, they could make that transition across much of their everyday lives at once, because the same system worked for the different kinds of payments they already made.

Obviously, none of that was easy or smooth. The big banks weren’t exactly excited. As early as 2020, they could see that free transfers threatened the money they made from transfer and account fees. Even a year after PIX launched, fees overall still accounted for about a third of revenue at Itaú, Bradesco and Santander Brasil.

And then there was COVID-19. In the middle of the pandemic, the central bank had to coordinate more than 700 institutions around one technical standard and one set of rules for how PIX would work in their apps. Each institution had to get its own systems ready so that, on launch day, a customer at one bank could send money to a customer at another in seconds. They still launched on schedule.

And then there was fraud, a major issue then and now. Scammers could move stolen money in seconds, and in so-called lightning kidnappings, people were forced to make transfers on the spot. The central bank had to react quickly. By the first anniversary, it had introduced limits on nighttime transfers and a process to help victims recover their money.

To me, that makes it even more spectacular. A team inside the central bank got the big banks to offer a service that threatened some of their fees, launched it on time in the middle of a pandemic, and moved quickly when criminals started exploiting it.

The effect on the whole ecosystem was gigantic, as you can see in the charts below.

But I’d go further. I think PIX was instrumental in turning Nubank from the company that issued your purple credit card into your primary bank, the place where you kept most of your money. Before PIX launched, Nubank had around 30 million customers. Today it has nearly four times that many in Brazil alone.

Here’s where PIX fits into this growth.

It removed the cash-in problem. Before PIX, getting money into a branchless bank could mean paying a boleto and waiting days for the money to show up, or paying your old bank for a TED transfer. With PIX, you could move money into Nubank in seconds and move it back just as easily. That made it practical to keep money in the same account you used for your credit card.

Fintechs won the opening round. In the first ten days of PIX key registration, customers of incumbent banks accounted for just 35% of registrations. Nubank alone took 24%. Registering a key with Nubank meant you could get paid directly into that account, even by someone who banked with Itaú. And every fintech that joined after launch got access to the same network.

It became the engagement engine. Once money could come straight into Nubank, you had a reason to open the app between credit card bills. You could get paid there and use that same money to pay rent or send it to someone else. Nubank later reported handling a billion PIX transactions a month. It said it accounted for 23% of all PIX transfers in Brazil and was on one side of 39% of transfers between institutions.

It became a credit product and pulled in deposits. Nubank let customers make a PIX payment using their credit card limit. By December 2023, more than 35% of its active credit card customers had used the feature. At the same time, using Nubank for everyday payments gives people a reason to leave money in the account. Nubank ended 2023 with US$23.7 billion in deposits, a low-cost source of funding for its loan book. So PIX helped Nubank on both sides of the lending business. It gave customers a new way to borrow and a reason to keep the money that helped fund those loans at Nubank.

Nubank is hardly an isolated example. Brazil has thousands of fintechs, and there are businesses being built on the PIX rails across just about every space imaginable. Think digital banks and wallets like PicPay, Mercado Pago, Inter and C6 Bank, or merchant acquiring, where CloudWalk built free PIX acceptance into InfinitePay.

Behind those products, there’s a whole industry in financial infrastructure. It spans banking as a service, core banking, issuer processing, fraud prevention, and back-office services for private credit and receivables funds. Think Celcoin, Pismo, QI Tech and Kanastra, and those are just some of the examples.

For small businesses, there’s a huge opportunity in billing and finance software. Think of something like Xero, with PIX built into the invoice. Asaas lets a business send a payment request with a due date and see exactly which invoice was paid when the money comes in. Jota takes it into WhatsApp. It runs a business account there, so you can start a PIX payment by sending a text, a voice note or a photo of a bill.

Take that across a border and you get another business. EBANX lets companies outside Brazil accept PIX from Brazilian customers, including for subscriptions. A customer can authorize the recurring charge once and pay from their Brazilian bank account each month. PagBrasil lets Brazilians pay participating merchants abroad through PIX, with licensed foreign-exchange partners handling the conversion.

Then there’s PIX-native credit. Pagaleve lets shoppers buy in installments through PIX without a credit card. It’s a multi-hundred-million-dollar business built around something really Brazil-specific.

And the list goes on. The financial services ecosystem in Brazil is unbelievably robust. Together, these companies have raised billions of dollars in venture funding and generate billions in annual revenue. They also represent some of the biggest success stories in Brazilian tech. Think Nubank, which became one of the world’s largest digital banks and is now entering the US. Or Pismo, which Visa bought for US$1 billion in one of Brazil’s biggest tech acquisitions.

That’s how big the impact already is. But it doesn’t stop there. The first generation of fintechs used PIX to move money. Think about what the next generation could build around those payments.

