Hernán Kazah on the power law, the AI valuation problem, and why two of Latin America's five most valuable companies are now tech

The J Curve is an English-language podcast and newsletter covering Latin America's technology and venture capital ecosystem, hosted by investor Olga Maslikhova. In this episode: Nigel Morris, who co-founded Capital One inside a small Virginia bank in 1994 and grew it from roughly $1 billion to nearly $20 billion in market value, then co-founded QED Investors and backed Nubank before it existed — after David Vélez pitched him "a Capital One in Brazil" over breakfast in Mexico City. He explains why Nubank, Revolut, Klarna and Monzo are all now aiming at the same market, why banks structurally cannot innovate, why a company growing 40% with an NPS of 89 currently cannot raise growth equity, and what he thinks genuinely changed in Latin America — which is not capital, regulation, or technology.

Why copy eBay and not Amazon or a search portal?

Because it was the only one of the three that Latin America could afford. In 1999 there were three legible internet business models, and Kazah and Marcos Galperin worked through them from Stanford. The portal model required serious server investment and access to engineering talent that wasn't available in the region. Amazon required capex, inventory, warehouses and logistics. eBay required connecting buyers to sellers, which is capital-light — so eBay it was.

The objection everyone raised was trust: would Latin Americans transact with strangers? Kazah's counter was that liquidity and ease of access would overwhelm the trust problem, and he says that's what happened. He is careful not to claim originality for any of it:

"We were not amazing at finding something nobody else saw. When we started there were other groups doing the exact same thing. We had about 80 competitors."

Hernán Kazah

What separated MercadoLibre from eighty competitors?

Three things, in sequence. First, capital — and specifically the ability to raise it. Two Stanford graduates with some family money and good connections could get to Chase Capital Partners, then led in the region by Susan Segal. Kazah's description of what that meant is worth quoting for anyone building in an emerging market today: "That was the Kaszek of the time, because everyone looked up to them in the region." The stamp of approval mattered as much as the money. Capital access alone took the field from eighty to five or six.

Second, the team — but specifically which team. Attracting talent to the internet in 1999 was easy for everybody; the differentiator showed up when the bubble burst. Competitors had hired people with excellent CVs who went back to consulting, banking and corporates when the market turned. At MercadoLibre almost nobody left, and many stayed years past the IPO and became wealthy. His explanation is that they had been recruited on a twenty-year vision rather than a boom.

Third, technology as the competitive advantage rather than marketing spend.

How did a technology mindset become the growth engine?

The origin is more accidental than the legend suggests. MercadoLibre started on off-the-shelf auction software, discovered within weeks that it was inadequate, wrote a long list of requirements and hired a consulting firm to build a better version. That version was obsolete the moment it shipped — so they hired the consulting firm into the company, and that became MercadoLibre's engineering team.

"The perfect product doesn't exist, because you can never reach it. You

aspire to it, but new needs appear, new consumer insights emerge, and new technology makes things that looked impossible possible."

Hernán Kazah

The payoff came when the Nasdaq fell 80% and they had to survive on their own resources. A product-and-technology culture meant they built organic acquisition engines instead of buying growth: an affiliate programme, and then search. Kazah's claim is a striking one — MercadoLibre was Google's first SEM customer in Latin America, and got precise enough at SEO that a search for almost anything in the region returned MercadoLibre in the top few results. They stopped spending on everything else.

What did the eBay partnership actually give them?

Context first: eBay was, in Kazah's phrase, "the star of a very dark era" — the one internet company generating cash when the market turned on the sector and started reading P&Ls. It had inherited Brazilian operations by acquiring a French company, decided it could not attack Europe and Latin America simultaneously, and ran what amounted to a beauty contest to find a regional partner. MercadoLibre wanted it more than its competitors did, because it would give confidence to the team and the investors and access to best practice.

What they got was better than that: observation rights without corporate obligation. They could sit in eBay's meetings and see what it was planning, what it was doing well and what it was failing to do — while remaining independent enough to act on any of it.

