Daniel Gross and Nat Friedman: Acquired by Meta | Microsoft Layoff 9000 People | OpenAI's Bombshell

Summary

• The NFDG-to-Meta exit is the market repricing AI talent in real time. Nat and Daniel built a $1.1B fund, 4x’d it in two years with only ~50% deployed — roughly $300M of gains already — and still walked. Rory’s math on what they abandoned: another 4x on the remaining ~$1.5B of investable capital implies ~$800M–$1B of forgone GP economics, “and obviously whatever offer they got had to be better than that.” Jason’s verdict on venture as a career for people of that caliber: “your highest and best use is not being the 50th venture guy.”
• Talent scarcity is the trade underneath everything discussed. Jason’s flat call: “It will be the biggest issue of 2026 in B2B AI — just the inability to recruit talent.” A ~$200M-ARR B2B CEO is giving 0.5–1% of the company per AI engineer; OpenAI’s stock comp ran $4.4B last year, 119% of GAAP revenue, and the panel expects the projected drop to 45% “is not going to happen.” The consequence for LPs: this wave’s SBC dilution will be materially worse for venture returns than prior generations.
• Inflated equity is a weapon — use it or waste it. CoreWeave buying Core Scientific for $9B (after Weights & Biases at $1B) swaps leases and rent for stock, taking it “from 100% leveraged on infinite data center demand to 80%” — the panel frames it against the Exodus mistake. Harry’s rule: a 20x-revenue stock with existential losses should be buying “everything you can that can address the bottom line” — and expects Circle to follow once past its first earnings, likely buying distribution to cut its Coinbase cost.
• Thoma Bravo taking Olo private at $2B — 6x ARR on a profitable ~20% grower — is the honest comp for the unicorn overhang. “Meat and potatoes”: public at $25 in 2021, out around $10. Rory’s warning: it’s a start, “but it’s not going to save 500 to 700 unicorns single-handedly,” and triple-triple-double-double companies without AI exposure “will have a discounted price… I’m sorry for them but it’s just the truth.”
• Carta’s 8-year low in VC deals and the AI frenzy are the same phenomenon: a flight to consensus. Rory: “attention begets more attention… no one wants nearly as good,” so capital crowds into the handful of breakouts and everything else struggles at any price — possibly “there’s no price at which you’ll do it.” Jason passed on a flawless triple-triple-double-double deal purely on opportunity cost: “I only have so many shots on goal. I don’t want to take this shot today.”
• Vanguard putting PE into target-date funds via Blackstone drew a one-word review from Jason: “terrible.” His line — “a target date fund is a hunt for folks that don’t even know how to spell alpha… and you’re going to put private equity in it” — plus Rory’s point that retail is least able to hold illiquid, questionably-marked private assets, makes this a greed tell.
• The AI-adoption purge inside companies is now explicit. Microsoft’s 9,000 layoffs swap relationship sellers for solutions engineers — a panelist sees 30–40% of one-to-two-call sales reps replaced by AI — and Canva’s “AI discovery week” for 5,000 staff is read as performative notice. A panelist on a public-company board was asked what to do with employees who won’t embrace AI: “You fire them.”
• On the macro, the panel says recession risk is priced right (~1-in-6 baseline) — but Rory reverses himself on the best question of the episode: expansions do get riskier with age, because “when times are good people pile up the dumb aggressive shit… manias, panics and crashes and you’re done.” Harry would still put $50–75K on no 2025 recession since “the technical definition of a recession is pretty tough to meet.”

