How LP Deployment to Venture Will Change in 2025 & Are Endowment Funds in Trouble?

Summary

• OpenAI’s rumored Windsurf acquisition is a “1% deal” in Jason’s M&A sizing framework: 10% of market cap is bet-the-farm (Adobe/Figma; Instagram and WhatsApp each ~10% of Facebook’s value), while 1% is “an SVP saying, ‘I’m betting my BU’” — a bet you “won’t even notice.” The logic: coding is the one front where OpenAI trails Anthropic, even as “ChatGPT has pulled away. You cannot catch it… you will never catch that revenue.” Rory isn’t convinced the deal closes, but even if it dies — “the mighty corporate intent has been stated,” and the line of coding apps outside OpenAI’s office “is going to go around the freaking block.”
• The narrative evolved from models are everything → apps are just AI wrappers → models are commodities → “the models have such market cap, they can buy the apps.” Jason’s conclusion: “no one knows nothing,” which is exactly why Altman’s “bias to action” is right — “Will it be PowerPoint or will it be Excite@Home?… Not making a move is akin to losing.”
• PMF can now lead to a $3B decision in 90 days: 90 days ago, Windsurf “barely existed” as a Chrome plugin called Codium; Cursor and Bolt both nearly died first. Once you lock in, you can jump from “this company is not going to make it” to turning down $3B in 90 days — because 100% of early adopters are in market and the products cost $20/month (“they make regular B2B look like a freaking ripoff”). Jason’s caveat via a Benioff conversation: it’s still early — 90% of enterprise is “just playing with ServiceNow.”
• Seed is structurally squeezed: Harry told a large LP he “wouldn’t touch” any SF seed manager because multi-stage firms’ seed product is “so good, so efficient” with such cheap capital that “they’re just crushing everyone at seed.” Even Green Oaks’ triumph — leading Windsurf’s seed, ~10% ownership with dilution — illustrates the pressure. In the Insight/Wiz example, a reported $500–600M return was less than a third of a $1.5–3B fund. Jason: with seed ownership compressed from 15% to 10% and funds doubled in size, “can a billion dollar outcome even return the fund anymore?… I don’t even think it can.”
• A revenue multiple without a growth rate is “an incomplete equation not worthy of discussion” (Jason, quoting partner Andy). 100x paying 3–4x growth with “high growth persistence” de-risks; 100x on sub-2x decelerating growth and “you’re so screwed your head will hurt.” The 2021 fiasco was paying for growth and not getting it — and Harry’s live example ($7M revenue priced at $700M) only works with one more great year.
• The endowment model is under pressure at Yale, its intellectual home — a reported ~$6B secondary sale that Rory likens to “Vanguard saying active management is the way to go.” Pattern: stress plus exogenous shock — “in ‘73 it’s the oil crisis, in ‘25 it’s the Trump crisis.” Mayur’s read: venture is “a rounding error… just juice”; the real stressor is PE, 5–10x the allocation with no liquidity — and Harry was shocked to find five-plus big-name endowments over 30% in privates.
• Hot hands will keep spinning out: “If you have a hot hand in venture and you’re not running the place, I would leave the next day” (Jason), and a new firm lets you “sever that thing like a stage of a rocket” — your own track record included. But Rory warns 2025–26 is a brutal vintage to raise: “You can want to buy a Ferrari… if you haven’t got the money, you can’t buy a Ferrari.”
• Jason’s conviction rule — take it: “Any deal where you have 100% conviction you’ll 5x, you should do it,” irrespective of ownership or valuation — crystallized by Harry’s confession of passing on 1% of 11 Labs at $25M (now $3B, a fund-returner for his $25M fund). Rory’s caveat: only solo shops can play the exception card — “the guy in Omaha who doesn’t listen to anyone is the richest man on the planet” — while peer-team firms that break the rule once “have broken it for everyone.”

