Paulo Passoni on why speed is the last moat, and why CloudWalk trades at an eighth of Ramp's multiple

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 is speed the last moat?

Passoni credits the idea to Aravind Srinivas of Perplexity. The argument runs against the consensus that data is the moat. Srinivas concedes data matters — Google's search history, YouTube, Gmail are proprietary and valuable — but argues that on a ten-year horizon even those are not sufficient to win. What is sufficient is shipping product faster than anyone else.

"Perplexity ships a new product every 90 hours. The culture required to ship at that speed is unique."

Paulo Passoni

The story he tells about how he learned it is the best illustration. Valor ran an investor webinar with Srinivas and collected the questions nobody could answer. Srinivas fed them to Comet — which has access to his email, his Slack, everything he has said publicly — and instructed it to answer as him. He sent the output back unedited. Passoni forwarded it to the investors, because it was good.

What does Perplexity do that almost nobody copies?

Three people in the entire company can approve a hire, Srinivas among them. If you want to hire someone, you have to prove to those three that you genuinely need a human being and that technology cannot do the job.

Passoni's reading is that this is not a cost measure. It is a forcing function: by deliberately restricting the ability to add headcount, you compel the organisation to reach for technology first. The speed is downstream of the constraint.

He connects it to a broader shift he says is visible across Silicon Valley — a move from a nine-to-five culture to something closer to China's 996, driven by a simple fear: if I don't build this now, someone else will. He notes Jensen Huang and most other technology leaders now talk about working every day of the week, and frame it as passion rather than work.

What is "eternity as a competitive advantage"?

This is the idea that organises the whole conversation, and it came to Passoni from three directions at once: from Luis Silva at CloudWalk, from a Swiss watch executive, and from his own mother.

Silva introduced him to the Long Now Foundation, whose emblem is the mechanical clock Jeff Bezos funded inside a mountain in Texas — engineered to run for 10,000 years with no human intervention. Passoni's argument is that the people who build durable businesses share the mindset the clock represents, and that this predates technology entirely: Buffett, the Walmart founders, the Costco founders.

"You are only going to work every day of the week and not burn out if you believe what you're doing is leaving a legacy for future generations."

Paulo Passoni

He is explicit that this is the thing he most wants to change about Latin America, and that it is a consequence of macro volatility rather than character. When the environment is unstable, founders rationally take the exit — sell at a hundred million, bank it, start again. He says he respects it and won't fund it, because a founder on a lifelong mission attracts different talent and has different stamina.

How do you detect that in a founder you have just met?

Two tests, and the second is a deliberate trap he explains on air.

First, the why. Most founders have never encountered eternity thinking, so he isn't looking for the vocabulary — he's looking for whether there is a mission underneath the business and whether the person can articulate why it matters to them beyond the money.

Second, he offers them secondary. He will ask whether the founder would like to sell some shares so he can increase his ownership, and the right answer is no — or a small yes, framed narrowly, to put children through school. His reasoning is arithmetic rather than moral: at Brazilian interest rates, $20 million in the bank generates enough that the hunger is gone, and the company becomes something the founder can afford to lose.

What he actively wants is the opposite instinct — a founder who resents dilution, wants to own as much as possible at the point the company turns cash-flow positive, and wants super-voting stock so they can never be removed. He notes this is a view he changed: at Third Point he was an activist investor removing underperforming CEOs. For technology founders he now thinks control is a feature, while conceding the obvious risk — if a founder with control checks out, there is nothing anyone can do.

He also looks for low ego about failure, and distinguishes between the rehearsed answer and the real one: "It's so popular now that you have to tell an investor you made mistakes... but the guys who really internalise them, you can feel the pain in their voice."

Why might Anthropic be the safer bet than OpenAI?

Passoni's framing is about dependence on external capital rather than about technology. Anthropic has a much shorter path to profitability and is focused on enterprise; OpenAI's path requires continuing to raise billions. He is explicit that he admires what OpenAI has done and thinks it has already changed the world — but that a plan requiring other people to keep funding it carries a specific fragility.

"If God forbid the revenue decelerates a bit, then nobody gives them money anymore."

Paulo Passoni

It is the same test he applies to founders: a business that cannot eventually generate cash is not a forever business, because it is permanently dependent on other people's agreement that it is attractive.

