Summary

The venture market has split into a barbell: Carta says Series A’s are down 81%, yet Owner just raised $120M on ~$40M revenue growing 10% a month, with the data room opened Monday and every term sheet in by Wednesday. Jason Lemkin’s meta-lesson: “triple triple double double still good enough,” and inside that box “there is just unlimited effing” capital — while for everyone outside it his verdict is “cry me a river… be best of breed, you’ll get five term sheets.”
Jason’s board-level data from Gorgias reframes the Klarna AI “walkback”: 20,000 SMBs have a literal AI-support slider, the average sits at 20%, and exactly two of 20,000 run at 100 — Klarna slid to 100, learned, and will rehire maybe 200 of 1,000 while “800 will still be AI’d.” Rory O’Driscoll, who disagreed a week ago, now concedes LLMs can take 50-70% of tickets with NPS improving: “a 5-year trend, not a 1-year moment.” Jason predicts: “by the middle of next year in tech almost every single person’s job will have changed.”
On OpenAI, Rory’s flat verdict is “they have won” (Jason: ~85% ChatGPT share; Claude “has fallen behind”), and Rory discusses concentrating a fund into it, citing David putting a third of Craft Fund I into SpaceX. Rory’s brake: at roughly 4 going to 12-ish in revenue, he questions the implied multiple, leaked forecasts outgrow Google’s, and the Nifty 50 and 2000 Nasdaq prove you can overpay for a right idea “and be wrong by five or six years” — plus the entry ticket was “show up with 250 million or don’t show up at all.”
Perplexity’s rumored $500M at $14B (Excel-led) on maybe $200M ARR is, in Harry’s framing, not a revenue multiple but an at-bat: “a credible one in three, not equally weighted” at a trillion-dollar prize, since Google is the comp and only OpenAI, Anthropic, and Perplexity are in the arena. Harry’s memo epiphany: add a trillion-dollar row to the IC template — “if it’s north of 2% odds, we do the deal” (his first-ever memo gave Pipedrive 5% odds of a $100M exit; it sold for $1.25B).
Clay at $1.5B is being bought out of fear, not vision: “every 2021 CMO that is scared they’re going to lose their job” is deploying it “like Hopin in 2020, but for marketers,” and fear finds six-figure budget “tonight.” But the knives are out — Jason met a Stanford dropout who said he had dropped out with his daughter, already at $2M with an easier-to-use competitor — so his advice is raise primary, not secondary: “I would armor up if I were Clay… and just scorch-earth everyone in the space.”
Tiger’s ~$12B, 300-350-deal 2021 fund is salvageable only through bet sizing — “it’s just hard to dig out of a $12 billion hole in $100 million increments” — but if it put real weight into OpenAI and Scale, “this could be the Tiger surge.” Coatue’s $50K-minimum retail vehicle splits the panel: Jason smells Boiler Room (“VTI is the perfect product for 99% of people”), Rory calls it inevitable — “there’s a PowerPoint on the desk at Andreessen, at General Catalyst, and at Lightspeed.”
Fewer, bigger winners is repricing portfolio construction itself: Rory is moving his target from 20 deals to 25-28 because exits now come at 300 rather than 150, two-three more years in which “one in three of these companies will fuck up” — more at-bats, fewer but bigger outcomes, all the way up to OpenAI reaching today’s-Facebook scale before it ever goes public.

Deep dive

Note: despite the feed title, this episode is the 20VC roundtable — Harry Stebbings with Jason Lemkin (SaaStr) and Rory O’Driscoll (Scale) — not the Figma CPO interview.

1. Owner’s $120M: inside “the box,” capital is unlimited

Jason led Owner’s seed and learned the round size only when he read it today — “I don’t even read the document… All I care about is my ownership.” It’s $120M (Harry had heard $90M at $900M) on roughly $40M growing 10% a month, and he’s candid that “in some ways it’s also a pre-AI company… it’s really good software.”
The process still gives him hives: founder Adam socialized VCs over time, then “open the data room on Monday, get two term sheets that afternoon, and get all of the term sheets by Wednesday.” Insiders alone could have filled $150M.
The meta-lesson: “triple triple double double still good enough” — and if you’re in that zone, “there is just unlimited effing” capital. “But you got to be in that box,” and getting into the box is harder than ever.