The best illustration is a story Kazah tells against himself. Pierre Omidyar visited and praised MercadoLibre's hidden feedback system, which solved a retaliation problem eBay was struggling with: rate someone badly and they rate you badly back, which corrupts the whole ratings system. Kazah accepted the compliment. The truth was that they had heard eBay discuss the idea internally, watched it not get implemented, and simply done it.

Was 1999 the right time to start?

Too early on penetration, exactly right on capital — and Kazah argues the second matters more. Five years later the internet would have been bigger and there would have been no money in Latin America to build with. The 1999–2000 window existed because the US boom spilled over: American VCs were busy at home, so other institutions built emerging-market vehicles to get exposure, and Chase Capital Partners, Goldman Sachs, Santander and GE Capital arrived in the region.

That window shut for roughly six years and reopened around 2006–07, when global investors who had watched Alibaba and Tencent emerge started asking what the next China might be. MercadoLibre's Nasdaq IPO landed in 2007 — a year before the market closed again for the financial crisis.

His generalisation is the most directly useful thing in the episode for a

founder:

“You cannot try to time things to perfection. If you're too early, maybe you don't have enough fuel to wait until the game wakes up. If you're too late, others eat your pie. For a company to be successful you need many things to align — and if you only move when everything is perfectly lined up, you'll miss it."

Hernán Kazah

Which is why he tells founders not to defer a raise for a better valuation. Take the capital when it is available, because the moment you need it may be the moment the market is shut.

Why is your default outcome failure?

Kazah's line about building in Latin America is climbing Everest on top of a rollercoaster, and he unpacks both halves. The Everest is ordinary entrepreneurship: you're building something that doesn't exist, on technology that may not work, for customers who may not want it, against competitors whose moves you only see once they hit the market. Even expert climbers don't control the weather.

"I always tell entrepreneurs: your default outcome is failure. You need to be aware of that, so you work against the odds and make the impossible possible. That's what entrepreneurship is."

Hernán Kazah

The rollercoaster is regional. His formulation of the capital cycle is precise and it explains a great deal about the region's history: when capital is abundant globally, some of it reaches Latin America; when it is scarce globally, there is none in Latin America at all. The excess arrives thinly and the drought hits hard. He adds a note that will interest anyone comparing regions right now — Latin America looks less unstable than it used to, partly because the rest of the world has become more so.

What travels from operator to investor, and what doesn't?

What travels is empathy with the founder's position — the practical problems and the emotional ones, since being an entrepreneur is its own rollercoaster. What doesn't travel is the thing that made him good as an operator.

"As an entrepreneur I was obsessed with winning. I would wake up, identify my closest enemy, and fight that battle with all my strength. Then the next one."

Hernán Kazah

As a fund manager that instinct is actively counterproductive, and he does the arithmetic out loud: every company has five battles, a portfolio has twenty companies, so you have a hundred battles. Try to win all of them and you win none. His image for the change in role is that in the best case you are in the passenger seat and usually you are in the back — in the car, but not driving.

"To build a successful career as an investor you need to be very, very, very right just a few times. Learning that, and living with it, was an effort."

Hernán Kazah

When do you roll up your sleeves, and when do you stay out?

Kaszek did a great deal of hands-on work in its early years — sitting with founders on marketing strategy, product, UX funnels, conversion — partly because it was what Kazah had loved doing at MercadoLibre and believed was productive. What they learned is that good founders mostly don't need it. They need the question: have you looked at this? And then they ignore you, which he says is fine, as long as they listened. Maybe one time in ten the suggestion lands and the founder builds something better than what was proposed.

What he is scathing about is the alternative failure mode, which he has watched on many boards:

"If you sit in a room and just clap and celebrate and cheerlead, you avoid the hard conversation. But what was your value add? Zero."

Hernán Kazah

Kaszek's stated posture is to provoke, from a base of genuine respect: I'll tell you what I think and why, you are running the ship and will decide, and once the decision is made we commit to it regardless of whose it was.