How much time does B2B software actually have?

More than he thought a year ago, and he says so directly — in the first episode he estimated two years, and now says perhaps five. The reason is customer behaviour rather than technology: enterprise customers are sticky, switching takes effort, and they tolerate a solution that merely works.

The technical constraint is accuracy. You cannot yet tell a model to build you an HR system and fire your vendor. And in domains where mistakes are unacceptable — compliance, moving money — hallucination is disqualifying in a way it isn't for a consumer support call, where you apologise and move on.

His bear case is not about the product at all. It is about talent: AI expertise remains scarce and is being hoarded by companies that can pay the most, which are the ones scaling fastest. He recounts a dinner with public-market software investors who were, in his word, depressed — and notes the chart doing the rounds, that semiconductors are eating software, having themselves been dismissed a decade ago as a cyclical business with unclear returns.

The warning he leaves is the one he keeps returning to: the fact that customers have not fired you yet is not evidence that you are safe.

“"You don't see it, you don't see it — and one day your old way of doing things falls off the cliff.”

Paulo Passoni, on the chart where horse use held for decades of engine improvement and then collapsed

Why is he bullish on the application layer when everyone else isn't?

Because that is where stock picking still matters. In the layers below — chips, data centres, models — the winners are known and the expectation is already in the price, so there is little delta available against what everyone knows. The application layer has far more companies, far more losers, and therefore far more room for a right answer to be worth something.

His selection rule is a single line: invest in the best-of-category or don't bother. He'll only back a fourth or fifth player if it has a technical innovation the leader lacks. He names coding as a category he avoids for exactly this reason — developers switch to the best tool immediately, so leadership changes hands too fast. Voice he is comfortable with. And he flags localised verticals — legal, accounting, healthcare — as structurally more defensible precisely because they are not globally scalable: what you lose in size you gain in protection from a US competitor.

The valuation test underneath it: at a very high entry multiple, how long until the company grows into ten times his cost? Two years is acceptable, one is excellent, five loses money because dilution and lower future rounds catch up with you.

What he got wrong about ElevenLabs

He tells this against himself, which is what makes it useful. He entered 2025 having told his LPs he would not do an AI investment because valuations were absurd. Then ElevenLabs came in at 30× revenue on roughly $100 million of revenue, and his reasoning was, essentially, thirty is not a hundred.

The investment committee had two doubts. Would Google or OpenAI build something similar and destroy them — the model assumed ElevenLabs would decay from 80% market share to about 20% over five to ten years. And how long to grow into a reasonable multiple — the model said two years to reach 10× revenue.

A year later it holds 80% share, unchanged, and it reached the multiple in six months, having gone from $100 million to $300 million of revenue. His explanation for the growth is focus: a horizontal product cannot build the details a sophisticated enterprise customer notices, so a vertical leader compounds — you ask what people use, they say ElevenLabs, it works, you stay.

"Being bearish because other people are making commentaries on social media, so you sound smart, is precisely what makes you dumb. You want to sound smart by being bearish? Go ahead, you'll get applause. But that's not how you learn, and it's not how you make money."

Paulo Passoni

What actually worries him about the Magnificent Seven

Not the valuations — he says they are not expensive relative to the cash they generate. What concerns him is what they do with that cash. Nvidia is not buying back stock or paying dividends; it is deploying capital into other companies, at a scale that multiplied roughly a hundredfold in a short period.

His analogy is Buffett: generating extraordinary returns is easy when you are small, and once you are in the whale-hunting business you must be patient, because there are not many whales and you cannot deploy constantly. Nvidia now has to hunt whales constantly, and whales are rare.

The second concern is that these companies are no longer only companies.

"Jensen is no longer just a CEO. He's our chief AI czar, in charge of winning the game versus China. When he deploys capital, I'm sure that weight is on his shoulders."

Paulo Passoni

Which makes it, in his description, closer to a development bank job — where returns are one consideration among several. He is careful to say he is not bearish, only noting that the role of these companies in society is completely different from ten years ago, when they simply maximised profit and market share.

