Summary

Concentration is core to the strategy, validated by Druckenmiller. Having just led Stripe at $50B, led OpenAI’s round, and agreed to lead Ramp below its 2021 price while “getting absolutely eviscerated by everyone in the ecosystem,” Kushner told Stan Druckenmiller he felt like he was “running into a burning building, taking my most favorite possessions, and everyone thinks I’m crazy.” The reply he’ll never forget: “That’s always the right decision. You just better fucking pick right.”
The Stripe trade shows the epistemic edge: ~$1.8B — Thrive’s largest check at that point — at $50B, a 50% discount to the prior year’s round, while every other investor obsessed over post-COVID margins and “the now.” Kushner’s logic: “It’s much easier to predict what is going to happen in the long term than in the short term… people buy more stuff on the internet every year.” O’Shaughnessy notes the position is likely back north of $100B in secondaries.
The OpenAI valuation math that answered the $29B skeptics: today’s $3-4T public companies “will likely, by the end of this decade, be seven to ten trillion dollar companies,” so private companies worth half a trillion or a trillion are “not inconceivable” — and since public-market value accrued to fewer than ten companies, Kushner believed the same concentration would happen in private markets as well. The commoditization/open-source critique was “the industry trying to fit a narrative” to how it was meant to operate.
Thrive’s current map is three buckets: AI-native businesses (generalized labs with OpenAI “the leading consumer company” and an enterprise “oligopoly at the hyperscaler level”; domain-specific models in robotics, drug development, life science; a deliberately thin application layer), infrastructure that benefits rather than gets disrupted (Databricks, Stripe powering agentic commerce, Wiz as identity becomes more confusing in an agentic world), and Holdings.
Holdings inverts the disruption thesis: reinforcement learning needs proprietary data plus in-house experts to fine-tune, so “disruption will happen from inside out” — buy businesses via a permanent capital vehicle, run them as technology companies, hold forever, and build “a differentiated cost of capital at scale.” It’s also a hedge: every finance innovation from Kravis’s private equity firm to Milken’s junk bonds to Apollo’s Athene got copied, and Kushner wants a moat before Thrive is arbed away.
His froth read cuts both ways: AI is “the most important paradigm shift of our lifetimes” — the internet democratized information, AI has the potential to democratize access to intelligence — but chasing it will produce “ultimately capital loss as well.” The discipline metaphor: investors are race-car drivers who must know when to drive the speed limit, when to change tires, and — as Thrive did the last couple of years — when to “put your foot on the gas pedal and gun it.”
The firm itself is the compounder: a firm managing ~$50B kept deliberately tiny (“the only way to have a team is to have a small team”), staffed by people who are “either first generation at Thrive or from the most hated state in the country, New Jersey,” and governed by a post-Instagram lesson attributed to someone likely named John Winkelried, posted on every computer: “never believe your own bullshit.”

Deep dive

1. Concentration is core — “you just better fucking pick right”

The episode’s defining anecdote: having just led Stripe at $50B, led OpenAI, and been asked by Eric and Karim at Ramp to lead a round below the 2021 price, Kushner drove uptown to Stan Druckenmiller: “We’ve always been concentrated in the businesses that we believe have the generational characteristics to compound… We sold a lot in ‘21, and we’ve leaned in very meaningfully in this moment, and I feel like I’m running into a burning building, taking my most favorite possessions, and everyone thinks I’m crazy.” Druckenmiller: “That’s always the right decision. You just better fucking pick right.”
The check sizes were never a top-down strategy, Kushner insists when Patrick probes for the “grand meeting”: “Stripe needed to raise six billion dollars or else they had to go public, and it was our job to find a way to get them that capital.” Same for OpenAI and Databricks — input orientation, not output orientation.
Concentration carries an obligation he states as a duty to both sides: discipline owed to LPs and total commitment owed to founders — “there’s only truth tellers in foxholes.”

