Summary

Chime drops its S-1 into a healed market at roughly 50-55% below its last round — the market is only 3% from all-time highs after “the fastest bounce back in the last 20 or 30 years,” and Rory O’Driscoll credits Chime for keeping the S-1 warm through April’s chaos: “weird shit happened, but it appears to be over. Proceed as normal.” The Information estimates a $7-8B valuation versus the $25B last private round; Rory thinks that’s low but plus-or-minus $10B still means the founder learns a hard lesson: “you thought you raised money at 25 billion, but in fact, if you go public at 12, you raised money at 12 and you just didn’t know it.”
Late-stage with IPO price protection is “the world’s best business” — Rory’s hypothetical mechanism: the only risk in a 99%-won’t-blow-up late-stage deal is overpaying, and mandatory-conversion price protection could negotiate that risk away, leaving liquidation preference in M&A, full protection in an IPO, and a 30% pop on the re-corrected price the same day. Harry’s conclusion, endorsed by Jason Lampkin: “seed is for suckers.” The residual risk is IRR, not capital — a 2021 entry distributing in 2027 is “six years to having a modest return.”
The IPO window is open but bifurcated by cost of capital: Klarna-type lenders need public-market access (growth decelerated to 13%, cost of borrowing way up), while Stripe, SpaceX and OpenAI have “infinite private capital at dirt cheap rates, why would you bother?” Jason’s tell: a retired top-10 LP told him “these guys just gotta sell or go IPO. It’s just time” — pressure he expects to cascade LP → GP → CEO for “anything sub-OpenAI.”
M&A buyers aren’t stepping up, and that breaks venture math: two of Jason’s portfolio companies got $500M offers, both walked, and one acquirer instead bought a company valued at $2M for ~$100M “because it was just easier.” Rory’s core fact: “we make an embarrassingly large percentage of our money once every seven years when you’re in the white heat of must acquire, must own high growth venture assets… That’s probably not this year.” And a fund returner doesn’t cut it: “we don’t get out of bed for a fund returner.”
Exit values 5-6x’d, but only because winners stay private longer — VenCap data shows top-1% exits growing from $1.4B (2005-09) to $10.2B (2020-24); Rory’s Google counterfactual (a $23B ‘04 exit that would’ve been $140B in ‘05) proves it’s compounding, not magic. If there aren’t four exits above $65B in 5-7 years, “the people who bought Stripe, SpaceX, Databricks, OpenAI, and Anthropic are screwed. So I don’t think they’re screwed.” Actionable rules: statistically sell 80% of the time at the last round price; Jason’s version — at $2B, sell unless you’re 100% sure it’s a SpaceX.
Rippling wins the Deal lawsuit — both say 100% — “the facts are too bad,” and Jason reads Deal’s counterclaims as “ironically a sign they’re gonna lose,” since counterclaims exist to offset damages, not to win. Rory’s escalation warning: keep yelling and civil drifts criminal — “suddenly someone opens an investigation, then you’re fucked. If I was the CEO of Deal, I’d be like, ‘How much money does it take to settle this thing by Friday?’”
Model providers are the anchor tenants; apps risk becoming “just databases” — Anthropic’s run-rate went $1B (Q4 ‘24) to $2B (Q1 ‘25) with customers up 8X, and one growth investor is trying to buy $1B of employee options with LP supply for five. Jason’s MCP threat: if agents can pull structured data out of HubSpot or Box, “they just become databases… I think they know it.” Rory’s pushback: the replacement cycle runs 10+ years, incumbency prints cash — he holds Salesforce: “I’ll be dead before we rip it out.”

Deep dive

1. Chime dropped its S-1 because the market healed faster than anyone could model

Jason’s opening question — why IPO now? — gets Rory’s “easy” answer: despite the last month and a half, the market is only 3% off its all-time high after “the fastest bounce back in the last 20 or 30 years,” up 17-18% in two or three weeks. “A month ago everything was doomed, now everything’s back… when policy changes that quickly, you really can’t triangulate on that.”
Chime gets credit for shrewdness: S-1 kept on file and updated through the chaos, so the message becomes “somewhere between we filed privately and today, weird shit happened, but it appears to be over. Proceed as normal.”
The numbers Harry pulled: 8.6 million active users, two-thirds using Chime as their primary account (“that’s amazing in my eyes”), $1.67B 2024 revenue, growing 30%+.

