Summary

Column is “a software company that also owns a bank” — the regulatory moat lets it build software nobody else can, powering the payments, deposits, and credit behind Built, Wise, Ramp, Brex, and Mercury, plus banks in dollarized emerging markets. Unlike banks, 90%+ of its money is from software, priced per API call, with the bank-side economics passed to customers — and Hockey’s thesis is that fintech is “probably going to be like the last area that is somewhat disrupted by AI,” collapsing into domestic fintech plus enterprise software.
The self-funding story is grittier than “billionaire buys bank”: the DOJ blocked Hockey’s attempted ~$5B Plaid sale to Visa, leaving him paper-rich and cash-poor, so he pledged over $1 billion of stock to borrow $70 million at SOFR+10% (~5% LTV), bought the bank, and “probably got margin called three times and almost went bankrupt multiple times.” Today employees and founder own 100%, with no dilution and no pref stack, and 25% of earnings buys back employee shares in an annual tender.
“VC money’s kind of like heroin” — feels good, nearly impossible to quit (“how many people do you know that raised a hundred million dollar Series A and then they’re like, ‘I’m done’?”). His test: the venture model only works if you can grow 30%+ off a billion-dollar revenue base — which is “99 times harder” than going $0→$100M in three years — and productively ingest hundreds of millions along the way. Most businesses can’t; he’s not anti-VC (“Plaid would not exist without VCs straight up”), just anti-default.
The risk has inverted: early employees now take more risk than founders. A 24-year-old leaving $400-500K at Google for 1% and $90K reroutes five years of their life, while the de-risked founder can sell secondary and exits with “CEO on your resume.” His fix is to make failure more expensive for founders, not safer for employees — and he argues the current safety produced timid companies: “pivot to AI and wrap Anthropic… that’s not bold.”
The dollar is a weapon and Silicon Valley should keep it American: ~75% of global trade is still dollar-denominated — even China’s imports of Russian oil and gas, even Qatar-to-Switzerland gas moving through US financial institutions. Generals “want to use a sanction before they use a missile” (Venezuela’s economy was collapsed by sanctions before anyone went in), and he explicitly opposes the Valley crowd trying to move financial power offshore: “we should still have the nuclear weapons of financial services.”
Contra the disruption narrative, US financial rails are good: he says there’s no “Cobalt” in the Fed and related systems, which can already clear money 24/7 “faster than stablecoins, faster than crypto.” The gap is implementation — a 50-person community bank can’t staff weekends — and the friction fintechs complain about “is actually a feature, not a bug,” built to protect the 5-10% of consumers vulnerable to fraud. Better detection models could remove that tail and make finance nearly instant for everyone else.
AI value accrues to distribution and fat brands, not AI companies — his railroad analogy: the biggest beneficiary of railroads was Standard Oil, “by an order of magnitude.” Big banks, being headcount- and technology-heavy with minimal physical assets and limited private-equity activity because of regulation, will be “either the most disrupted or probably actually the largest beneficiaries.”
The edge is boring specialization: he mined one product idea worth “millions… hundreds of millions” from a 2,000-page book on 19th-century Chinese banking — the kind of work “you cannot Gemini deep research your way through.” Attack industries where “the dumbest people make the most money,” and treat YC’s quest for startups as “a list of startups you should not start” — by the time it’s consensus, capital and talent have swarmed it.

Deep dive

1. Column: a software company with a bank inside — 90% software revenue

The model in one line: “we have this interesting regulatory moat with a bank that most other people don’t have, and we’re going to build this incredible software behind it that nobody else can build because they’re not a bank.” Column is the backend for payments, deposits, and credit at Built, Wise, Ramp, Brex, and Mercury, and extends to “anybody wanting to do things with the global dollar” — international fintechs and emerging-market banks that need dollar rails.
The Built example makes it concrete: flip the card over and it says issued by Column; the account and routing numbers behind your rent payment are Column’s. The customer builds the app and the marketing; Column handles “everything behind the scenes that has to deal with the Federal Reserve or TCH or the card networks or Swift.”
The economics are deliberately un-bank-like: “we are technically a bank, but unlike banks, we make 90 plus percent of our money off of software… it’s a per API call. It’s a pure play tech business” — with most bank-side economics passed down to customers.
The strategic backdrop: vertical software gets rolled by AI, so software must go deeper into the business — “people actually need to control the underlying finances,” as Brex and Ramp proved. His related call: fintech is “probably going to be like the last area that is somewhat disrupted by AI,” collapsing into domestic fintech and enterprise software.