2. GitHub and Instagram — the formative trades

GitHub, 2014: a $20M initial check, big for Thrive then. A month in, CEO Chris Wanstrath moved on from much of his leadership team, and Nabil — fresh from running finance for Heinz at 3G (“working on ketchup… a lot of similarities to software”) — became effective CFO. The outside view said troubled; Thrive’s time-intensive read said noise. When early founders wanted liquidity, “Thrive was the only person who wanted to buy the shares… Thrive owned about ten percent, and no one knew.”
Instagram: a $40M fund put $12M into a $500M round that sold two days later for $1B. The durable lesson wasn’t the double — it was someone likely named John Winkelried (now TPG’s CEO) calling the 26-year-old fresh off the Wall Street Journal cover: “I’m gonna give you the greatest lesson that you’ll ever get from me, which is never believe your own bullshit.” Kushner wrote it on Post-it notes and stuck it on all ~10 employees’ computers for Monday morning.

3. Stripe at $50B: the long term is easier to predict than the short term

The terms: $50B with John and Patrick — a 50% discount to the round done a year prior — and roughly $1.8B invested, Thrive’s largest check at that point. The reasoning: “It’s much easier to predict what is going to happen in the long term than it is in the short term. If there’s one thing that I could predict between now and the day that I die is that people buy more stuff on the internet every year.”
What the process revealed about everyone else: because of the check size and the moment, “everyone asked to speak to us” — and every conversation fixated on that quarter’s multiple and post-COVID margin structure. In a founder-oriented style, conviction rests on “people’s ability to figure things out”; “the idea that anyone could ever doubt Patrick and John is blasphemy.” O’Shaughnessy notes the position is probably back north of $100B in secondary markets.

4. OpenAI: “I just couldn’t unsee it”

Kushner first approached Altman in March 2022 — before ChatGPT — reasoning that “if compute and research was all that mattered, they were most set up to ultimately succeed.” Months later, Sam gave him the ChatGPT preview, intended only to “show the world what these models were ultimately capable of”: “I just couldn’t unsee it.” At one or two in the morning at his kitchen table, he told Carly it was going to change the world.
The answer to the $29B skeptics was a repricing of the whole opportunity set: multi-trillion-dollar companies “will likely, by the end of this decade, be seven to ten trillion dollar companies,” so private companies worth half a trillion or a trillion are “not inconceivable” — especially since “a disproportionate amount of the value that has accrued in the public markets has accrued to less than ten companies,” and Thrive believed privates would rhyme. The LLM-commoditization, open-source, app-layer-wins critique was “the industry trying to fit in a narrative… that would play to the way in which the industry was meant to operate.”
The service is literal: Altman’s vignette, relayed by Patrick — a 2am call, and Kushner got on the phone with an IC engineer being recruited against a rival lab from 2:00 to 3:30am; the engineer signed. Proximity also taught a darker lesson: “Sam, Greg, OpenAI have the ring, and everyone is willing to do whatever they can to take it.” Asked flatly whether power corrupts: “Yes.” His own antidote: “True power doesn’t shout. It acts quietly and decisively.”

5. The current map: three buckets

AI-native businesses: first, a generalized lab — OpenAI is “obviously the leading consumer company,” with “somewhat of an oligopoly” at the hyperscaler level for enterprise. Second, domain-specific models — robotics and embodied intelligence, drug development, life science; Isomorphic, supported with Demis after Alpha Fold, is the exemplar: “creating a company whose end goal is to cure disease is probably the most important thing that we could ever be a part of.” Third, a deliberately thin application layer — “we’ve done very few investments,” because “Apple has launched the iPhone, and they don’t know what applications they want to own versus what they’re willing to let other people own.” What earns conviction there: reinforcement learning plus memory applied to user preferences.
Infrastructure that benefits rather than gets disrupted: businesses like Databricks — but also “things like Stripe that need to kind of power agentic commerce,” and identity as agents proliferate: “companies like Wiz become increasingly more important.”