Deep dive

1. Windsurf at $3B: the rumored 1% bet

• Jason’s M&A sizing framework from his big-tech VP days: 10% of market cap is bet-the-farm — Adobe/Figma, Instagram and WhatsApp each at “the magic number,” 10% of Facebook’s value — while “1% is like an SVP deal. This is an SVP saying, ‘I’m betting my BU.’” Where is OpenAI weak versus Anthropic? “It’s in coding. It’s the one area it’s weak” — while “ChatGPT has pulled away… you will never catch that revenue.” So 1% to catch up: “you won’t even notice 1%.”
• Jason’s zoom-out: at a $300B market cap playing to be one of the three or four $2T companies, owning a major use case makes sense. The proof no one can see ahead: conventional wisdom evolved from “models are everything, apps are just AI wrappers” to “models are commodities” to “the models have such market cap, they can buy the apps” — across roughly a year. “No one knows nothing.” What he admires in Altman from a distance is “a bias to action”: direct chat, tick; coding, hmm; customer success, TBD later.
• History offers both endings: Microsoft bought the elements that became Office for “tens of millions” in the early 80s — “who cares” — while Excite@Home was “a deal as dumb as rocks.” “Will it be PowerPoint or will it be Excite@Home? That’s why they play the game. But not making a move is akin to losing.”
• Rory isn’t convinced the deal happens, and Jason disputes that Cursor was unbuyable: Altman “can spend 10% — $30 billion,” which is 3x the last round — “I guarantee you the VCs at the 10 billion round will take a quick 3x.” Rory: even if Windsurf falls through, “the mighty corporate intent has been stated… the line of other coding apps outside the OpenAI office is going to go around the freaking block.”

2. Platform peace now, a ten-year grind later

• Jason: an acquirer needn’t run “the brutal Salesforce–Oracle strategy.” Ship “Windsurf powered by OpenAI,” put a thousand engineers on it, but “if you want to use Cursor, if you want to use Lovable… we still love you” — run different teams and let the market decide.
• Rory’s pushback, worth keeping: fine short term, but “in the end there is a grinding you down element” — after 20 years of the PC wars, only Adobe, Quicken and some security apps sold productivity software independent of Microsoft at scale. “It’ll be a long 10-year grind if you’re the independent.”

3. Seed funds are being crushed — even the winners’ math barely works

• Harry’s cap-table takeaway: Green Oaks led Windsurf’s seed and doubled down at the A. His answer when a massive LP asked for a SF seed manager to back: “I wouldn’t touch it. The multi-stage fund product at seed is so good, so efficient, and their cost of capital is so low that they’re just crushing everyone at seed.”
• The brutal arithmetic: Green Oaks’ fund is $1.5–3B and owns ~10% with dilution. In the separate Insight/Wiz example, a reported $500–600M return was less than a third of the fund on the smaller number. To 5x you need “just 15” Windsurfs per fund. Rory refuses the systemic read — it’s “extraordinarily good picking from a very connected investor,” and anyway “I’ve been doing this for 30 years. Everyone cashes the check.”
• Jason’s structural worry: seed rounds compressed from 15% ownership (12.5 + 2.5 for angels) to 10% (7.5 + 2.5) while seed funds doubled in size — “a compounding set of risk pressures.” The old line was a billion-dollar outcome returns the fund: “I don’t think it’s true of a lot of seed managers anymore… I don’t even think it can return the fund.”
• Rory’s leveler: “Everyone is looking at everyone else’s spot and going, my spot is hard, my God, theirs looks easy… the dirty little secret is it’s hard everywhere” — no one is staying in their swim lanes. Jason adds, per Barton Biggs, “there’s no business so good that excess capital can’t ruin it.”