2. If Monday is your first meeting, you already lost

Harry’s dilemma — “great founder, we need to decide by Wednesday, and it’s Monday. I can’t write a $15 million check” — answers itself, per Rory: you’re up against someone who met them last round and did the work the prior week, so for them “Monday is just confirmatory.” Two interactions can maybe assimilate the data; “it’s very hard to assimilate the person.”
Rory’s discipline: a Salesforce hot list of 10-20 named companies he could imagine backing within 12 months — without it, “I’m probably not doing my job.” You can woo 10 people, not 200; “if you’re just wandering around hoping shit’s going to turn up on Monday that will make you money on Wednesday, it’s not going to be successful.”
On preempting a seed months later, Jason sees no problem: diligence done, two investor updates in, the company’s grown 50% — and Sequoia has made an art of “second and third and fourth chances” (it just did Clay at $1.5B).
Rory admits the bias — paying “twice as much as someone 3 months ago… feels like an idiot, which is why we haven’t done it” — then questions himself: the best companies raise at the highest velocity. “We put in 100 million, the model got smart. Let’s do a billion. Got smarter. Let’s do 10… you’re going to double down until it don’t work.”

3. Series A’s down 81%: “cry me a river”

On Carta’s 81% drop, Rory’s structural read: seed is the “believe in the team” round, the A is “show me the traction” round — “belief is easy to manufacture and traction is hard.” Smart founders now behave like barristers: “never call a question unless you know the answer” — don’t go out unless you’re pretty certain you’ve got it.
Jason is blunter: Series A conversions are down — who cares. “Go make your startup S-tier… be best of breed, you’ll get five term sheets.” He lived the cycle himself as a founder: “easy to get funded, impossible, easy to get funded, impossible” — crying about a hard A “is a B-tier way to look at it.”
Rory’s analogy comes from his own stage-4 colon cancer at 40: survival statistics interest doctors managing populations, but “to the patient it’s 0.01. You either make it or you don’t.” Startups are identical — either you have something worth funding and the statistics don’t matter, or you don’t.
Why he came back after ~18 months of chemo (taking calls lying on the floor because it hurt): no yurt, no Everest — “I like my work and I just like to keep doing it and you just do it until you can’t.” His yay-venture footnote: he was among the first people to get Avastin, the Genentech drug — and Kleiner funded Genentech.

4. Tiger’s surge — and whether a convicted Sam gets re-funded

Jason and Rory on whether OpenAI and Scale rescue Tiger’s ~$12B, 300-350-deal 2021 fund: it’s purely bet sizing. “If they did 1/350th in OpenAI, it’s not going to bail them out even if it’s a 10X or 30X”; 10-15% of the fund into something that 7-8x’s and “maybe they’ve snatched victory from the jaws of defeat.” Otherwise, “it’s just hard to dig out of a $12 billion hole in $100 million increments.”
The vintage nuance: any 2021 exit was really a 2018 late-stage deal — buy at $2B in ‘18, sell at $6B in ‘21 — so late-stage funds will post a very nice ‘18 fund while earlier funds missed the window.
Jason thinks the criticism is overdone: hold OpenAI, Scale, and a decent back catalog and “they will 1x. They won’t lose money.” Rory grants that digging out counts for points: “like Bush in Iraq — a mistake to go in, but at least in 2006 he surged his way out. This could be the Tiger surge.”
The FTX detour: Sam’s early positions in Anthropic and Cursor — “astonishing… before the ChatGPT moment” — would have saved the fund’s performance absent the commingling; “he had an excellent business, blew it with fraud, and did some great venture. Quite a polymath.” Would he be re-funded? Jason: someone will — “just a little hundred at a billion post.” Rory: no — “once you pass the convicted criminal stage, the bar goes way up”; that’s a different category from WeWork-scale hubris.