On what the US needs in order to stay ahead: speed, and engineers. He points out that a large share of the relevant talent comes from China and Russia, that the country is currently hoarding it, and that immigration policy therefore matters more than it appears to. His reassurance is provisional — nobody in Latin America is yet saying they want to study in China, but he expects that eventually to change.

CloudWalk versus Ramp: same revenue, an eighth of the multiple

The cleanest data point in this archive, and Passoni lays it out as he did to his own LPs.

CloudWalk: $1.4 billion of revenue, roughly $200 million of annualised net income, growing about 100% a year. Valor invested at approximately three times revenue and seventeen times earnings.

Ramp: $1.1–1.2 billion of revenue, not profitable, growing at a comparable rate. Its most recent round was at roughly twenty-four times revenue.

He concedes they are not identical businesses, then gives the explanation anyway:

"How come one thing is 24 times revenue and there is three? It's because people just don't like Brazil right now. That's what it is."

Paulo Passoni

He rates both as very good companies and says the difference is entirely the entry price — he expects to make considerably more money in CloudWalk, and is willing to carry what he estimates as a roughly 5% tail risk over ten years that Brazil goes the way of Venezuela in exchange for the risk-adjusted return.

The secondary explanation is that the market misreads CloudWalk as another PagSeguro or Stone.

Who is CloudWalk actually serving?

Hustlers, in Passoni's term — self-employed individuals rather than small businesses with staff. He argues PagSeguro and Stone created a market by serving small businesses that incumbents thought could not be served profitably; CloudWalk is doing the same thing one layer further down, and the two are routinely confused.

The macro observation underneath it is the good part. Brazilian unemployment is falling while formal employment is not rising — so where did everyone go? They are working: massage therapy, personal training, social-media influencing, side businesses that mostly did not exist twenty years ago. CloudWalk gives that person payments, banking and cash-flow management in one product. He estimates 40 million such people in Brazil, of whom CloudWalk serves six million, and expects AI and automation to produce more of them, not fewer.

Two things made it work now rather than earlier. Tap-to-pay removed the card terminal, so there is no hardware to finance or service and the product is pure software — which changes the economics of serving a one-person business entirely. And CloudWalk got there first. The model travels: its US launch has already produced more revenue in its first year than it did in Brazil.

He also flags, half-seriously, that saying any of this out loud may invite copycats — and tells a related story about PagSeguro's founder paying a regulatory fine rather than publish financials, so competitors would not learn he was making money, right up until the IPO made it obvious.

Why doesn't global capital come to Latin America?

Passoni has spent two years fundraising as well as investing, and his diagnosis has two parts.

The first is a category error. Nine out of ten allocators form a macro view and then express it through the investment — which he argues is simply the wrong instrument for venture and growth. The journey from 0.1% market share to 20% is idiosyncratic: it depends on this founder and this team. Only after 20% does the business become macro-driven. Judging a pre-scale company through a country view is measuring the wrong thing.

The second is an agency problem, and this is the part rarely said out loud by someone raising money:

"Most allocators are not the owners of the money."

Paulo Passoni

The people who commit, in his experience, are of exactly two kinds. Family offices, where the person deciding owns the capital. And people twenty years into a seat at an endowment who have nothing left to prove. Everyone else is early in a career where being different and wrong is career-ending while being conventional and wrong is survivable — so they manage the career rather than the portfolio, and demonstrating access to US venture funds is what builds it. He says he understands it completely, and notes drily that his success rate with an agent is lower than with a principal.

Separate from both, he adds a structural point: the dollar has strengthened for fifteen years, which was a persistent headwind on dollar returns from the region. If that reverses — and he argues US policy no longer wants further real appreciation — the same portfolios produce materially better numbers.

How do you change an allocator's mind?

Slowly, and by showing your work. Passoni's method is to run SPVs alongside the fund, and his description of what that is for is the most transferable idea here:

"When we do an SPV, it's like opening our kitchen. Let me show you how we cook the sausage — my working model, the scenarios, what I think the risks are, how I calculate expected returns. Ask me any questions you want."

Paulo Passoni

The point is not the deal. It is that an investor who watches how you reason about one investment can infer how you will reason about every other one — and may conclude they don't like it, which he accepts. Transparency creates the trust; the trust is what eventually converts. He notes several SPV participants have since come to Valor's annual meeting.