4. 100x is back — and meaningless without the growth rate

• Harry walked out of an IC where a company doing $7M revenue was priced at $700M — “we’re back, huh?” Jason (after calling Harry a hypocrite who pays the highest multiples): his partner Andy “will refuse to have a conversation about revenue multiple unless you state also the growth rate. It’s an incomplete equation not worthy of discussion.” Forward multiple here: ~33x, meaning it’s 3x-ing.
• The mechanics: venture starts at an infinite multiple and everything eventually trades at 5–6x revenue growing 20% — “you hope to God the growth rate stays higher long enough to de-risk the multiple before you intersect the public markets.” 100x growing 3–4x with “high growth persistence” gets you out of the risk zone; 100x at sub-2x and declining, “you’re so screwed your head will hurt.” 2021’s fiasco: “people paid up for growth rates at 100x and then didn’t get the growth.”
• Why SaaS’s decade was golden: predictable sales-and-marketing input to revenue output, sticky revenue — “apply capital and grow into the multiple.” Jason’s younger partners’ unspoken verdict: “You idiots, you made money when it was easy… It’s hard today, brother.” His concession: “they’re right.”
• Bryce’s (OATV) puzzle — if AI makes companies cheaper, why are seed rounds bigger than ever? Mayur: VCs “love companies that don’t need their money” (Accel’s genius was owning 20–30% of bootstrapped Atlassian), and founders are “utterly insensitive” to astronomical prices. His rule: “stop at $100 million valuation if you’re not sure you’re going to IPO” — the kids “see no risk in raising at 3, 10 billion.” Mayur: “We want to invest in capital-efficient companies… in a capital-inefficient way. It’s a paradox, but it’s true.”

5. PMF can now lead to $3B fast

• “90 days ago Windsurf barely existed — it was a Chrome plugin called Codium.” Cursor almost died; Bolt almost died. Rory’s frame: the “walk in the woods period is indeterminate” — 6 months or 5 years, seed finances the journey — but where SaaS locked into triple-triple-double-double, now “you go from ’this company is not going to make it’ to ‘oh my god, I think I’m going to turn down three billion dollars’” in 90 days.
• Why the odds at the craps table changed: “AI hype is bigger than all the other hypes put together.” Two years into the internet, Krugman was still asking whether it mattered; two years into AI, “every company is saying, shit, I got to do something here” — no one tells their board “AI really matters, but I’m a bit nervous.”
• Jason, off a catch-up with Marc Benioff (“I got to tell you, it’s so early for others”): 100% of early adopters are in market — while “90% of the enterprise is not even there. They’re just playing with ServiceNow.” And it’s cheap: Windsurf is $20/month against ~$20K just to get Atlassian to engage — “these products make regular B2B look like a freaking ripoff… I don’t think it’s as crazy as it sounds.”

6. Hot hands leave: the spinout logic and its expiry date

• On Bucky leaving KP, Jason speaks from experience: “If you have a hot hand in venture and you’re not running the place, I would leave the next day.” Tomasz Tunguz (ex-Redpoint, now managing toward a billion solo) told him he should have done it earlier — versus “$400K to a million a year plus waiting 22 years for some carry… why would anyone stay?”
• Rory’s correction, knowing Jason: “Delete the first part of the sentence. If I’m not running the place, I’ll leave. Some people just want to run the place.” The leadership obligation is centripetal: “if you’re not putting them inside the tent as quickly as humanly possible, you’re an idiot” — well-run firms make damn sure hot hands get promoted and cut in, “and if we’re not doing that, shame on you.”
• Why now, per Rory: five-six years of markdowns mean juniors are “just digging out of someone else’s hole”; LPs barbell — “$200 million into mega fund and I give 20 million bucks to Tomasz. I feel good.” And the unspoken bonus: a new firm severs your track record “like a stage of a rocket” — Fred Wilson was “wildly unsuccessful” at Flatiron in the dot-com crash, “went on to do a new thing and killed it from day one.”
• The expiry date: LP appetite for new funds is lower than it’s been in a long time — endowment withdrawal is real, “fines are coming… and tax-exempt status is at risk.” Rory predicts “a little bit of clinging to the lifeboats” over the next one to two years: “You can want to buy a Ferrari… if you haven’t got the money to buy a Ferrari, you can’t buy a Ferrari.”

