Summary

The operating system Escobari took from the 3G founders is spearfishing: decide where to anchor years in advance, “let little fish go by,” and strike in seconds when the big one appears. The three founders—likely Jorge Paulo Lemann, Marcel Telles and Beto Sicupira—picked beer five years before Brahma came for sale, closed in a week when Swiss owners panicked before an election over Lula, and turned $80M into $60B-plus through Antarctica, Interbrew and Anheuser-Busch. “Every four or five years there’s a once-in-a-generation opportunity that you have to be ready and be willing to move quickly to capture.”
On the AI bubble, his direct answer to whether the top is in: “It’s not crazy enough.” Capex-to-revenue is still tame versus railroads and dot-com, revenues are materializing, and the spend is funded by “the magnificent six who are printing money” — not junk-bond speculators or levered telecoms with retail money. An “invested too much” morning is coming — “I don’t know if it’s three years away or 18 months” — but for sure it hasn’t happened yet.
General Atlantic’s cycle discipline — “we try to make different mistakes in each bubble” — is to deploy AI across 200+ portfolio companies first (500 projects this year, a third AI) and pounce only on proven ROI. First pounce: code generation, where coding revenue at Anthropic went from $200M to over $4B in 12 months — “that kind of growth has not happened ever.” He concedes the approach “may prove to be too conservative.”
The valuation call: the US has never been this expensive — 26x earnings for 4% forecast growth (97th percentile of 25 years), the dollar two standard deviations from neutral, debt at 125% of GDP heading to 145%, “higher than Greece and Italy,” with no recession since 2009. Against that: Europe at 14x, Brazil at 9x, Mexico at 10x, and “40–50% growers at 12–14x EBITDA serving dollarized clients.” “The case for global diversification has never been stronger; the price for global earnings has never been lower on a relative basis.”
GA’s risk product is the anomaly: a 4% loss ratio on capital versus the 20–40% common in venture and growth, built by refusing binary risk — the modeled worst case is “a company grows into the valuation we paid for it.” Escobari keeps 95% of his own net worth in two assets: GA funds and treasuries.
Structure as edge: hybrid evergreen fundraising (no cliffs, no forced selling into risk winters), GA as the largest investor in its own product, with employees holding ~8% of AUM ($5B+), a “communist system of compensation” policed by “if you’re not pulling your weight, you’re not on the boat” — and deliberately no Latin America fund, because a dedicated fund “buys at the top and sells at the bottom.”
Two people calls: if you have a young mind, go work in AI now — “dog years,” seven years of learning in one, “compressed learning that only happens once every 20 years.” And he has never met a top performer without trauma-based drive: “Still looking for her or him. Zero.”

Deep dive

1. Spearfishing: pick the anchor five years early, close in a week

The framework comes from a 2003 book he wrote with Harvard mentor Don Sull, pairing ten hugely successful ’90s Brazilian companies against ten look-alikes that started similar in size and ended at least 5x apart. The defining pair: Brahma vs. Antarctica — Antarctica the “better, more profitable, more valuable company” in 1989, yet when they merged a decade later, Brahma’s shareholders kept 95% of the equity.
The metaphor as told: “You drop down with no equipment other than the spear… You let little fish go by because you’re not there to hunt little fish. You’re waiting for the big fish” — and near the end of your oxygen you get two or three seconds to strike. Step one is deciding where to anchor.
The mechanics: the 3G founders were minting money on inflation at Banco Garantia, knew it would end, and decided five years early they wanted a low-inflation, rising-consumption asset — beer. The Swiss owners called two weeks before an election and panicked over Lula: “Can you do a deal in a week? We’d like to get out of town.” Antarctica followed ten years later in three months amid a devaluation, then likely Interbrew, then likely Anheuser-Busch: $80M became $60B-plus excluding dividends.
The distilled rule: “Every four or five years there’s a once-in-a-generation opportunity that you have to be ready and be willing to move quickly to capture.”