Fraud, identity and recovery are some of the most interesting spaces to build and invest in. Fraud scales with the rails. In the first nine months of 2026, Brazilian banks confirmed over US$1.1 billion in PIX fraud through the refund system. Victims got back just 13 cents on the dollar. And what’s really exciting to me is that the opportunity goes well beyond Brazil. Brazilian banks are already dealing with sophisticated fraud at a scale few markets have had to contend with. If you can build software that works in that environment, there’s a good chance it’ll be relevant to banks and fintechs facing the same problems around the world.

And then there are agents that can run payables and collections, with chat as one way to tell them what to do. Add cash-flow lending, installment credit and supplier finance, and there’s still a lot to build. Time will tell how big that pie really is.

For companies operating across countries, moving money gets expensive and complicated fast. Every new market can mean another bank, another currency, another payment rail and another layer of reconciliation.

Jeeves is built to remove that complexity. Its stablecoin-native financial operating system brings payments, FX, corporate cards and treasury into one stack — allowing global companies to move and manage money across markets without rebuilding their financial infrastructure country by country.

For Orinter, one of Brazil’s largest B2B travel operators, the impact was 87% lower international transfer costs.

The less it costs to move money globally, the more capital you have to grow globally.

So if you’re a founder, CFO or finance leader building across markets, stop building financial infrastructure country by country. Go to tryJeeves.com to see what a global financial OS can do for your company.

I sat down with Vitor Olivier, Nubank’s former CTO and now co-founder and CEO of Decade, to talk about what founders get wrong about growth, AI and the future of finance. At Decade, Vitor is building toward what he calls “self-driving money,” money that can act on your behalf and look out for your interests.

We get into why giving people access to banking didn’t automatically make their financial lives better, why Brazil is a strong place to start building AI-powered financial services, and what’s holding those products back even as the models get better. We also talk about why having the money to scale doesn’t mean it’s time to scale, and whether global investors actually have a Latin America strategy. Check out the full conversation on Spotify, Apple Podcasts and YouTube.

Nubank posts a record $1.1B quarterly profit and reaches customers in all 50 U.S. states.

Gross revenue rose 37% year over year on a currency-neutral basis to $6.3B, and return on equity reached 32%. Nubank ended the quarter with 143M customers globally, including nearly 120M in Brazil, 16.6M in Mexico and 6.1M in Colombia. Its U.S. offering launched in September with a savings account, a cashback credit card and fee-free transfers to Brazil, Mexico and Colombia.

Clara announces $105M in financing to expand its corporate spend platform across Latin America.

The package includes $65M in Series C equity and a $40M renewal of a growth-financing facility from General Catalyst. Citius, Kaszek and Germany’s DEG anchored the equity round. Clara plans to use the financing to expand and build AI tools for company finance teams.

U.S. venture firm Canaan leads a $24M Series A for Brazil’s Nova.

Nova lets customers use their existing credit cards to pay expenses like rent and tuition through PIX. The company says its app has been downloaded more than 500,000 times and it’s processing over $179M in annualized payment volume. Nova was founded by Rafael Matos, a former Neon executive, Melissa McCoy, who led product for the Google Pay API, and Jean Sigrist, Neon’s former CEO. Propel Venture Partners, Canary and Monashees also joined the round.

dLocal processes $18.3B in a quarter, up 77%, while revenue grows 41% to $398M.

The Uruguayan payments company connects global merchants to local payment methods in emerging markets. Ride-hailing, delivery, financial services and remittances helped drive volume growth. Revenue was roughly flat from the previous quarter.

Amazon opens its first AWS cloud region in Chile as part of a $4B-plus investment plan.

AWS opened the region to select customers on November 11. It’s the company’s third in Latin America after Brazil and Mexico, with three separate availability zones in the Santiago metropolitan area. Amazon first announced the investment plan in 2025 and expects to spend more than $4B over 15 years.

Read — Refounding Ro to $3B+ Run Rate, by Zachariah Reitano.

A really interesting turnaround story. In 2021, Ro was a walking zombie. It was growing 9%, burning cash and still had most of the roughly $800M it had raised. Zach says they’d hired too fast and tried to do everything at once. In 2022, they cut the team, froze hiring and narrowed what they were building. When demand for GLP-1s took off, Ro had the clinical and pharmacy operations to serve it. Today it’s at a $3B-plus revenue run rate with fewer employees than it had in 2021. Great read.

Read — What Shaped Anthropic CEO Dario Amodei’s Thinking on A.I.

In 2007, after reading Ray Kurzweil at Princeton, Dario Amodei told a friend he thought AI could surpass human intelligence by 2040. His father had died the year before, and Amodei had thrown himself into studying the retina and thinking about how to speed up scientific discovery. The Times traces how that urgency collided with another conviction. Amodei thought AI could be dangerous enough that people who took the risks seriously had to get there first. That helps explain how the same person could build Anthropic, call for a slowdown in AI development and push back on the Pentagon over how its models could be used. Now he’s taking the company public. How much room will he have to make those same calls when they cost the company business?

Thanks for reading,

Olga 

🎙 The J Curve  is where LATAM's boldest founders & investors come to talk real strategy, opportunity and leadership.