2. San Francisco and Beijing’s consensus culture

Hockey endorses Dan Wang’s “harsh criticism” that SF and Beijing are the two most consensus places he’s been to — and calls that both “a huge crutch for us” and “probably our most valuable asset.” Consensus is a great operating environment for ideas the Valley believes and the world doesn’t yet (AI, stablecoins), but “we have completely lost touch with how the rest of the world operates, or even the everyday American.”
His sharper framing: Silicon Valley today is “an elite dominated society… probably more akin to Wall Street of the 1990s than it is to what we want it to be, which is a research lab in Cambridge in the 1950s.” Elites build software for elites, drink their own Kool-Aid, and the Valley’s ability to build things that resonate outside SF and New York “is probably at the low point in the entire time I’ve been here.”
Hence Kinshasa. He claims 90% of his ideas come in the shower, walking around, or in “a random emerging markets country” — because in San Francisco “you can’t walk around and not get completely hit with AI FOMO 24/7.” Meanwhile in DRC, mobile penetration is still under 25% and banking penetration under 5%: “there’s stuff we need to do before we think about embedding LLM in everybody’s brain.”

3. Financial innovation thrives under constraint

The pattern: “financial services tend to be most innovative and most progressive in the worst countries” — Argentina, Iran (“they have built a lot of bespoke stuff just for themselves because they do not have access to global financial markets”), and Africa’s M-Pesa, doing mobile payments “decades ago… well before Venmo.” Constraint breeds a creativity that abundance in London, Vienna, or SF can’t.
Talent is misjudged too: if brains are distributed equally, Congo has proportionally as many smart people as France — but “there’s no Anthropic to go to,” so top talent flows to job safety and money: “the brewers and the banks.” His claim: an emerging-markets bank executive team is “hands down way better than what you see in the western world.”
The cross-sell case study is Kaspi in Kazakhstan: bought a bank, then “did everything” — largest e-commerce company, largest bank, you pay taxes and renew your driver’s license on it. Congo’s Rawbank has a mobile app “way better than anything we have here in the US… Imagine JP Morgan doing that.” Because these economies are dollarized, Column can innovate with them like a fintech peer — and the rest of the world is a big part of revenue, though the US market remains “so good.”

4. The anti-VC playbook: “VC money’s kind of like heroin”

Hockey rejects the Valley’s binary — venture-backed-and-ambitious versus bootstrapper running “a cute lifestyle business” with subscale hires: “you can be highly ambitious, hire the world’s best talent, and build a massive company without being addicted to venture money.” Column grows by its earnings, with 100% of ownership held by employees and himself.
The signature line: “VC money’s kind of like heroin. It feels good… but you got to keep shooting up.” The hamster wheel forces strategy drift — “stablecoins are cool this year, let’s do a stablecoin strategy… I need an AI strategy” — a rational but “pretty windy way” to your goal when you must impress the next round.
What independence bought: he acquired a regulated bank during the first Biden administration — “a relatively non-consensus bet” that required two-three years of not focusing on revenue, “not a fundable thing.” He can make 10-year-payback investments comfortably: “if we grow 80% versus 110%, it doesn’t really matter that much.”
It also buys unscalable employee economics: $2,000/month toward rent or mortgage if you live within two miles of the office, and every year 25% of earnings runs a tender buying back employee shares. His mental model: “imagine our profits are our funding round… each January 1st we raise a massive round” — except with zero dilution and no pref stack.