7. Yale sale tests the endowment model

• Yale — “the intellectual godfather of the endowment model,” whose Swensen book Rory read 20 years ago and considered definitive — is reportedly selling ~$6B of assets in a secondary. Rory’s analogy: “this would be like if Vanguard said, we’ve been thinking active management is the way to go.” His ascending severity ladder: (1) just need cash, having “misjudged the amount of illiquidity you could afford”; (2) the 17% return is really 13–15% — “I’m taking the risk and I’m not getting paid”; (3) catastrophic: “I actually need the damn money.”
• The exogenous trigger: “In ‘73 it’s the oil crisis, in ‘25 it’s the Trump crisis.” As a hypothetical Ivy CFO facing 30–40% of revenue disappearing, anyone selling illiquids gets “get the freak out of my office… I’m thinking bonds here, dude. Index funds. Accessible cash at a moment’s notice.”
• Mayur’s two-part read, from LP conversations pre-dating Trump: everyone’s “cash planning was just wrong” — fine with paper returns, but nobody modeled the liquidity drought lasting this long. And venture isn’t the issue: “venture is a rounding error in most endowments… it’s just juice.” PE, at 5–10x the allocation with deals not going public in 2–3 years, “is an order of magnitude bigger issue.”
• Raising his fund nine months ago, Harry was shocked to find five-plus big-name endowments over 30% in privates — he’d assumed 6–10%. Rory: not crazy for multi-century institutions (Bologna, Oxford, Cambridge — “the longest-lived institutions in the world other than the papacy”) dispersing ~3% a year; they “just didn’t plan for the situation where the world could change utterly.” Harry won’t bail them out — “they poked the bear… it’s not my problem” — while Harry, a naturalized citizen whose wife warned him not to get “singled out to the president,” is “more than mildly sympathetic”: foreign students paying $60–70K a year is “one of the best products we have in the country. It’s not clear to me why killing this particular golden goose is a good idea.”

8. AI rollups: “they didn’t pick you”

• Mayur is “modestly skeptical,” ironically opening with his own success: SpeechWorks-turned-Nuance built its 2005–2015 medical transcription business buying “crappy little mom-and-pop transcription companies, injected the AI and made it work.” “But — and this is my but — I think it’s a crappy model.”
• The mechanism: you’re buying customers “that weren’t picked by you… they didn’t pick you because of your AI, because you didn’t have the AI when they picked you.” Maybe 3–4 of 10 convert to pure software; the rest churn or drag, you develop no new-deal muscle, and you pile up services at a low multiple. Jason: this predates AI — overvalued companies have always bought terrestrial assets on the cheap — and “it’s financial engineering is all it is in most cases.”
• Mayur’s pushback with a live specimen: one of his fastest growers went 0 to $30M revenue in two years rolling up real-estate management with AI tooling — customers “pretty much identical,” zero ambiguity, margins from 5% to 40% in six weeks. Rory concedes the key sentence is “all the customers are exactly the same and it’s all tuned to your technology” — but broad-based plays like BPO rollups are “probably a lot harder than you think.”

9. Where nobody’s looking — and the fatal error of missing a competitor

• “All the esoteric areas are full” — rockets, defense, healthcare. Rory’s partner’s contrarian nudge: “everyone’s running away from consumer. Maybe you should spend some time there.” Rory’s two: gnarly true-enterprise problems — “there’s not going to be that hundred kids in SF that want to build the next ServiceNow” — and vertical SaaS is still underinvested: outside legal and sales tools, “you’re going to see five competitors, but you’re not going to see 500.”
• The crowding is real: Harry tweeted that five years ago every deal had 2–3 competitors, now 10–15 — that day’s IC market map for an L&D/GenAI-security tool stunned him. Rory: LMS “was already overcrowded before AI… too many vendors for a midsize TAM — that’s the worst area to invest in.”
• Rory won’t pretend he meets every player before a term sheet, but defines the diligence bar: know the universe and how they’re doing via “jungle telegraph” and references. “The fatal error… is you fast forward 12 months and the number one competitor is someone you hadn’t heard of. At that point you should be committing ritual suicide at the boardroom table because you screwed up.”
• On market structure: consumer skews winner-take-all; enterprise apps skew oligopoly — HubSpot and Salesforce both won CRM (“Salesforce was an investor in HubSpot early on”) because sub-markets have real nuance; infrastructure is more winner-take-all since “you don’t need a separate router for healthcare versus banking.” Jason’s billion-dollar exits were all “brutally competitive” oligopoly markets.