His framing of the stakes: a public-markets investor who dislikes a manager simply redeems. An LP is committed for twenty or thirty years — longer, he points out, than many marriages — which is why the decision takes so long to make. He credits the growth of secondaries as the market's genuine innovation here, giving LPs a way to reverse a twenty-year bet.

What does culture actually mean, and why the first thirty hires?

Passoni's original template is not a technology company at all. As an undergraduate in Brazil, the exemplar of good culture was the 3G founders running what became Ambev — meritocratic and flat, which was radical for a beer company at the time. He says the specifics don't transfer to technology but the primacy of culture does, and that it is decided by roughly the first thirty hires.

Which produces his sharpest piece of advice to founders:

"You should spend half your time recruiting. And you should be more excited that you got someone amazing to come work with you than about raising money."

Paulo Passoni

His reasoning is that hiring is harder than fundraising, and structurally so: an investor is diversified across a portfolio, while an employee is entirely undiversified in you. They are committing time, the only resource that cannot be replenished — and the better they are, the more alternatives they have.

He connects this back to speed. If speed is the moat and speed comes from culture, then changing speed means changing culture, which in many cases means changing people. He does not soften it: you rip the bandaid, because there is no other way.

Why hobbies might be the answer to AI

The most unexpected section, and Passoni means it seriously. His argument starts from a problem: as adults we stop trusting other adults, because most encounters carry an agenda — networking, business, someone wanting something. So the guard goes up and the connection doesn't form. The charts showing how alone people are at the end of life follow from that.

A hobby dissolves it. Go deep enough into something — watches, cars, bridge, poker, art — and you meet people equally obsessed, from entirely different walks of life, with nothing to gain from you. The relationships form fast because the shared passion is the whole basis.

"I think the future of humankind is hobbies. Hobbies create communities that are authentic."

Paulo Passoni

He says that if he were starting a company it would be one that helps hobby communities flourish, noting they are already booming through WhatsApp groups, in-person events and multi-day rallies. And he closes the loop on social media with a genuine nuance: it produced the disconnection, and it is also how he found his own watch community — Instagram to acquaintance to friendship to community.

Which is the counterweight to everything else in the conversation. In a year defined by removing humans from processes, his conclusion is to become deliberately more human.

Rapid fire: four questions

The highest-ROI use of his time this year. Learning to use AI tools — which he admits he cannot quantify, but says matters because tinkering is the only way to imagine where businesses should be going. Without it, he stays in the past.

His three learning inputs. Travel, because he is constantly between countries and reads in transit. X, where he follows a small number of genuinely curious people who post charts worth thinking about. And podcasts, constantly — he singles out Acquired, which SoftBank Latin America sponsored when he was there, letting portfolio founders introduce their own companies in the ad slots rather than the fund talking about itself. What he thinks it teaches: how to tell a story, and how to learn.

What he consumes less of. Everything social. Work, family and his hobbies fill the plate, and he says he has never worked harder than in the last three years.

His goal for next year. Not an investment. He wants to stop losing human connections — and to repair the ones he has lost without knowing. He describes learning that a founder he met once for thirty minutes six years ago has disliked him ever since, which he only found out second-hand, and reaching out on LinkedIn to ask for a reset. He offers his relationship with his ex-wife as the evidence that it works: they have dinners, attend each other's birthdays, and he stays at her house because of their daughter.

"Forgiving people in life is critical for happiness."

Paulo Passoni

COMPANIES AND INSTITUTIONS

Managing partner at Valor Capital Group, the cross-border firm investing between the United States and Latin America. Previously a managing partner at SoftBank's Latin America fund, and before that an investor at Third Point, where he anchored PagSeguro's IPO. His portfolio work spans both sides of the comparison in this episode — CloudWalk in Brazil, and ElevenLabs and Perplexity in the United States. This is his second appearance on The J Curve; the first was the most-listened episode the show released in 2025.

Hernán Kazah, Kaszek — building Latin America's $100B+ companies · Nigel Morris, QED — Capital One, Nubank and what actually changed in Latin America · Stelleo Tolda — How MercadoLibre beat Amazon in Brazil ·

Pedro Conrade, Neon — building a $760M bank for the customers investors avoid

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