Summary

Figma’s 250% pop was less a recoverable $3 billion pricing error than a scarcity-and-FOMO event created by the IPO itself. Figma had roughly 40X demand against a tiny float, but nobody had bid at $98; Rory O’Driscoll’s key line was, “The $98 price only happened because the IPO happened at $38.” The company might have captured another few dollars, but the spectacular opening valuation was not accessible the night before.
The real IPO pricing trade is a modest discount in exchange for a durable institutional cap table. Brian Halligan recalled choosing HubSpot’s $25 price even after being told Fidelity was “in at 24, out at 25,” because founders want Fidelity, T. Rowe Price, Wellington and Capital Group holding for years. The Figma wrinkle is that a price instantly above institutions’ internal targets may force selling; Rory expects that, while Halligan thinks long-only funds kept their scarce “toehold.”
Direct listings would not necessarily prevent mega-pops, because the public print changes demand psychology. Rory said amended SEC rules now allow capital raising through a direct listing, and cited Amplitude as a possible example, but argued that a direct listing might have captured $2-$3 more without putting Figma immediately near its euphoric trading price. With public-market money now potentially cheaper than private capital, Halligan’s advice to Canva was blunt: “Run, Forrest, run.”
Founder compensation is moving from low-risk RSUs toward moonshot packages, but stock-price triggers remain badly designed. Halligan argued that RSUs behave like cash and dampen risk-taking, while PSUs can recreate option-like upside appropriate to a founder whose normal peer compensation is immaterial. Figma’s targets reportedly ran to roughly $118 and were immediately reached by the IPO pop, leaving seven-year vesting intact but making the “performance” hurdle vanish.
Meta and Microsoft’s AI spending is being financed by extraordinary legacy businesses, not yet justified by AI application revenue alone. Rory framed the permission structure as: “Real men with $70 billion of free cash flow get to spend 40 billion of that on servers.” Against perhaps $25-$30 billion of application revenue, annual AI CapEx of $400-$600 billion implies eventual retrenchment and casualties among marginal or leveraged players, even if the ten-year technology thesis is correct. Halligan separately noted that, from August 26, 2022, Oracle and SAP stock prices were up about 230% and Shopify’s about 300%.
AI has found product-market fit in prosumer tools and large enterprises, but the SMB middle remains conspicuously uncracked. Lovable, Replit and Gamma work with limited bespoke setup, while enterprise deployments can fund teams and forward-deployed engineers; ordinary small businesses can do neither. Jason Lemkin’s investable challenge is to turn something that can take six months to train into a product that self-configures “in 60 seconds.”
Cognition’s rumored $15 billion round prices scarce AI distribution while its Windsurf deal exposes the human cost of improvisation. Harry cited roughly $85 million of revenue for each business, while Jason later framed the acquired revenue as about $80 million. Jason read the deal as buying a brand, revenue and three-to-nine months of market entry—not primarily the team. Laying off 30% and offering the rest nine months’ severance or an 80-hour, six-day office schedule showed how vulnerable employees become once outcomes depend on “the kindness of strangers.”
Ramp’s $22-$22.5 billion financing and CRV’s retreat to early stage illustrate two rational—but opposite—uses of capital. Ramp may genuinely need billions of funding capacity because interchange requires financing customers’ revolving balances; $500 million also costs only about 2% dilution. CRV, meanwhile, is choosing clarity and faster carry over a distracting opportunity fund—the broader lesson being that firms need to know “what game they’re playing and play it well.”