2. Bubbles run on dog years — seven years of work in one

His first mission from the 3G founders: invest $500M in 20 companies, “you got 18 months. Go.” By month three he’d defected to the bubble itself, after his friend Dennis — at a dot-com trading at 20x revenue — grabbed his hand: “Martin, it feels better than being outside the bubble.” Within three months he’d raised $80M for likely Submarino.com, “Amazon.com merged with Alibaba to take on Brazil.”
The dot-com’s real lesson, which he says he only recently articulated: “You can do seven years of work in one year.” Submarino opened in six countries with warehouses and teams in twelve months — a three-to-four-year job at normal speed. When Elon Musk says compress the ten-year plan into one, “he’s like, crazy. Look what he’s built.”
The GFC replay: everyone dropped out of bidding for a dominant Brazilian fixed-income exchange with 80% margins, so he doubled down and closed in two months. “If we’re not willing to buy a dominant platform at six times EBITDA… it means the world is ending. The world is not ending.”
Hence the “unambiguous recommendation”: if you have a young mind — at any age — go work in AI. “You’re going to live through dog years… compressed learning that only happens once every 20 years,” regardless of whether that company wins.

3. GA’s AI playbook: deploy in the portfolio, pounce on proven ROI

This is his “fourth or fifth bubble,” and the shape never changes: “the promise was spectacular, the short term was disappointing, and the long term delivered more than expected” — with fortunes made and destroyed in between. GA’s 45-year discipline: “we try to make different mistakes in each bubble.”
The mechanism: 200+ portfolio companies, a 100-person operations team, 500 portfolio projects this year — a third of them AI — watching the front lines for use cases with real ROI and real revenue to the provider. “And then we pounce.”
The first pounce is code generation (Cognition and Cursor come up): coding revenue at Anthropic went from $200M to over $4B in twelve months of B2B — “that kind of growth has not happened ever.” The new reality is hyperproductive humans “working alongside agentic programmers who have no moral north star and do not sleep.” Also live: marketing optimization (Liftoff, Insider) and AI-turbocharged data (an Israeli company he calls VI).
The hedge, kept intact: “we’re probably going to look back and say we weren’t bold enough in going for the killer app soon enough — but we’ve been bold before and it didn’t pay to go very early.” Meanwhile Bill Ford is pushing him into robotics and humanoids: “I’m like, Bill, too early. No, we need to go.”

4. “It’s not crazy enough” — why the top isn’t in yet

Patrick pushes directly: couldn’t we already be at the we’ve-invested-too-much moment? Flat no — “it’s not crazy enough.” Against railroads and the dot-com, capex-to-revenue is “still not crazy” and new revenue streams are materializing.
The structural difference is who’s funding it: “the magnificent six, who are printing money out of their dominant positions,” not “junk-bond speculators or thin-margin telecom companies levering up the wazoo with retail money. It’s really profitable companies. So I think it’s got more legs. Will it be bumpy? Yes.”
He still expects the hangover — “I don’t know if it’s three years away or 18 months” — and his doctrine for the fog in between: it slows you down but doesn’t excuse paralysis. “You just have to engage with the unpredictability until you see something before others — and you strike for the fish.” You never invest big without visibility; you also never shut down because it’s “too unpredictable.”