2. The business is real — but 75% of revenue rides a Durbin-amendment umbrella

Rory’s bull case: the internet lets you give people a bank account so cheaply — no branches, no branch staff — that you can skip overdraft fees and monthly nickel-and-diming, make ~$250 per client per year mostly on debit interchange, and still be profitable. JP Morgan couldn’t run that model on debit alone; loans and cross-sell are the upside from here.
The “minor negative” is structural: post-‘09 Durbin amendment arbitrage. Banks over $10B in assets can charge roughly 50 bips on debit; sub-$10B banks charge an effective ~1.2%. Chime isn’t a bank but partners with small deposit banks, so the product generating 75% of revenue enjoys a legislative umbrella. No sign of change, “but I gotta believe if you’re Jamie Dimon, you wake up every morning spitting mad that these dudes are able to take your customers because you’re not allowed to charge what they can charge for exactly the same product.”

3. $25B to ~$10B — and the term that decides who eats the markdown

The last private round was $25B; The Information estimates a $7-8B valuation. Rory thinks that’s low, but even at plus-or-minus $10B you’re looking at a deal pricing 50-55% below the last round.
The question Rory actually researched (he pulled the S-1, ran out of time on the articles of incorporation): what are the mandatory conversion terms — does the $25B round get price protection that adjusts it down to the IPO price? His framing of why this matters: in late stage, “99% of them won’t blow up. The only risk you’re running is the risk that you overpay. And if you can negotiate a term that effectively says, ‘Hey dude, if I overpay, you gotta give me more shares such that I didn’t overpay,’ then it’s the world’s best business.”
Harry’s datapoint: General Atlantic (in the Chime round, and lead of Shein at $100B) is “incredibly diligent” about those protections. Rory notes we’ll know before the final S-1 — the adjustment must be disclosed and quantified, “because otherwise the CFO goes to prison.”

4. Ratchets aren’t emotional — and the real risk is IRR, not capital

Jason’s case for calm: if you got ratcheted from $25B to $10B on a 3% position, you take ~3 points of extra dilution “because you lost the bet, but you still won the bet because you got the money.” Rory ran the numbers on the recent example everyone wrote up — ServiceTitan — and agrees: “It’s an economic term. It’s not an emotional thing.” The one part that stings the CEO: the ratchet resets to IPO price, then the 30% pop puts the investor up 30% on the corrected basis “in two hours’ time.”
Jason’s deeper cut: the 2021 vintage still loses on time. “Even if they get a 30% pop and they distribute by 2027, that’s six years to having a modest return.” Rory’s taxonomy: seed runs ~60%+ loss-of-capital risk, his stage 30-35%, but late stage writing those checks should see 90%+ of deals return 1X-plus — “the risk you’re running is primarily IRR risk, not lose-your-capital risk.” The hedge-fund crossover guys literally speak a different language: not “I need a 2X,” but “I want a return of 30% a year.”
Harry’s synthesis, delivered deadpan: “You can overpay by double and still get your 1X protected with a shortened time to liquidity and more money at work. Seed is for suckers.” Jason, the seed investor, doesn’t fight it: “I’m not grumpy about it anymore. I just sign the documents. I don’t even read them anymore ‘cause it don’t matter what’s in them.”