5. The margin-call story

The myth-puncturing sequence: Plaid “attempted to sell to Visa for like $5 billion… we got blocked by the DOJ. And I did not sell my company. Thus, I did not have any money.” People assumed he was a liquid billionaire; his liquidity-to-paper-wealth gap was “pretty extreme.”
So he funded Column with debt: “I pledged over a billion dollars of stock to get $70 million” — the best terms he found were SOFR plus 10% at roughly 5% LTV — and bought the bank for $70 million. “In the process, I probably got margin called three times and almost went bankrupt multiple times.” He salutes his lenders while noting margin lending against private stock is not a great business: the moment you’d seize the collateral is exactly when you don’t want it.
The psychology he took from those years: “when there’s literally only one door in front of you, you don’t have a choice. You have to go in. And in that fear… creates creativity, it creates inspiration.” He kept checking himself against the old line that markets can stay irrational longer than you can stay solvent: “which side of the equation am I on today?”
The residue is total concentration: “I own two things. I own Column and Plaid… I don’t even own a majority of my house.” At 36, with a 6-month-old son, he concedes he “should probably not be this concentrated” — but driving toward something, even “solvency,” is what keeps him motivated daily.

6. We de-risked the founder and left the employee holding the bag

The inversion, spelled out: a 24-year-old leaves $400-500K total comp at Google or Meta for 1% of a startup and $90K — no house, no vacations, roommates, a four-to-five-year trade-off. The founder, meanwhile, can likely sell secondary next round and, if it dies, lands “at a great company with CEO on your resume” — while “first employee at a failing company, that’s actually not a great resume line item.”
His prescription is not to de-risk employees but to make failure much more expensive for founders. His standing irritation: second-time founders who’ve made $100M, put $1M at risk, and raise $500M — “if you believe in this so much… why aren’t you going all in? If I’m an employee, I look at them and I’m like, you’re asking me to go all in, but you can’t.”
The downstream cost of safety: the YC-playbook path — right high school, right college, $3M seed, a soft landing available — “has created a lot of value, but I’m not quite sure created a lot of great founders… there is no risk in that proposition.” Result: “super safe companies. ‘We’re going to pivot to AI and wrap Anthropic.’ That’s not bold.”
His answer to the kindest-thing question ties it together: his parents, whose own life “was not the easiest,” insulated him yet taught him “it’s okay to get punched in the face” (he’s lost real teeth). His worry about his peers’ perfectly optimized children: “we may be creating children that can do linear algebra at seven… but is that what we’re going to need in 20 years? Or do you want kids that are pretty good at taking risks?”

7. Be the best in the world at the most boring thing possible

The specialist creed: “compounding on yourself is probably the best investment to make… I’m probably the best in the world at a couple small boring stuff — really confusing boring-sounding companies that are really hard to explain on podcasts.” His builder-versus-investor stance: “if I’m investing or doing something else, there’s a me on the other side of that trade.”
The study habit Patrick highlights in the intro: Hockey read about Japanese banks in the 1800s and “a very boring 2,000-page book on the history of banking in China in the 19th century.” The yield: maybe one small thing per 2,000 pages — but “that one little thing can create millions of dollars of value… hundreds of millions” when you own the business it applies to.
His best founder-selection heuristic: “can they find the most boring thing humanly possible interesting — over a multi-decade period?” Generalist topics (AI-disrupts-vertical-software, geopolitics) have a thousand people with compelling takes and no capturable value. The valuable niches “require you to read hundreds of thousands of pages that you cannot Gemini deep research your way through… you have to suffer in silence.” His partner asks what’s wrong with him for enjoying that book; his reply: “if I don’t, you and I are going to be super poor.”