What makes a general partnership work?

Kazah had worked alongside Nicolás Szekasy for a decade at MercadoLibre before they founded Kaszek together, which he treats as the only reliable form of diligence on a partner. You cannot learn from interviews how someone behaves under pressure; you can only learn it by having been under pressure with them. What he watched for was whether they took shortcuts when the situation was worst, and whether they held to long-term thinking when short-term thinking was available.

The partnership's operating principle is the more surprising part, and it inverts what most firms say they want:

"Typically we don't like consensus. When there's too much consensus we get nervous, because there must be an elephant in the room we don't see."

Hernán Kazah

His reasoning: innovation happens at the margin, not at the core. Broad agreement is a signal you are sitting at the core — which is the right place to be if you run Procter & Gamble or Coca-Cola, and the wrong place if you run a venture fund.

When does a fund become an institution?

Kaszek's Fund I targeted $50 million — chosen as the minimum capital needed to build a portfolio at all, expected to come from the partners' own money and their MercadoLibre colleagues — and closed at $95 million in 2011. Kazah notes drily that this is now roughly a seed round for one Silicon Valley company.

The intent was always to outlast the founders, but the method is what he recommends to entrepreneurs: extreme clarity about the long-term destination, extreme clarity about the single next problem to solve, and deliberate vagueness in between, because you will learn things that make plan A irrelevant while leaving the vision intact.

The evidence he offers is unusually testable. At Kaszek's tenth anniversary they pulled out the 2011 deck and presented it to their LPs. Everything in it had happened — a growing regional ecosystem, technology companies appearing on the list of the region's most valuable. It had simply taken ten years instead of the five they projected.

Why is judging people still the hardest part?

Kazah says his mistakes have consistently been about team selection, and he does not claim to have solved it.

"I've been interviewing people for more than 25 years and I still feel I'm average at it."

Hernán Kazah

The thing he wants to know — how someone behaves under pressure, over years — is exactly what an interview cannot show. Reference calls are the best available proxy. The practical adaptation is duration: Kaszek now brings founders in for a full day, meetings in the morning, lunch, more meetings, an early dinner, so that the interaction runs five to seven hours net. Anyone can prepare for an hour, or perform for two. Six is harder.

Looking back at Fund I, he says some of those teams would not clear the bar today — good, committed people who lacked either the depth of commitment or the technology instinct Kaszek now treats as non-negotiable, on the grounds that product and technology are the only durable leverage.

"Successful founders come in all formats. That makes our job harder, because there is no formula that identifies the ones with a better chance."

Hernán Kazah

The pattern-matching exercise that would have killed his best companies

At a year-end review Kazah ran an analysis of every Kaszek investment that had failed, looking for common features. He found them, and presented the list to the team. The response was the obvious one: screen every new company against these markers and kill anything that matches.

Then he applied the same screen to the companies that had worked.

"Applying that algorithm to the rest of the portfolio, we would have ruled out many of the big, big winners."

Hernán Kazah

The lesson he draws is not that analysis is useless but that in this business the failure patterns and the success patterns overlap heavily, so a screen tuned on losses will discard winners. And the asymmetry makes that fatal rather than merely costly: a bad investment loses one dollar, while a missed winner forgoes a thousand or ten thousand.

Which is a more rigorous version of a thing many investors say loosely — and it is the passage to send anyone who thinks venture can be run as a checklist.

At what point can you no longer sell a deal to yourself?

Olga asks what the internal conversation sounds like when Kaszek loves the team, the company and the market, but the price makes the maths more art than science. Kazah opens with "you're touching a very sensitive point," and says they are having exactly this discussion in investment committee repeatedly.

The AI opportunities in front of them look extraordinary on the old criteria: small bright teams, revenue arriving far faster than it used to, genuine customer enthusiasm, and unit economics that work from the start rather than promising to work at scale. Historically Kaszek invested pre-product-market-fit and hoped; now the evidence is there early.