5. Who should IPO now: the window is open, but cost of capital decides

Rory scales the signal properly: Chime at $1.7B revenue is bigger than maybe 90% of talked-about IPO candidates, so this doesn’t open the window for a $200M-revenue company. But the billion-plus revenue cohort now clearly has a choice — and oddly, both answers make sense. Klarna, “fundamentally a lender,” needs continuous capital access and should go public sooner than OpenAI or Stripe; a high-growth AI company with “infinite private capital at dirt cheap rates, why would you bother?”
Harry’s sharper version, which Rory largely accepts: Klarnas and Chimes can’t raise infinite private capital at good terms — Stripe, Databricks, Anthropic, OpenAI can — “and that’s why those go public and the others don’t.” Jason adds the Klarna caveat: growth just decelerated to 13% (per Harry’s partner Paul’s analysis) with cost of borrowing way up. Rory’s punchline: “most companies are more like Klarna than OpenAI… most companies aren’t singularities.”
Jason’s vibe check from EF’s US demo day: a retired, legendary LP — “no axe to grind” — went through his portfolio saying “these guys just gotta sell or go IPO. It’s just time.” When a top-10-of-all-time LP says that, “it may trickle down to the CEOs”: GPs stop tripling down on “anything sub-OpenAI.”

6. M&A: strategics won’t step up, and venture’s profits come once every seven years

On the apparent contradiction between Orlando Bravo’s “cold, quiet year for M&A” and Salesforce paying nine figures for sub-one-year-old Convergence, Rory refuses to arbitrate forecasts (“people’s prognostications of what’s gonna happen in the future are pretty worthless. Including mine”). Both can be true: AI tuck-ins by scared software incumbents will keep coming (see Moveworks); PE buying venture portfolios won’t — “they’ve got a fair amount of indigestion from the stuff they already have.”
Jason’s ground truth: two of his portfolio companies recently got $500M offers from big tech — one a smidge above the last round, one a smidge below. Both walked; one acquirer instead bought a company valued at $2M for just under $100M “because it was just easier,” and “will lose years due to it.” His conclusion: without buyers who “pay 3X the last round,” venture “kind of collapses a little bit.”
Rory’s framing — the quote of the episode: “One of the pressing facts about venture is we make an embarrassingly large percentage of our money once every seven years when you’re in the white heat of must acquire, must own high growth venture assets. And the trick in the other six years is surviving and keeping all the little companies alive… so that when that moment comes, you have inventory to sell. That’s probably not this year.” Wiz is the exception that proves it: they created the leverage, and the buyer paid the step-up.

7. Exit values 5x’d — but it’s compounding math, not a rising tide

Jason’s ownership in those two $500M companies averages ~10% — and he still shrugs: “We don’t get out of bed for a fund returner. A fund returner just returns the fund… I don’t think they’re so great.” The classic seed pattern — one 1X, two 0.05Xs, drib and drab to 3X — doesn’t buy the Marylebone carriage house.
The VenCap analysis (Rory emailed David for the underlying data): the top-1% exit value grew from $1.4B in 2005-09 to $10.2B in 2020-24. But Rory kills the linearity: 2000-04 had exits as high as $3.3B, so it dipped for a decade before exploding. The mechanism is boring: “the longer you hold the company, the more compounding takes place, the more dispersion takes place. The big get bigger, and the shittier ones are crap. It’s just math.”
His killer counterfactual: the largest 2000-04 exit was Google at $23B in ‘04. Had the Google CFO “had a heart attack” and the IPO slipped 18 months, Google’s ~$140B end-of-‘05 market cap would swamp the whole dataset. And the forward bet: 2020-24 had two exits at $65B+; “if there’s not four exits above 65 billion in the next five or seven years, then the people who bought Stripe, SpaceX, Databricks, OpenAI, and Anthropic are screwed. So I don’t think they’re screwed.”
What it does not prove: that mega-funds all work. “The bigger your fund, the more imperative it is you have to be in those six deals, which explains why capital is so easy to raise for those companies… Does that translate into all the funds making enough return to make the late stage math work? Not as clear.” Harry’s worry stands: “so few companies in that 99.9% decile — a world of concentration unlike any we’ve seen.”