8. Earnings are the product: the venture-math test every founder should run

In financial infrastructure, “our customers pay us for safety and our companies pay us for longevity” — 10-15-year partnerships with brutal switching costs. Profitability is a customer feature: “the risk falls on me. I can’t pass my risk to some other venture fund or their LPs.” The competitor risk he sells against: your critical vendor gets acquired, or its founder rationally takes an Anthropic offer.
Earnings get split three ways — employees, growth, and a war chest. He asks how to remain good if something goes wrong for “10 years.” Markets zigzag; “there’s so many incredible companies that just couldn’t survive through a couple bad periods.”
His core discipline for founders choosing capital: going 0→$100M in three years is celebrated on LinkedIn, but “what’s much harder? Going above 30% a year off of a billion-dollar revenue base. That is 99 times harder” — and the venture model only works if you can do that while productively ingesting hundreds of millions of dollars. Most markets don’t work like an ATM where “you put in a dollar, you get a dollar thirty back.”
He’s careful not to be cast as anti-VC: Plaid was rejected by roughly 80-90 investors before its first round, and “Plaid would not exist without VCs, straight up.” Venture is “a really good asset class in Silicon Valley” — it’s just not the right structure for every company, and the industry is opaque about which funds truly have long timelines.

9. The dollar is a national-security weapon — keep it that way

The stat that reframes trade: officially most economies trade under 10% of GDP with their neighbors, but counting unofficial flows “it’s actually closer to 90%.” He separately argues that the dollar’s reach is global: gas from Qatar to Switzerland crosses through US financial institutions, and even China’s imports of Russian oil and gas are “still denominated for the vast majority in the US dollar… They almost hate each other as much as they hate us.” Roughly 75% of global trade remains dollar-based.
The hard-power framing, sourced first-hand: “you can ask any general — and I have — they want to use a sanction before they use a missile.” Venezuela is his proof case: sanctions had “fundamentally destroyed the economy before” any intervention, so he says Venezuelans were probably less likely to resist. France, lacking that lever, has “two options: don’t drink our wine, and here’s some missiles.”
His heresy against crypto-adjacent Silicon Valley: he “fundamentally disagrees” with those who want to move financial power outside the US. “We should still have the nuclear weapons of financial services.” And US finance is a decent baseline for that trust: JP Morgan isn’t that dominant, whereas “95% of Canadians have four banks” — the system should keep fragmenting across US corporations, not concentrate in officials and elites as it does in most countries.
He also demolishes the “US rails are broken Cobalt” narrative: “I rack my brain hard with the Federal Reserve… there’s no Cobalt in all of these places. The Fed has a pretty good tech team.” The US can already clear money 24/7 “faster than stablecoins, faster than crypto. We’ve had that for decades.” The gap is business-model implementation — a 50-person rural community bank simply can’t staff weekends — not the underlying infrastructure.

10. AI value goes to fat brands and banks — so build where the dumb money is

His railroad heresy: value from AI won’t accrue to “AI companies” but to whoever has massive distribution and massive costs. “The value in railroads accrued to the oil companies. Standard Oil was the biggest beneficiary — by an order of magnitude.” His hot take: “the biggest, fattest, most inefficient brands are going to be the best beneficiary of AI.”
Banks fit the template precisely: their business model is “too good,” which made them laggards, but they’re headcount- and technology-heavy with de minimis physical assets (unlike railroads, where cutting conductors from 1,000 to 500 “doesn’t change the equation at all”) and private-equity activity is limited by regulation. So the best-run banks with the biggest distribution and cost structures become “either the most disrupted or probably actually the largest beneficiaries.”
On UX: slow money movement “is actually a feature, not a bug” — make it free and easy for grandma to send money to Nigeria and you’ve optimized for romance scams. All the friction protects “the 5 to 10% of consumers that can get hurt”; if better AI detection models match humans and run instantly, “we can actually take that tail away” and make financial services “almost entirely instant and entirely friction free” for everyone else.
Founder timing advice: with “90 plus percent” of the last YC batch AI-related, “it’s probably a pretty good time to be a founder in a non-AI related place.” His screen: “look at every single industry and say, who is the dumbest people and what makes them the most money? Attack that area.” And the anti-signal: YC’s quest for startups “should be a list of startups you should not start” — consensus attracts capital and smart people like moths to light.

Why Scale Will Not Solve AGI | Vishal Misra - The a16z Show Two Legendary Founders: Travis Kalanick & Michael Dell Live from Austin, Texas