6. Holdings: disruption goes inside-out

Origin: right after the OpenAI investment, the company had $50M of API revenue, and Thrive pitched New York private-equity firms on using the API for efficiency inside their portfolios — “we did not get meaningful traction.” Karim’s response: “Let’s just start doing this ourselves.”
The thesis inversion worth keeping: reinforcement learning needs “the data that is proprietary to the company but also the experts that exist at the company to fine-tune the model,” so “on a go-forward basis, disruption will happen from inside out” — rather than outside-in disruption. Hence a permanent capital vehicle to buy businesses, run them as product-and-technology companies, and hold “forever” — targeting “a differentiated cost of capital at scale.”
The motive is candidly defensive: “One of my greatest insecurities with Thrive is that we fundamentally look very different than the businesses that we try to invest in.” Every financial innovation got copied — Kravis and private equity, Milken and junk bonds, Apollo creating Athene — and Kushner wants the provocative moves to compound into “an element of true defensibility and a true moat.”

7. Froth check: paradigm of a lifetime, capital loss ahead

The base case: AI is “the most important paradigm shift of our lifetimes.” The internet democratized information — “anyone with a mobile phone living anywhere in the world has more access to information than the president of the United States did at the turn of the century” — and AI has the potential to democratize access to intelligence; enterprise adoption will lead to efficiency until, a decade out, “the only way to compete… will be, are you a better product or are you a cheaper product?”
The hedge, stated without smoothing: the transformation “will lead to a lot of people trying to chase the paradigm shift and ultimately capital loss as well,” and “it’s very dangerous in moments like this to both be exuberant or of a more negative perspective.” His operating metaphor: the best investors are race-car drivers — know when to drive the speed limit, when to pull over and change tires, and, as Thrive did over the last couple of years, when to “look both ways, put your foot on the gas pedal and gun it.”

8. Firm-building: small teams and chips on shoulders

A prominent endowment once passed on Thrive because Kushner was “too much of an entrepreneur” — “I was told that investment managers building investment firms were not meant to be ambitious.” He rejects the implied critique that ambition means asset bloat: the focus is inputs — culture, process, product for founders — and “there will be moments in time where the opportunity does not exist to write very large checks, and we won’t.”
On why everyone there is so driven: “People are either first generation at Thrive or from the most hated state in the country, New Jersey.” But the insecurity engine has a governor — “Insecurity is the greatest driver of innovation… but it can’t be the end goal,” because proving something to the world is “a never-winning game” and “the world doesn’t actually care.”
The team design: “the only way to have a team is to have a small team,” with “10X people” in legal, finance, compliance, and engineering so everyone feels “pride and authorship” over decisions — “artists like to live in artist colonies.” Against the new risk of praise (a recruiter survey put Thrive #1, with more than 2x the votes of the next-highest firm, which Patrick thought might be Sequoia), the Marcus Aurelius filter: be “the same man in both good times and in bad.” And the ambition check: “I think Thrive is a very small percentage of its potential.”

9. Loyalty, art, and the spine underneath

Competition as a swim race: “if you look to your left or if you look to your right, you’re gonna lose… just swim as hard as we can in one direction and don’t save anything for the swim back.” The ethos: never sell against another firm, never invest in competitive businesses, trust by default — “but if someone breaks that trust, they’re out.” The line he gives the team: “Kindness without toughness falters, and toughness without kindness corrodes.”
The A24 investment carries Thrive’s self-image: “Our founders are our heroes. We’re not Da Vinci, we’re Medici.” Frank Ocean — no album in thirteen years — models intentionality (“if there’s nothing for us to create, then we won’t create”); The Beatles model the team: a small group’s seven-year run, with songs like Hey Jude “only performed live once.”
The perspective source: his grandmother Rae — his new daughter’s namesake — escaped a Belarus ghetto through a 600-foot tunnel dug with a spoon after watching her mother and sister murdered, lived with the Belsky partisans, spent four years in a refugee camp, and arrived in America with nothing. “Nothing I will ever accomplish in life will be greater than what she pulled off.” And the closing story: in Guatemala City, a gunman on a motorbike took his friend’s watch and phone, looked at Josh’s Swatch skin, handed it back, and drove off — “I’ll never wear another watch.”

Mathematical Superintelligence: Harmonic's Vlad & Tudor on IMO Gold & Theories of Everything February 2026 Random Ramblings