5. The US has never been this expensive; the rest never this cheap, relatively

The numbers: US equities at 26x earnings for 4% forecast growth — the 97th percentile of the last 25 years — with the dollar two standard deviations from neutral despite a 10% fall this year. Debt at 125% of GDP is the highest in the OECD, above its post-WWII level, and on current plans hits 145% within five years — “higher than Greece and Italy.” No recession since 2009. “Are you sure you want to have 95% of your assets in the United States of America? I don’t.” He still wants half there — just not 90%.
The other side of the ledger: Europe at 14x earnings, Brazil at 9x, Mexico at 10x, and GA is “finding 40–50% growers at 12–14x EBITDA, many of them serving dollarized clients.” “The case for global diversification has never been stronger; the price for global earnings has never been lower on a relative basis” — over ten years, “those who achieve some diversification will be rewarded.”
His pitch to local families holding 90–95% of total net worth at home: “Close your eyes. Imagine you’re not Brazilian — what percent would you put in Brazil?” Answer: 3%. “How about I help you get a little closer to three than the 95 you’re in.” People overinvest in what they understand, and “diversification is the only free lunch.”
Ex-US craft: far more volatility, so agility; low-trust cultures, so references are much more important — the hack is asking a family whose money you manage (“No way. He’s a crook, son of a crook” — because they have money with you, the truth comes out); and the opportunity: humongous lower-hanging fruit. On China: underweight five years, “we just did two deals,” picking it up as he believes tensions have stabilized — binary geopolitical risk intact.

6. Feeney’s DNA: dolphins in a sea of sharks

Founder Chuck Feeney — “the accidental billionaire” of duty-free — answered “what is the purpose of wealth?” with: improve the human condition now, not tomorrow, because “the present value of a happier life for more people now is very valuable.” His dream: “my last check will bounce.” GA exists as the wealth-creation engine for giving it away — 500+ companies over 45 years, half outside the US.
The founding documents commit to being “good partners to each other, to our founders, and to our clients” — hence “sometimes we’re accused of being dolphins in a sea of sharks. And I love to be a dolphin. Who wants to be a shark? Dolphins have a much better life.”
His answer for why GA makes money in Latin America: “we don’t have a Latin America fund.” A dedicated fund forces you to “buy at the top but sell at the bottom”; routing regions through one global IC does the opposite — but that only survives when “the culture demands good partnership,” which is why it is hard to replicate.

7. Structure as edge: evergreen money, communist comp, 19,000 competitors

Alignment first: employees hold about 8% of the funds — over $5B of their own capital. “This does not feel like managing other people’s money… I’m managing my family wealth first and foremost.” His own book is two assets: GA and treasuries, 95% of net worth.
The five-year fundraising cycle is the industry’s hidden distortion: after a three-year risk winter, “you’re out of dry powder exactly at the time things are on sale.” GA’s hybrid evergreen — normal funds every two to three years plus always-open managed accounts into the same forward portfolio — kills the cliffs. The costs, honestly stated: onboarding “is a painful experience” and fundraising never stops, versus competitors’ once-a-cycle “binge dieting.”
Compensation is “a communist system” — everyone paid on total, not individual, performance. He hated it: “I’m a spearfisherman… this communism didn’t work in the Soviet Union.” Then he watched the collaboration it unlocks; the check on free-riding is “if you’re not pulling your weight, you’re not on the boat.”
The context: 19,000 GPs — “more GPs than McDonald’s in the United States” — so “you can’t be a generalist anymore.” GA’s answer is its power alleys (18 across five sectors in one telling, 16 in another), showing up with 32 case studies, the 100-person ops team, and 15,000 vetted executives on tap. “You can copy things that work, but you don’t know the things that didn’t work that we tried.” Loss ratio through all of it: 4% of capital, versus 20–40% norms — because “we don’t take binary risk.”