Two things stop them anyway. The moat looks thin. If building software keeps getting easier, the right to win erodes — and in many cases the customer demanding the solution today could plausibly build it themselves tomorrow. And the price has moved faster than the progress. His numbers on the seed-to-A step are the sharpest data in the conversation: revenue up perhaps 20, 50 or 100%, valuation up five to ten times, landing at $100–500 million pre-money — with none of the original question marks resolved.

So Kaszek is deliberately barbelling: more seed, where prices are elevated but still in a reasonable zip code and a right answer overcompensates for the entry price, and more Series B, where sustainability is demonstrable. They are walking away from Series A rounds — not on the team, not on the opportunity, but on risk-reward.

What makes an AI company defensible — and is SaaS dead?

If the ability to build the software is no longer the moat, Kazah looks for two substitutes. A local friction point: a licence, a permit, a regulatory right to operate, or access to a dataset others cannot get. Entrenchment: being embedded in a process such that unplugging it is expensive, including the cost of retraining the customer's own team.

He walks through the failure case concretely. A vertical AI tool solves an isolated process well, the CFO adopts it immediately because it beats the brute-force alternative — and then a year or two later someone asks what the $2 million a year is buying, whether it's just running on the company's own data, and whether they could build it. If the answer is "with some effort, yes," there was never much there. The defensible version brings knowledge from outside — a system trained on many customers' data, not one — and is painful to unpick.

On SaaS he is contrarian and explicit that it's a bet. Multiples have compressed on the assumption that AI makes every SaaS product replicable. He thinks the market is underpricing how much value sits in understanding a problem deeply, solving it well, and integrating with everything around it.

"If you pick the right one, SaaS could be a great arbitrage opportunity — something mispriced today because the market thinks replacing that software is easy."

Hernán Kazah

The same instinct governs his answer on valuation gaps between comparable companies. Asked about two similar businesses trading at wildly different multiples, his response is discipline rather than opportunism: "My first reaction is to try to understand why, rather than to think there's an opportunity."

Does Latin America have an exit problem?

Kazah's answer is a direct rebuttal to the standard objection, and it rests on his own history. When MercadoLibre was being built there were no exits at all — they assumed eBay would acquire them, it didn't, so they went public anyway, at a size and in a market where nobody had done it before. His argument: if it could be done then, it can be done now.

The evidence has moved. Two of Latin America's five most valuable companies are now MercadoLibre and Nubank, where a decade ago the list contained no technology company at all. He expects three or four of the top five within five years, and something like nine of the top ten within ten to fifteen — with a note on how long it takes: MercadoLibre needed twenty years to get there, Nubank about eight.

Part of Kaszek's underwriting is asking whether a company could become a public company attracting global investors, with a US listing assumed because of the depth of the market and the analyst coverage. He is blunt about the hit rate: wrong nine times out of ten, which is fine, because once is enough.

On the window, he expects Latin American listings in 2027 and 2028 rather than immediately, and thinks the market genuinely reopens if the largest US private companies go public. Kaszek has portfolio companies that are profitable, growing, and waiting.

What does compounding actually feel like?

Kazah says he was always good at maths and understood compounding theoretically, and that living through it is a different thing entirely. His illustration is uncomfortable for his own asset class. MercadoLibre went public at $18 a share and trades around $2,200. Private investors who came in around a dollar and exited at eighteen did well. Public investors who bought at eighteen have done far better — and despite holding for longer, at a much better IRR.

The Nubank episode is the practical case study, and it is the most useful passage in the archive for anyone thinking about distribution policy. When Nubank went public the market fell, and Kaszek judged it was being mispriced for reasons unrelated to the company. So they held, knowing many LPs sell public shares on receipt as a matter of policy. Some LPs were pleased and assumed an information edge; others objected that they had bought exposure to private companies and wanted the shares. Kaszek eventually distributed in instalments rather than all at once.

His assessment is even-handed in both directions: distributing immediately would have produced lower returns, and holding until today — Nubank hit a record price the week of the recording — would have produced higher ones. He is careful not to claim genius, and names the structural limit honestly: once the company is public, our investors can buy it themselves. They don't need us.