10. The 100%-conviction 5x rule — and who’s allowed to use it

• Career honesty first: pressed by Rory, Jason counts “only three, maybe four” deals returning 10x in distributed cash. Rory: “It’s so damn hard to have four… so many people don’t get any.” Jason is “more proud of my 2xs and 3xs in 2004 and 2005-6 than any 10x that kind of sailed into the 21 bubble and made me a fortune.”
• Rory on crises of confidence: “All the time. My most recent, today.” Out of his first five deals he thought he’d lose money on four — three years of barely sleeping in ‘95. The discipline: never say “I was great then, so it’ll work out” — ask “am I playing the game correctly for where it is today?” His recent mistakes trace to skipping a step: “when you skip a step, it bites you in the ass.”
• Harry’s confession — offered 1% of 11 Labs at $25M against his $25M fund, passed on ownership grounds, and it’s now a $3B company (“a fund returner on one”) — draws Jason’s rule: “Any deal where you have 100% conviction you’ll 5x, you should do it, irrespective of ownership or valuation… You’ll never regret this sure-thing 5x.” Rory’s engineering translation: “you’re only as accurate as your least accurate variable” — informed conviction outweighs the price debate, and when you know “this is a thing” (his own: every app will be rewritten as SaaS), “finding a way to monetize that bet is actually your job.”
• The institutional caveat: exceptions corrode peer-team firms — “if you start breaking the rule once, you’ve broken it for everyone,” which is why high-variance shops are “single leader dominated”: “the guy in Omaha who doesn’t listen to anyone is the richest man on the planet.” In a broadly equal team with seven check-writers, they stick to strategy; the real agony isn’t missing a Hail Mary outside the strike zone but missing — or worse, turning down — the 5–10x inside the sweet spot.

11. Jason and Mayur: growth filters, one more Wiz, and an AI associate

• With Rory gone, Harry raises the Bay’s 82 tech billionaires (likely Henley & Partners) and rising. Jason, who loved pandemic beach living in SoCal: “the density here” — a DM yesterday from a new-generation tech billionaire, a downtown meeting tomorrow that “would never have happened if I wasn’t here.” “SF is so back.”
• Mayur’s heterodox filter: he’s done competitive diligence exactly once (Pipedrive, his first deal) — “not once” since — because he only invests at top-0.1% growth plus a founder he believes in: Talkdesk went 1 to 15 in five quarters, Algolia grew 20% a month for two years. He’ll never give on growth or on an S-tier CTO (“or I’m off”), gives entirely on competition (“super loose… it’s the box I just completely have given up on”) and education (“I don’t care if you went to high school”). He’s known he wanted to invest before the first meeting “100% of the time” — and has never met a passionate outlier-growth founder who turned out dull.
• The business model behind it: Mayur is only in it for one big massive win — “All I want to do is one more Wiz” — a real $8B outcome, “not fake 8 billion on paper.” Owner is the most competitive market he’s ever backed; the counter-lesson is Olo — “an enterprise play in an SMB market… one billion and it’s been 15 years, brutal slog” — because the consumer end of B2B is niche (even Shopify gets only 25% of revenue from big brands).
• His deal flow is now machine-assisted: an AI reviews inbound decks, benchmarks growth and TAM against his portfolio, states his check and ownership parameters, and pings him when metrics clear the trigger — “better than an associate… and the AI does not judge you.” Contrast at growth stage: Meritech’s diligence on a shared deal ran to ~100 customer calls — “the entry ticket for growth firms to get a meeting” — while Mayur talked to two customers, after the term sheet.