Deep dive

1. Figma’s planned small explosion became a 250% detonation

Halligan’s inside view begins the night before an IPO, when exhausted founders must choose both price and holders. HubSpot was 27X oversubscribed and Figma roughly 40X, yet the allocation book still reflected a negotiation: founders wanted long-only institutions, while bankers also wanted their hedge-fund relationships accommodated.
At HubSpot, Morgan Stanley recommended $24 because Fidelity was supposedly “in at 24, out at 25.” Halligan played chicken, held at $25 and got Fidelity anyway; the stock opened around $33. He did not allege collusion, but Morgan Stanley and Fidelity were functionally “on the same side of the table” against the issuer.
Founders deliberately want some upside after pricing. A 15%-20% pop makes long-only institutions feel good and encourages long holding; in Jason Lemkin’s separate example, employees also benefited from opening-day gains. Figma instead produced what Rory called “this absurd 250% pop,” apparently the largest since 1999. The mechanism intended to create a small explosion accidentally created a spectacular one.
Jason Lemkin’s corrective to social-media outrage: “There’s no one dumb at Figma.” Dylan Field, his finance team, CFO and advisers understood the standard trade-offs; the company had mostly secondary stock, very little supply and intense demand. The result exceeded their design, but that is different from a rookie team blindly mispricing its company.

2. The opening price was not money Figma could have raised

Harry’s question was whether Figma was underpriced or simply overpriced once exuberant buyers pushed the shares toward the roughly $98-$145 levels cited during the discussion. Rory rejected the simplistic “$3 billion left on the table” arithmetic because the institutional order book contained bids around the offer—not at $98.
Rory’s causal claim is the episode’s sharpest: “The $98 price only happened because the IPO happened at $38.” Had Figma tried to raise at $80 the prior evening, “they wouldn’t have had a book.” The realistic debate was Halligan’s familiar choice between pricing perhaps $2 higher and losing Fidelity, or $2 lower and securing it.
The more serious problem comes after trading begins. Institutions entered with internal exit targets; Rory doubted any exceeded roughly $110. A fund underwriting $50 in two years must reconsider when its position reaches $100 in two days: “At least half those mutual funds” might sell, undermining the stable ownership the discount was meant to purchase.
Halligan pushed back from his experience watching HubSpot’s register: so-called long-only funds do trim, but he doubted Figma’s core institutions immediately abandoned such a scarce allocation. His alternative culprit was hedge funds—and “Harry Stebbings that sold”—while institutions retained a valuable toehold they could build over time.

3. Direct listings cannot eliminate public-market FOMO

Rory said amended SEC rules now allow capital raising through a direct listing, and cited Amplitude as a possible example, but disputed that this would have put Figma straight onto its euphoric public valuation. The initial analysis still might have cleared around $39-$40; a direct listing could capture a few extra dollars without guaranteeing the $80-times-revenue psychology that appeared after trading.
His deliberately strange analogy was that mega-pops may be intermittent natural phenomena “like earthquakes.” Circle, CoreWeave and Figma clustered as markets rapidly moved from April pessimism to summer risk appetite; a limited set of speculative assets then attracted more capital than cautious pre-IPO price discovery could anticipate.
The venture analogue is a branded investor winning a round at 120, followed six weeks later by newcomers paying 350 although nothing operational changed. The brand’s participation creates the evidence others needed: “This is kind of the public market equivalent of FOMO. That’s how FOMO manifests.”
Halligan still considers direct listings risky unless a company has Google- or Facebook-level scale. The conventional process costs about 7% to the bankers, but properly markets the business to long-only buyers. The fee is small relative to the enduring downside of beginning public life with the wrong owners.

4. Timing and investor selection can shape a company years later

Before HubSpot’s IPO, a non-deal roadshow converted nearly every institution the team met. Skipping T. Rowe Price in Baltimore and Capital Group in Southern California proved costly: it took three or four years to persuade them to buy. Halligan’s lesson was literal as well as strategic—meet important prospective holders in person.
HubSpot went public in a decent window and secured the long-only base it wanted. Zendesk listed three months earlier during a shaky week, required VC participation and remained heavier in hedge funds and individuals. Halligan believes timing was “more luck than skill,” but “having a good cap table’s underrated.”
Rory reconstructed the downstream implication: Zendesk’s weaker register was not the sole reason activist pressure eventually helped force a reluctant sale, but it may have mattered at the margin. Halligan’s willingness to concede a dollar to Fidelity was therefore not generosity; it was insurance against being poorly supported when the business later hit turbulence.
Jason added an overlooked beneficiary of underpricing: employees. At his first startup, staff struggled to find $50,000-$70,000 for the directed-share program, then effectively made about $100,000 each on opening day. Whatever the dilution debate, “it was a magical moment” for people who cared about their stock rather than abstract issuance efficiency.