Deep dive

1. NFDG’s exit: a very clean fund wind-down, and the market speaking

• The setup, per Jason: a $1.1B fund, 4x’d in two years, ~50% deployed per the WSJ, about to close another fund — “for 99% of the venture world this is beyond a dream outcome.” The math Rory walks through: $500M deployed at 4x is $2B, $1.5B of paper profit, roughly $300M in gains already for two partners plus a few others.
• The offer’s structure is what makes it clean: Meta buys 49% of the fund, each LP elects how much to sell, and anyone taking the offer locks a 2x floor — “even if the whole thing goes to zero you got a 2x.” What Nat and Daniel give up is ~$1.5B of investable capital; another hypothetical 4x implies ~$800M to $1B of forgone economics, so the Meta package “had to be better than that.”
• Harry’s read on LP sentiment cuts against the tidiness: “I do not think LPs are happy about this outcome” — they’re losing the stewardship and future funds they were excited about. Rory’s rejoinder: “getting jilted with a 2x is a lot less painful… if getting screwed over is getting 2x in cash and a ticket to ride on the other half in two years, we should all be so lucky.”
• Harry’s sharper pushback: an SSI investor whose round Daniel led as an operator “a couple of weeks” ago has a right to be pissed. The panel’s answer is incentives, not propriety — “you shouldn’t make decisions based on someone else quote-unquote doing the right thing. All you can do is evaluate their incentives” — with Gary Tan leaving Initialized (a ~10x fund with Flock Safety and Rippling) for YC as the precedent.

2. Will Meta’s talent machine work? Yes — the harder question is the business

• Jason thinks the accumulation play succeeds because it already worked at X: “they’re running a little bit of the Elon X playbook of just being insane, cracked, creating this mega mecca for talent.” His iron law: “the best only want to work for the best” — and it’s why struggling unicorns are in “an existential death spiral,” unable to attract anyone great once growth falls to 10%.
• A supporting example from the OpenAI books: one of the first emails (likely from Ilya) to Elon was essentially “if you would lend your name to this project… we would hire more engineers quicker cuz you’re cool.” Prominence has always been the recruiting flywheel; Meta is just paying cash for it.
• The reservation is the one worth holding: these hires are “members of the mythic inner circle of people who know the magic spell — they will make the product. The interesting question as a business decision: will being the fourth or fifth broadly capable LLM be a compelling business for Meta? To me that’s a much more difficult, unclear question. But I don’t think they’re agonizing about that now. They just feel the existential need to play.”
• Jason’s one doubt echoes Washington: “I also worry this is going to be like the Trump administration — everyone’s going to quit too… do I really want to go work for Meta for four years?” A once-in-20-years moment, “like 1999 except it’s not going to implode on us in 12 months” — because this time incumbents will monetize your quitting for a hundred million bucks.

3. The talent war’s bill comes due as dilution

• A multi-billion-dollar founder told Harry that morning: “my single biggest challenge today is Cursor” — vibe-coding leaders paying through the nose for everyone. Jason’s response is the episode’s headline call: “It will be the biggest issue of 2026 in B2B AI — just the inability to recruit talent.” And his tell for fake answers: “if a startup says I want to hire a VP of AI, I’d like to sell all my shares on any secondary market that exists.”
• The concrete cost: a CEO Jason knows at ~$200M ARR “basically has to give half a percent or a percent of my company to each AI engineer now… I got no choice” — at exactly the stage engineers normally get 0.001%. Harry’s conclusion: this wave “will be highly damaging to venture returns because the employee stock-based comp dilution is going to be so much more significant than in prior generations.”
• The OpenAI data point: $4.4B of stock comp last year, 119% of GAAP revenue, projected to fall to 45% this year — “it’s not going to happen… it’s going to still be in the triple digits.” The panel’s accounting caveat keeps it honest: SBC is priced off a rising 49A, so the dilution may be far less catastrophic than the GAAP number — at a $300B valuation, $10B of stock is 3%.
• A panelist’s absolution for the spenders, verbatim: “No one ever said to Winston Churchill, ‘Did you bring World War II in on budget?’ They just said, ‘Did you win World War II?’… When it becomes existential, you do what you have to do to win. Not getting there is the fatal error.”