How would you sell Latin America to an American allocator?

Asked to make the case to a global investor who mostly reads macro headlines, Kazah declines to lead with a tailwind — that claim, he says, might last a day. He builds it in four parts instead.

The technology opportunity is larger in Latin America than in developed markets, because the gaps are larger. Commerce, finance, education, healthcare — technology doesn't close them entirely, but it narrows them, and the value created by narrowing a large gap exceeds the value of optimising a small one.

Monetary policy has been more conservative than in many developed markets over the past decade — central banks more sensitive to inflation, more careful about printing. He presents this as a genuine and underappreciated reversal of the region's reputation.

Supply chains favour the region. Shortening supply chains predates the pandemic, accelerated through it, and matters more under US-China tension. Latin America is close to the US, close to Europe, a natural trade partner that he thinks was over-ignored for two or three decades — and it has the commodities, mining and energy everyone is now worried about securing. His comparison is to Africa, which he says offers similar land, minerals and energy with less rule of law and more instability.

And the geopolitical tensions defining the rest of the world aren't present here. He notes elections across the region have trended pro-market, with the caveat that swings have been large everywhere lately.

"Latin America is one of those regions where you can do well and do good at the same time. You do not need to choose."

Hernán Kazah

What would you tell a founder in a garage in 2026?

Three things. Dare to have very wild dreams — he puts this first. Hold the long-term vision even while solving the immediate problem, because time and resource pressure make founders collapse into the next task. And then a warning he says he sees constantly:

Resist the temptation to over-innovate."

Hernán Kazah

His version of the problem is precise. Getting from zero to 95% is hard and exhilarating. Getting from 95 to 100 takes far longer and is boring, and it is where the durable value actually sits — so entrepreneurs, who became entrepreneurs because they like creating things, drift toward the next idea just as the current one is about to pay.

"When you think your job is getting boring, that's when you're really making it."

Hernán Kazah

Rapid fire: five questions

The most underappreciated founder skill in Latin America. Courage.

What he underestimated about building Kaszek as a business. The importance of portfolio management.

What he believed about venture that was completely wrong. That coming from a successful company meant returns would be spread across many investments — that they would pick well and help well enough to get ten winners out of twenty. The power law does not care. If you are a good investor, most of the result comes from one or two per fund; if it doesn't, you aren't one.

The question founders should ask investors and don't. What happens when things go badly. Partnerships are easy in good times, and character only shows in bad ones — a point he makes from experience, since MercadoLibre nearly died twice, once because some of its own investors wanted to shut it down.

The most under-the-radar company in the region. He resists — it's like naming a favourite child — then picks one from the newer cohort that he says doesn't get enough credit, praising the team, the leadership and a product-and-usability mindset.

COMPANIES AND INSTITUTIONS

⠀PEOPLE

⠀IDEAS AND FRAMEWORKS

Co-founder and managing partner of Kaszek, the largest venture firm in Latin America, which he started in 2011 with Nicolás Szekasy. From a $95 million first fund it has grown to nine funds and more than $3 billion raised, backing over 130 companies — Nubank, the world's most valuable digital bank outside Asia, QuintoAndar, Kavak, Creditas, Loggi, Nuvemshop, MadeiraMadeira, Bitso, Konfío, Clara, NotCo, Wellhub, Kushki, Cora, and Technisys, acquired by SoFi. Before Kaszek he co-founded MercadoLibre in 1999 with Marcos Galperin and Stelleo Tolda, serving as Chief Operating Officer and later Chief Financial Officer through its 2007 Nasdaq IPO.

Stelleo Tolda — how MercadoLibre beat Amazon in Brazil · Nigel Morris, QED — Capital One, Nubank and what actually changed in Latin America · Stelleo Tolda — How MercadoLibre beat Amazon in Brazil · Brian Requarth, Latitud — why ~50% of LatAm founders now go global ·

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