8. Trauma is the engine — in China, in Bolivia, in everyone who wins

The China unlock came over drinks with an entrepreneur holding an anthropology PhD: his explanation was that this generation of founders “are all children of the Cultural Revolution. Everything was taken away from these families… they think something was stolen and they will get it back” — drive he likens to post-WWII refugees, and the claim applies to 98% of the entrepreneurs there.
His own stack: his mother’s Jewish family fled the Russian Empire; his father’s landed Bolivian family lost everything in the 1952 revolution, house and farms burned. Then 1980s Bolivia: seven presidents in ten years, four coups, 35,000% inflation — his mother grounded him for a year after her salary lost half its value in the hour he paused while exchanging it. A bruising disorder made every activity “a risk-reward tradeoff since the age of five… I know how to price risk.”
Patrick’s test: ever worked with a great performer with no trauma-based drive — just well-adjusted and happy? “No. Still looking for her or him. Zero.” It needn’t be rags-to-riches: one fierce competitor spent a lifetime driven by mowing lawns for the buddies who got the girls. “I don’t judge. That’s a real trauma. Pain is pain.”
Harnessing it “requires either therapy, writing a journal, or meditating — you should do two of the three” (he does two; meditation is the third). His mentoring is the same medicine at speed — “crash therapy… three years of therapy in one hour” — a habit begun at his own darkest post-dot-com moment, when Endeavor’s Linda Rottenberg told him “it is precisely at your darkest moment that you mentor.” Billionaire mentees now say “you really hurt me, but thank you.”

9. Educated intuition: run the checklist, then close your eyes

Newly chairman of GA’s investment committee, he asked both elder statesmen for a framework. Steve Denning: build a checklist from the best deals. Dave Hodgson, the same day: “avoid the temptation to use a checklist. If it was as simple as a checklist, we wouldn’t get paid millions of dollars.” His discussion references Thinking, Fast and Slow, The Checklist Manifesto, and Israeli Defense Forces checklist work: the best super-interviewer completed the checklist, closed it, closed her eyes and went with her gut — “perfect scores.” The synthesis he uses: “educated intuition.”
The checklist itself: huge TAMs, business models that create economic value and have moats, teams with go-forward capability, inorganic growth to get, and strategic value — “someone will overpay to have this capability if we are successful.”
His personal filter — deals must make the world better — has paid: XP, backed “when there were only 80,000 people that owned stocks in Brazil” and now 10 million do ($10B market cap); an edtech in northeast Brazil — a “dark alley” outside the power alleys — that went from 80,000 to 8 million K-12 students; a fraud platform used by 97% of Brazilian financial institutions, mentored through Endeavor for eight years before it was investable. “It’s beyond money. It’s energy.”
On heart: he’s never met a great investor who is mostly heart — heart is for moving thousands of people, not picking. “Part of being a good investor is to not fall in love… love is a treacherous thing.” The one time he skipped the checklist was his wife, who “didn’t score very high” on it: “I had the wrong framework. She was perfect in every way.”

10. Inside the IC: 190 people in the room, a robot with a vote

Every Tuesday the IC is open to the whole firm — about 190 sign-ups, materials out by Friday, no presenting: “we come in directly to Shark Tank, just questions.” His junior-years trick was predicting each committee member’s questions and votes, reaching 80–90% accuracy within a year — “apprenticeship on steroids.”
The sixth member is the IC robot, trained on 45 years of GA data; she has voted on every deal for three years and backtests “much better than humans.” His caveat, verbatim: “someone who’s been trained in the past is very good at the past” — he’s hoping she’s better than him by the time he retires “in about 10 years.”
What he probes as chairman: the founder’s motivations, the true basis of competition, and above all the tails — because in 10% of deals GA gets 5x-plus and “all of them are better than the upside case in our memos… good things happen that we did not see coming.” His standing question: “where are the lottery tickets?” The host then asks whether the CEO is someone who can spearfish.
What seniority changes: patience — “nothing scares me and I know a big fish will come. Calm down. No pressure” — plus the coach-player flip of living vicariously through younger partners (“Tom Brady likes to be on the field”). Hodgson’s recipe for a young brain in his late 60s, which Escobari has adopted: “I refuse to think like an old man… I’m always learning, experimenting, and playing.”

Verification Notes

Raw captions garble the phrase after “Anthropic revenues” (“and coaching”); whether it says coding revenue is unresolved.

The Interview: How Wikipedia Is Responding to the Culture Wars Market Overview November 25, 2025