4. CoreWeave’s playbook: turn a meme multiple into a balance sheet

• A panelist liked the $9B Core Scientific deal (following Weights & Biases at $1B), framed against Exodus in ‘98–‘99 — the first hosting company, huge valuation, then “absolutely bankrupt because it had leases and debt.” CoreWeave “uses equity to take out rent expense,” going “from 100% leveraged on the upside of infinite data center demand to 80%. That’s a smart slight de-risking.”
• Note the hedge Harry scoffed at: “Who would take the bet that demand for data centers would go down?” A panelist: “I would put that bet at least on the table as being plausible at some point in the next 3 or 4 years” — people find they’re ahead of their investment schedules and slow down.
• Harry wants ten more CoreWeaves: “if you have a public company trading at 20x revenues that has existential losses, you got to use your stock as your currency ASAP… corp dev is going to be looking for 20 assets.” Circle is the obvious next candidate — $156M of net income, still inside its first 90-day post-IPO window, and its stock “is burning a hole in their pockets”; a panelist expects it to buy distribution, since Coinbase is its biggest cost.
• The unifying principle, per Jason’s “Maslow hierarchy of caring”: pre-profit, dilution barely matters — “if you don’t make profits, it doesn’t matter because you’re screwed.” Only once you cross to profitability does allocating value become a real economic decision.

5. Olo at 6x ARR is what rescue actually looks like

• Thoma Bravo takes Olo private for $2B: ~$320M ARR, GAAP profitable, ~20% growth, defensible vertical SaaS. Jason’s deflation of the trade: “6x ARR doesn’t buy the biggest house in Atherton… at Yellowstone Club we’re just going to get a condo.” Rory: “a meat and potatoes deal — a 20% grower, profitable, 6.5x, all’s-right-with-the-world sensible.”
• The round trip stings: public in 2021 at $25 a share, out around $10. The PE playbook from here is what PE “does better than us” — bolt on five adjacent products and take share of wallet from the top 500 restaurant chains, underwriting maybe a 2–5x.
• The caution both agree on: “it’s a start, but it’s not going to save 500 to 700 unicorns single-handedly” — and the bar is unforgiving: above 20% growth and profitable. “There’s too many unicorns that are profitable but not above 20.”
• Jason’s structural aside off the Toast comp ($2B vs $25B): in B2B2C, “the long tail is sometimes where the biggest dollars are” — even Shopify still gets only 25% of revenue from enterprise — against SaaS-board reflexes that always push upmarket.

6. The paradox resolved: frenzy and famine are the same market

• Harry’s framing: crazy fund payouts, PE shopping, meme stocks — and Carta showing VC deals at an 8-year low. Rory says they go together: a “flight to consensus… the running in the early AI markets has been attention begets more attention. If you start to pull ahead, provided you continue to execute, it’s very hard to catch up” — Cursor being the case in point, with 20 rivals “left in the dust… because the sad truth is no one wants nearly as good.”
• Jason’s confession makes it concrete: he passed on a triple-triple-double-double deal with “nothing to criticize” — right numbers, right space, would have done it in early 2023. “I only have so many shots on goal. I don’t want to take this shot today.” People want to swing for the fences or not at all.
• Rory’s pricing formalization is the keeper: if one game carries an embedded 10% chance of a $20B Cursor outcome and the other doesn’t, how much cheaper must the second be? “I fear the answer is either it has to be a lot cheaper or even worse there’s no price at which you’ll do it.” Non-AI compounders “will have a discounted price… I’m sorry for them but it’s just the truth.”
• Harry’s pushback — worth keeping: at Scale he’d go 1,000% for TTDD, where the odds of winning are high, rather than fight Thrive, Founders, and Andre for AI halo deals. Rory concedes only conditionally: it has to pencil, and going early is dangerous because “some of your destiny is outside your control” if follow-on rounds don’t come. Jason’s scar tissue: everyone’s 2017–21 cohort has “decelerated at scale” into $100–200M revenue at 10–30% growth — “the one thing you don’t say to yourself is: this is so much fun, let me add to that collection.” Jason’s closing warning: the industry is repeating its 2021 error with AI — “extrapolation of the current growth rate to the sky” — and should at least price the 20% scenario where data center spend slows.

7. Vanguard’s PE-in-target-date-funds is a greed tell

• Jason’s verdict arrived before the question finished: “terrible.” The structure — partnering with Blackstone to put PE into target-date funds — bothers him precisely because of who’s buying: “a target date fund is a hunt for folks that don’t even know how to spell alpha… and you’re going to put private equity in it.”
• Rory is more clinical: this is the industry hunting new capital as endowments and pension funds wobble — call it “another 20% more money, what the heck” — but venture is structurally hard to fit into liquid retail products (“you can’t target your return date”), private marks are already contested (a likely reference to Elise Stefanik’s crusade over Harvard’s PE accounting), and “the less sophisticated the investor, the less able to take the long view. The retail investor is least in a position to do that.”