5. IPO economics make founders surprisingly indifferent to dilution

Halligan described pricing night as the low-battery endpoint of two weeks, 12 countries and six daily pitches, immediately followed by a party nobody wants. The next morning brings another status fight over who stands on the exchange platform, then hours of watching monitors while the opening auction slowly finds a price—about two hours generally, and six hours in HubSpot’s case.
When HubSpot finally traded, co-founder Dharmesh Shah showed Halligan the app and said, “Brian, look, we’re worth a billion dollars.” Halligan replied, “Take a screenshot. We’ll never see that again.” His practical exchange choice was equally human: NYSE and Nasdaq seemed virtually identical, but NYSE let them ring the bell.
Asked whether he worried about IPO dilution, Halligan answered: “My net worth went from X to 100X. I would just not care.” Rory called that the reason reform is difficult: founders do this once, bankers do it daily, and the party with life-changing wealth has little incentive to reconstruct the system over one disputed dollar.
Founder liquidity remains “a drip and a drab,” not immediate monetization. Halligan has sold the same number of shares monthly since listing to avoid signaling, while later equity grants replenished some ownership. By contrast, Scale had to distribute HubSpot shares and missed much of the public run from roughly $1 billion to $10 billion and then $25 billion.
The same timing lesson appeared in HubSpot’s 2009 Series C. After roughly 20 meetings on Sand Hill Road during the recession, Halligan said every big-name firm passed; Rory and Rob offered a term sheet at a $66 million pre-money valuation when HubSpot was doing roughly $7 million-$10 million and growing about 2X year on year. Buying when others were not buying was presented as the investor’s advantage.

6. CEO pay needs real upside, not disguised cash

Halligan called CEO compensation “pretty broken” because the post-2006 move from incentive stock options to RSUs replaced asymmetric upside with something close to cash. RSUs fluctuate, but remain valuable without a major win; options make the executive “swing for the fences,” a risk posture growth companies should sometimes actively want.
Peer benchmarking compounds the problem. HubSpot might pay its CEO near the 75th percentile of 20 comparable companies—roughly $20 million—but the same amount would be only about 0.3% relative to Dylan Field’s stated wealth context and would “not move the needle an iota.” Halligan favors matching incentives to the CEO’s net worth, as with Elon Musk, not merely Mary Barra’s $29 million compensation.
Rory explained PSUs as RSUs that vest only after a performance condition, usually a stock-price threshold. They recreate an option-like payoff without an explicit exercise price. He agrees with their direction, but Figma’s pre-IPO targets reportedly topped out near $118 and were immediately achieved by the pop, even though the shares still vest over seven years.
Halligan’s preferred design uses net-new ARR plus an earnings floor. Rory strongly preferred those operating goals to market price, while acknowledging disclosure problems: targets can guide analysts and circumstances change. HubSpot previously used measures such as NPS, which became even trickier to disclose and administer.

7. Moonshot grants are becoming part of growth-round pricing

Jason sees private growth investors institutionalizing moonshot packages: invest at $3 billion, then promise founders perhaps another 7%-10% if the company reaches $30 billion. The award is not simply a replacement for dilution; the investor is offering a large package while requiring a roughly 10X outcome.
Rory’s objection is that 10X’ing an already elevated valuation bases compensation on “a chimera.” If revenue triples and operating income turns positive while multiples normalize, he expects the founder to demand a recut. Jason wagered the opposite: founders now knowingly sign extreme hurdles, and “grouchy VCs” will not waive them.
Halligan supports restoring ownership for an excellent founder diluted during four slow years before growth arrived. He also favors modest secondaries: selling a couple of million dollars of HubSpot stock was financially regrettable but stiffened his backbone when Salesforce approached. A $50 million Series B secondary gets “wobbly,” yet prudent liquidity can align both founder and investor.