5. Coatue’s $50K retail fund: democratization or Boiler Room?

Jason’s instinct is Boiler Room: “Let’s just rip off the retail investor… ripping people off for 50 grand that don’t know what they’re getting themselves into.” His standing advice to every individual, including prospective LPs in his own funds: don’t — “no one understands the illiquidity… VTI is the perfect product for 99% of people. You cannot beat it.”
Rory’s wider frame: Blackstone-style PE already raises an expanding share from individuals, and the whole thing is round-trip madness — “these companies should be public and then Fidelity Growth Fund could do them at 50 bips… this is a workaround for a problem that would be better solved some other way.”
Jason’s three failure modes: manager (fine — Coatue is “in the flow” and will get deals), timing (“if the next 5 years are tough, no one can save you”), and structure — redemption gates, as the Blackstone/BlackRock real-estate fund showed. The fees (Jason quotes 12.5% carry and 1.6842% fees) blend out because half the assets are public: “It’s not cheap. It’s market.” Neither got the Dell/Bezos anchor bit: “Jeff Bezos and Michael Dell make the 50K sophisticated-investor threshold.”
Will it spread? “Anything that works in finance gets copied immediately.” Rory: “I’m sure there’s a PowerPoint on the desk at Andreessen, at General Catalyst, and at Lightspeed… the movie’s inevitable. There’s no point getting all moralistic about it.” Jason’s residual test is smell: good greed aligns on the cap table; SPACs and certain SPVs “smell too greedy.”

6. The Klarna slider: two of 20,000 run support at 100% AI

Jason defends his $100K jobs bet with data from Gorgias (likely; he’s on the board): 20,000 SMBs have a literal AI-support slider — the average is 20%, invested customers hit 40, and exactly two of 20,000 went to 100. Klarna did what those two did: slid to 100, learned, slid partially back — “he did not say I’m rehiring everybody”; maybe 200 of 1,000 return, “800 will still be AI’d.”
The direction is toward more AI: “every 3 months that slider’s going to move closer to 100,” and the new “five-person billion-dollar startups” will accept wrong orders and wrong answers rather than hire humans. Rory, who disagreed last week, converts: depending on where companies are with LLMs, they can get to 50-70% “without any deterioration in service” and NPS improves. “This is just a 5-year trend, not a 1-year moment.”
Harry’s strategic read: it’s IPO messaging. Duolingo’s founder told him public markets are binary — AI winner or AI loser — so Sebastian sang AI-first heading toward an IPO; no longer going public, “he can save that for the next day when I do want to go public.”
Jason widens it to shock therapy: Microsoft cut 3% — “It’s not enough”; big-company CEOs privately say “I’m not sure I need 80% of my team today”; Fiverr’s CEO declared every job at risk including his own. His forecast: “by the middle of next year in tech almost every single person’s job will have changed.” His only urgency fix for a 10,000-person company: 100% RTO in 30 days, keep only S-tier engineers — while admitting he goes in two days a week himself.

7. Two non-technical CEOs atop OpenAI — “I just still think it’s weird”

On Fiji Simo as CEO of Apps, Jason shrugs — “They weren’t going to start being normal now… 10 years into the least normal startup on the planet” — and Jason’s objection isn’t the title (Salesforce has had division CEOs; “every VP of sales is a CRO today”) but that both CEOs are now non-technical. His life experience: non-technical CEOs “almost all fail” because the rate of change outruns them — “I could spend 100 hours with the smartest people in the world. I will never understand how OpenAI works.” Sam’s offset: he “can recruit like no one on planet Earth.”
Rory won’t reweight the odds: “They have won.” Jason: ChatGPT has ~85% share and growing — “I love using Claude… I can already see how far it’s fallen behind.” Rory updates one prior: he doesn’t know if they’ll do customer success (“so idiosyncratic”); the next apps stay broadly horizontal — consumer, shopping. “You thought Evernote would know everything about you. You ain’t seen nothing yet.”
On structure, they’re not reverting to nonprofit: the operating company becomes a Patagonia-style public benefit corporation, the nonprofit foundation keeping board control and a large economic stake. Rory’s kicker: “that structure is actually what Anthropic did from day one. They got it right.”
Microsoft’s deal, once mocked as a backdoor license around M&A antitrust, now looks “epic” — everything plus 10% of all revenue until AGI. Converting caps and rev-shares into a simple ownership percentage, with Microsoft holding a block and strong leverage, is the hard part: legal bills “could be 10 million a month,” and “a legion of children are going to go to college on this one.”