8. Universities get squeezed — and VCs quietly pocket a QSBS windfall

• On Harvard’s billion-dollar funding gap and Stanford’s $140M of cuts (blamed on federal research funding plus a potential endowment-tax increase), Rory is blunt about who’s being punished: “a whole bunch of humanities kids yelled in the close about political issues — you’re firing the poor guy who’s been in his lab for 10 years trying to cure cancer… The administrators ain’t getting whacked at scale.” Taxing endowments is, in a sense, taxing venture.
• Meanwhile the same tax bill handed venture its own carve-out: QSBS raised from $10M to $15M of federal-tax-free gains per exit (asset cap $50M→$75M). One speaker, cheerfully compromised: “I got bribed in the tax deal” — stacked across his five trusts it’s “probably $40 to $50 million per exit with no taxes… my quiet motivator to do early stage investing.”

9. Microsoft and Canva show the two speeds of the AI workforce purge

• Microsoft’s 9,000 layoffs replace generalist salespeople with solutions engineers. A panelist’s sizing: “30 to 40% of one-to-two-call sales reps are going to be replaced by AI” — small in the enterprise, where the model is Clay’s forward-deployed engineers instead: “we’re not going to have a guy that doesn’t know our product in the age of AI show up to big deals.” Rory notes it was couched as replace-with-better-people, not replace-with-AI — “and it’s hard to argue with that.”
• Canva’s AI discovery week — 5,000 employees released from their jobs to learn AI, after the CPO pointed out they’d had free ChatGPT and Claude for a year — splits the panel. One panelist: “reskilling doesn’t work”; Canva is basically saying get on it or get off. A panelist on a public-company board was asked what to do with employees who fear AI: “I said you fire them… if you need a rediscovery week, I promise you you’re not going to work.”
• Rory’s synthesis: the week is “performative — we’re-letting-you-all-know expectations,” i.e. notice before the HR conversation. “Don’t over-agonize how you do it. Two, three, four years from now you’re not going to have people who say ‘I’m too busy to use AI’ — they’ll be long since gone.” Harry’s gentler alternative at 20VC: a standing Friday 4–5pm hour to try new tools, then a 5–6 show-and-tell.

10. Kalshi quickfire: base rates, aging booms, and two personnel bets

• On recession by end-2025 (soft landing 76%, high unemployment 16%), Rory calls the market “about priced right”: a Bayesian prior of roughly 1-in-6 or 1-in-7, spiking toward 1-in-4 when the VIX climbs (as with tariffs) and mean-reverting after. Harry would still bet $50K — “maybe 75” — on no recession, because “the technical definition of a recession is pretty tough to meet.”
• Jason’s question — do recession odds rise the longer you go without one? — produces the episode’s best change of mind. Rory first cites Australia’s 17-year run and the line “booms don’t die of old age,” then reverses on air: “Jason is more right than me… when times are good people pile up the dumb aggressive shit to make money… it’s the Charlie Kindleberger thing — manias, panics and crashes — and you’re done.”
• On Shaun Maguire leaving Sequoia this year, the market says highly unlikely; Jason takes the yes side anyway, for real money — not on the merits of the controversy but as a tell: “if this keeps going week after week, it’s a sign his head is not into the investing.” Rory’s counterargument: “a lot of his returns have come from working with Elon, and Elon loves this shit… he might get promoted for this.”
• On Linda Yaccarino, Harry says she stays through end of 2026 despite seeing little case for her (“on a social media platform she’s chosen not to be insightful in public — a bad tell”) — because Elon, who just fired his head of European sales at Tesla, can’t run core X himself: “if it’s not broken, he’s not going to make a change.” And on Musk’s America Party: the Tesla reaction was “shareholder barf,” but Jason doubts it’s a real party — “he’s not going to run candidates in all 50 states; this is a tactical move at the edge of what a political party is.”