8. Canva should treat the public window as perishable

Halligan’s recommendation was unambiguous: “Run, Forrest, run.” IPO demand is strong, valuations are good and timing is “oddly seasonal,” so Canva should line up a transaction. Jason noted the founders’ philanthropic commitments, existing liquidity and profitability make their calculus more complicated than a founder simply needing cash.
Rory’s governing rule was “price clears all markets.” During the prior three years, private investors might offer 5X revenue with less hassle, while public markets offered 10X revenue with more scrutiny. Figma’s offer valuation around 18X may be privately reproducible, but its roughly 80X trading valuation is not; at the margin, public money has become cheaper.
Halligan disputed the mythology that public life is inherently terrifying. HubSpot exchanged “quirky, slightly misaligned venture capital investors” who were deeply “in our shorts” for quirky public investors who were less intrusive. Activist episodes at Twilio, Autodesk and Zendesk are memorable precisely because Halligan considers them relatively rare.
Rory’s hierarchy was typical public investors as the most benign, VCs as a bigger pain, and public activists as worse still. Listing also creates daily liquidity rather than annual liquidity “by appointment only.” Halligan’s non-financial case was stronger: IPO day and the company celebration afterward rank among the two or three most meaningful days of a founder’s life.

9. The old SMB software playbook does not transfer cleanly to AI

Halligan chose SMB partly because selling to CIOs was “soul-crushing work.” He believed the internet disproportionately benefited small businesses relative to large ones, at least when HubSpot started, making success depend more on “the width of your brain than the width of your wallet,” and evaluated the model through CAC and lifetime value rather than a superficially unattractive P&L.
HubSpot, Shopify, Block and monday.com demonstrated that Silicon Valley’s anti-SMB consensus could be wrong. Yet Halligan cautioned founders against copying HubSpot’s inbound marketing, freemium, PLG, culture or under-the-enterprise go-to-market wholesale: “It worked at the time,” and innovation means finding the next lever rather than reenacting the last one.
Jason’s AI challenge is deployment. Enterprise buyers can fund training teams and forward-deployed engineers; prosumer products such as Lovable, Replit and Gamma work with little bespoke setup. Halligan sees less success between those poles, where a restaurant or conventional small company plainly does not have an “AI team.” A panelist cited a HubSpot report saying 80% of the surveyed SMB/VC group had an AI team, but the discussion immediately questioned whether a true small business does.
Jason wants AI that self-trains, compressing a Palantir-grade six-month rollout into 60 seconds. Rory’s more optimistic sequence is that large companies first discover the required features, after which vendors can sell SMBs pre-trained phone answering, order dispatch and other “pre-canned, pre-baked” capabilities: turn it on and “you too can sound like a big co.”

10. Hyperscaler profits can fund AI excess for another year

Harry cited a quarter with roughly 38% adjusted-EPS growth, 22% revenue growth and a 22% free-cash-flow decline. Rory’s interpretation was not that AI already caused the operating success; the existing business is so strong that management can finance its AI ambition for as long as that cash machine continues working.
His deliberately macho summary: “Real men with $70 billion of free cash flow get to spend 40 billion of that on servers.” Meta and Microsoft indicated that spending would continue for another year because they can afford to secure a place in the next platform, even before the new revenue fully arrives.
The mismatch remains enormous: perhaps $25-$30 billion of current AI application revenue against $400-$600 billion of annual infrastructure investment. Rory could imagine the application layer reaching $300-$400 billion in a decade, making the long-term thesis real while still allowing a nearer-term period when capacity outruns demand.
Halligan also highlighted stock-market performance rather than revenue growth: from August 26, 2022, Oracle and SAP had each risen about 230% and Shopify about 300%, outpacing the other SaaS companies he named. Rory noted that some of this reflected their lower starting base, while saying Oracle had so far made itself relevant in cloud by investing free cash flow in GPUs.
Halligan sees “some kind of bubble,” especially where Silicon Valley companies trade revenue among themselves; he prefers ChatGPT reaching ordinary consumers and Harvey or Rogo selling to lawyers and bankers. Rory expects marginal, debt-financed players to burn, the giants to retrench for a year or two, then growth to resume: capitalism “spending money trying shit that works.”