8. Hold or sell: statistically sell 80% of the time — unless you’re sure it’s SpaceX

Rory reaches for Mary Meeker’s old IPO analysis, because “private late stage companies in 2025 are just the same asset class as IPOs in ‘95 to 2005”: most companies barely beat their IPO price a year later; a small number compound and cover everything. So if “the great secondary gods” offer you last-round price on everything, “statistically 80% of the time you should sell” — but “compounding is a very forgiving thing”: the second-best strategy is holding them all, provided you have one of the good ones — “the compounder that forgives all sins.”
Jason’s spreadsheet-derived rule: “at 2 billion, sell unless you’re 100% sure you shouldn’t, as a seed manager. I know it sounds goofy, but it ties to doing better than returning the fund.” Rory endorses the secondary push outright: “I don’t think you can take the risk of doubling down ad nauseam when you’re 10 or 12 years in. It sucks, but there you are.”
The counter-story is Veeva: the only real investor held 30% at IPO — worth $2.4B on a roughly $250M fund — LPs were mad they didn’t distribute everything, and today it’s worth $25-40B: “created billionaires by holding.” Jason’s hedge on knowability: he knew Peter Gassner was “fucking off the planet in terms of quality as a CEO” — best of a batch that included David Sacks — “but I didn’t have the numbers.”

9. The death of IPOs is a capital-allocation problem — and private capital’s true price is hidden

Rory’s real-time realization: public markets let everyone choose — GP distributes and holds, LP sells, “choice leads to optimal outcomes.” Private, “I either have to sell now, which maybe is not what I want, or ride it out for five years, which is maybe not what my LP wants. That’s inducing tension in the system.” Jason adds taxes as a hold incentive — ~40% long-term all-in in California without QSBS, while most LPs pay none. (Cue the Wall Street Journal chart Jason cites: 90 US companies worth over $1B totaling $2.5 trillion vs. the EU’s $333 billion — a big chunk of which is Stripe.)
Rory’s fix-the-market ideas: time-based voting (the Texas exchange concept — share weighting proportional to holding period, to blunt short-term arbitrageurs) and Google’s old trick of simply not doing quarterly calls. Jason counters with Brian Halligan’s take as HubSpot chairman: at $30B scale, “it’s not much more work being a public company than it was being late stage.”
Rory’s concession lands on economics: “CEOs respond to cost of capital signals,” and private capital remains bizarrely cheaper than public — except when it isn’t. The Chime kicker: with the ratchet, that $25B round’s true cost “was twice as high as you thought — you didn’t give away 4%, you gave away 8%. We may be in this little bubble where we actually don’t know the cost of capital for some of these late stage rounds.”
On leaving California for tax reasons, Rory’s math is final: “If I were to reduce my income tax by 25% by moving, I would also reduce my net worth by 50% ‘cause my wife would be staying behind. I’m at peace with paying whatever Gavin Newsom needs.”

10. Prediction markets: OpenAI converts (eventually), GPT-5 splits the panel

On Harry’s prediction-market picks (likely Kalshi; heard as “CalSheet”) — will OpenAI stop being a nonprofit? Rory bets yes on a technicality: “the question is weakly phrased enough… there’s no timing.” His fuller view: the nonprofit halo was a recruiting device — OpenAI and Anthropic embraced it because “the most important audience for both of them was talented AI engineers, and all of them shared the religion.” Endgame: “some half-assed cobble compromise whereby the entity will be a PBC, the not for profit will be one level up… It’ll just be a wild and wacky journey,” with Brett Taylor smart enough to untangle it.
Jason’s no, from experience: “The folks I’ve seen on nonprofit boards are not gonna give up this power… there’s no money in it, so it’s all about the power.” And his standing amazement: “I’ve never seen a dysfunctional company that’s more successful than OpenAI. All the founders left. They fired the CEO, brought him back… The momentum’s crazy.”
ChatGPT-5 revealed this year: Jason says no — merging all the models “makes no sense at a consumer level” yet, “it wouldn’t be a surprise to me if it pushes a year or longer” — while conceding OpenAI’s talent is “so next level.” Harry takes the yes: “the velocity is insane.”