8. Fewer, bigger winners — and whether to just buy OpenAI

Rory: “I am so sure that OpenAI is going to be a 1.5 to 2 trillion dollar company.” His precedent: David put roughly a third of Craft Fund I into SpaceX — “it seemed like a nutty move, but it was highly concentrated can’t-lose.”
Rory’s portfolio math for this world: target count moves from 20 deals to 25-28, because “instead of exiting at 150, we’re going to exit best case at 300” — two or three more years in which “one in three of these companies will fuck up.” More at-bats, three winners instead of four, but bigger — and at the top of the stack, OpenAI staying private might reach today’s-Facebook valuation before ever listing.
His brake on the trade: OpenAI at roughly 4 last year and 12-ish this year prompts questions about the implied revenue multiple; it’s tracked Google side-by-side to now and leaked forecasts show faster growth than Google — but “you can overpay for growth assets and be wrong by five or six years. The Nifty 50 in ‘68 didn’t come back for 10 or 15 years… the Nasdaq didn’t come back for 14.”
The practical bar: the last round demanded “show up with 250 million or don’t show up at all.” Rory, with his $30M check size: “I can make a phone call, but I don’t think they’ll get back to me. I’ll leave a voicemail.”

9. Perplexity at $14B: buying a one-in-three at-bat at a trillion

The rumored round: $500M at $14B, Excel-led, on ~$100M ARR announced recently — “let’s say they’re at 200 today, which is still generous.” Rory likes the asset (“the closest to a Google on steroids”) but questions whether a distant number three deserves this premium versus even Anthropic.
Harry’s case for paying anyway: only three players are relevant — OpenAI, Anthropic, Perplexity — and Google is the trillion-dollar comp. “What Perplexity is selling is an at-bat, a credible one in three, not equally weighted… OpenAI is clearly going to win, but maybe you can be third, and that’s worth the downside if the prize is a trillion bucks.”
Harry’s ’90s rhyme: four search engines IPO’d in ‘98 — Lycos, Excite, Yahoo — “you don’t remember any of them except Yahoo,” but all had a chance. Plan B is M&A: “there’s a couple of trillion-dollar companies who are like, shit, maybe I’d like to mess with Google’s head.” Harry (an investor — “otherwise my CFO will kill me”) adds the overlooked piece: European telecom partnerships give default access to consumers’ phones, exactly Google’s early distribution playbook.
Harry’s memo epiphany: his first-ever deal, Pipedrive in 2013, was written up as “5% chance of a hundred million-dollar exit” — it sold for $1.25B. The template needs a new row: odds of a trillion — “if it’s north of 2% odds, we do the deal.” Rory’s inversion: all the risk of a classic private company with no embedded upside means pass — “it’s the pixie fairy dust that lands on our portfolio every 5 or 6 years that makes the math worthwhile.”

10. Clay: selling into fear works — until the ground is waterlogged

Harry floats Clay ($1.5B employee secondary with Sequoia) as the next credible threat to Salesforce; heads shake. Rory’s product reality: a brilliant pre-gen-AI rev-ops tool that consolidates data sources into one clean list (your 10,000 B2B CFOs), now adding “Claygents” moving down-funnel toward AI SDRs — “so far from Salesforce CRM that I can’t even think about it in any useful fashion… which should never get in the way of a good story from a VC.”
Jason’s ground truth from the SaaStr pre-event: the pull is fear — “every 2021 CMO that is scared they’re going to lose their job” is hiring Clay consultants and writing checks “like Hopin in 2020, but for marketers.” Fear finds budget instantly: “All I have to do is come up with 200k for Clay. Done. Send over the contract tonight.” Rory’s caveat: selling fear ahead of the product is legit “provided the product gets there” — Hopin’s end demand died: “if you seize ground but the ground turns out to be waterlogged, you’re toast.”
The knives are out: Jason met a 19-20-year-old Stanford dropout who said he had dropped out with his daughter — already at $2M in a couple of months with an easier-to-use Clay competitor. Gong’s edge took the market 4 years to decode; “now it’s happening in days… or weeks.” His prescription: forget secondary — “I would armor up if I were Clay. I would raise another 100 million and I would just scorch-earth everyone in the space.”
The closing disagreement is video: Rory calls complex AI video editing “so far off”; Jason’s jaw “drops with Higgsfield” — “I don’t even think there’s a point in having a static marketing image after Higgsfield” (Jason says capability advances “5 or 10% a year”; Harry counters “5 or 10% a month”). Jason says differentiated 20VC clips become more valuable when everyone can make clips, while Harry says infinite AI content makes “discoverability a massive problem.”

Ed Helms Answers Your Hard Questions Sundar Pichai, CEO of Alphabet | The All-In Interview