11. Jensen created the new stack while Satya re-founded Microsoft

Halligan’s CEO-of-the-year choice was Jensen Huang, partly because he is rewriting the CEO playbook. Rory distinguished leaders who created new categories—GPUs and models—from Meta and Microsoft leaders brilliantly using existing cash machines to buy relevance in the new world.
Jason nevertheless argued Satya Nadella deserves more credit than Zuckerberg because a non-founder must orchestrate bureaucracy rather than simply command it. Bringing Sam Altman back in, committing Azure to AI and structuring the unusual OpenAI relationship required years of internal persuasion that a controlling founder can largely bypass.
Rory’s praise centered on accepting an uncomfortable constraint: Microsoft could not build the core technology internally. Nadella instead made a roughly $10 billion commitment, which Rory described as producing 49% economics, secured resale access and inserted Microsoft into AI. “He’s like a re-founder,” Halligan concluded—someone who gets things done with founder-like speed despite lacking founder control.

12. Cognition bought acceleration, then imposed a cultural reset

Cognition’s rumored pricing moved from $10 billion to $15 billion, much as Anthropic rumors moved from $100 billion to $170 billion. Harry cited Cognition and Windsurf at roughly $85 million of revenue each, or $170 million combined; Jason later described the acquired revenue as about $80 million. Rory viewed the repricing as a private IPO: scarce, perceived-premium AI assets attract enough demand that price becomes the allocation mechanism.
Jason read the Windsurf purchase as brand and distribution rather than acquihiring. Devin works for some difficult engineering teams but remains a limited deployment beside tools such as Claude Code; acquiring a brand, roughly $80 million of revenue and a broader platform saved perhaps three-to-nine months and looked effectively non-dilutive at the new valuation.
The post-deal terms were severe: 30% reportedly laid off, while roughly 200 others had until August 10 to choose nine months’ severance or an 80-hour week across six in-office days. Rory allowed that this could be an explicit cultural statement; Jason saw proof that the team itself was not the primary asset.
Windsurf’s founders and investors reportedly supplied $100 million after Google did not, followed by an eleventh-hour sale and employee buyouts. Rory’s broader warning was that once a transaction moves away from normal cap-table rights, everybody starts improvising “fairness.” Non-key participants become dependent on “the kindness of strangers,” whose definition of fair can change.

13. Ramp needs balance-sheet capital; venture firms need strategic clarity

Ramp raised $500 million at roughly $22-$22.5 billion after five or six rounds in about 18 months, bringing stated capital raised to $1.9 billion. Rory emphasized that corporate cards are not pure software: somebody must finance the customer during the float period, so faster interchange growth mechanically consumes more capital.
Using the panel’s rough figures—$700-$800 million of revenue, 2%-2.5% interchange and perhaps a 15-day rotating balance—Rory estimated a possible $3-$4 billion funding requirement. That makes a $500 million round less cosmetic than it looks, while its roughly 2% dilution can still be cheaper than other financing or obtaining a banking license.
A high-priced round becomes “suicide” only if the company soon needs another. If sufficient capital lets it grow for four or five years, investors can absorb the consequences of overpaying; trouble begins when a company raises too much or too little at a high valuation, returns six months later worth about $1.5 billion and cannot reconcile the down round.
CRV’s $750 million early-stage fund, smaller team and decision not to raise a late-stage select vehicle represent the opposite discipline. Jason said 90% of his carry will come from his main fund; an opportunity fund might add only 10%-15% for disproportionate work. Rory noted that an early-stage manager can still raise an annex fund later around an outlier investment, so declining a vehicle now is not necessarily permanent.
Rory’s final framework also covered Benchmark: specialists risk losing attention, full-stack firms risk deploying too broadly, and either can win if it knows its game and executes consistently. Specialists face noise and lost at-bats; large platforms face the risk that volume obscures weak returns. The consistent strategy, rather than a particular fund structure, is the essential requirement.

PE Perspective on Insurance Brokers - [Business Breakdowns, EP.225]47. A Conversation with 测测's 任永亮 About His 99-Point Ambition: Tools, AI, and the New Continent of Robotics