11. Rippling vs. Deal: unanimous — and the counterclaims are the tell

Both investors call it flatly: Rippling wins, “100%. The facts are too bad… They stole trade properties. This is a classic case.” Rory, who read the counterclaim: “It was some version of you did it too… the fact pattern looks crap, and at some point sense prevails, and you say, ‘Whoopsie, sorry,’ and settle.” Harry’s disclosure — Deal shareholder, Alex a dear friend who’s “been abroad for many years” — gets Rory’s reply: “He ain’t coming to Dublin to testify on this puppy.”
Jason’s litigation mechanics — worth keeping: US counterclaims can offset damages even when time-barred, so losers stuff everything in. “They’re doing it not because they think they’re gonna win. They’re doing it because they know they’re gonna lose… All those counterclaims are ironically a sign they’re gonna lose.” An innocent CEO does what Sam did with Elon: “‘Sorry we misunderstood each other, Parker. Happy to have a beer and talk it out.’ That’s what an innocent CEO says.”
Rory goes darker: the allegations could read criminally, and noise attracts the FBI’s attention — “suddenly we’ve drifted into some kind of theft of trade properties, and suddenly someone opens an investigation, then you’re fucked. If my ass was on the line and I was the CEO of Deal, I’d be like, ‘How much money does it take to settle this thing by Friday?’”
The shared war story on lawyers: every case starts “super strong” and converges to 50/50 at the courtroom door — Jason: “it’s right around $2 million of legal expenses they change their tune.” Rory’s wife, a former criminal defense attorney, supplies the coda: “the worst defendants are defendants who start talking about principles.”

12. SaaStr’s reset: ban the past, embrace the reaping — and own the anchor tenants

Jason banned the past across all 300 SaaStr sessions — “you were only allowed to talk about AI today and tomorrow” — and 2024 turned out to be “the end of the Debbie Downer era.” Rory’s historical rhyme: post-crash survivors were too scarred to think big, “and many of those companies, as a result, didn’t make it.” His verdict: “this whole SaaS is dying is bullshit. It’s changing, and you better be AI forward or dead” — the right moves can re-accelerate growth. From Jason’s CMO event: everyone expects 20-30% of their teams replaced by AI and, behind closed doors, is “borderline excited” — “no one was regretting the impact on culture. Seriously, no one.” Rory’s label: Jason is “the happy Grim Reaper — ‘I love my work. Let’s do some reaping here.’”
Moats are weaker everywhere — even Microsoft open-sourcing VS Code is, per Rory, “a sign of relative weakness, not absolute weakness… When you have a dominant position, you don’t have to be nice.” The playbook is the “run fast deal”: use the two-to-three-year AI-magic wedge to grab distribution, then build defensibility before the core commodifies — Gong being the model, riding voice into forecasting and CRM-updating and now re-accelerating. Rory’s self-aware caveat from ‘96: “I can remember being a snarky little 30-year-old VC making snide comments that Amazon was just a bookseller.” Still, his partnership is deliberately adding N-of-1, high-IP deals — “I don’t want to wake up with every deal being GPT Plus and 27 competitors.”
The layer that isn’t commodifying: model providers. Anthropic’s run-rate grew from $1B in Q4 ‘24 to $2B in Q1 ‘25, customers up 8X past 100,000; a growth investor told Harry that morning he’s “trying to buy every employee’s options… I’ll get a billion dollars’ worth. I’ve got supply from my LPs for five.” Rory never bought the commodification argument: “there’s gonna be two or three of them, not 10… they’re the anchor tenants of the AI economy,” even as OpenAI says it will burn at least another $44B before 2029 profitability.
The losers and the survivors: Chegg went $12B to $95M (“why would I pay Chegg for X when I can literally type it into ChatGPT?”). Jason’s next-Chegg mechanism is MCP: if agents can pull structured data out of HubSpot or Box, “they just become databases… I think they know it. That’s why everyone’s stressed.” Rory’s pushback is on the word instantly — the replacement cycle “plays out over 10-plus years,” and incumbency prints cash: Oracle is “a fricking database” at 43% operating margins. He holds Salesforce: “I’ll be dead before we rip it out.”

OpenAI Enters the Hardware Business, The Challenges and Opportunities for Jony Ive, Takeaways from Google I/O 2025 Bitcoin Above $100k, Is Alt Season Here?