May 8, 2025 · 1h 19m · 20vc
Benchmark vs a16z: Why Stage Specific Firms Win · 20VC with Harry Stebbings
0:00 / 0:00
▶ Watch on YouTube →
gold bands on the timeline = statements, start to end. Hover to read, click to jump. CC turns on captions
In this episode of 20VC, Harry Stebbings, Jason Lemkin, and Rory O'Driscoll dissect the defining strategic divide between stage-specific focus funds and multi-stage mega-funds, while debating the macroeconomic limits, labor displacements, and market defensibility of the generative AI boom.
How this conversation actually went
Every chapter scored 0–10 on four independent dynamics. Hover any point for the reasoning behind the score. Harry holds 13.3% of the talking time here. How this is scored →
Harry as informed peer 5.0 Guest teaching 5.4 Guest disagreement 3.6 Harry pushing back 3.3
speaking balance: gold is Harry, purple is the guest (3 minute bins)
Sharpest disagreement ▶ 32:25 Jason's $100k bet on labor displacement
Jason forcefully rejects Rory's moderate economic framing, betting $100k that knowledge workers will be displaced in 12 months and ranting about tech work ethic.
Hardest push from Harry ▶ 1:17:14 Harry challenging Koppelman's relevance thesisHarry explicitly rejects Josh Koppelman's premise that deal activity is the only way to remain relevant, arguing media content builds relevance without deploying capital into bad deals.
Biggest teaching moment ▶ 55:40 Rory's breakdown of option vs intrinsic valueAfter Harry admits he doesn't understand Decagon's valuation, Rory breaks down the fundamental venture dynamic of paying up for option value upside rather than bounded intrinsic value.
Harry holds his own ▶ 9:24 Harry citing Singerman on compounding doublesHarry demonstrates keen portfolio math expertise by citing Brian Singerman's rule on the value of the final double, earning immediate validation from Rory.
the scores for every segment, with the reasoning behind each
| Chapter | Topic | Harry as informed peer | Guest teaching | Guest disagreement | Harry pushing back | Why |
|---|---|---|---|---|---|---|
| Transition and Video Title Sequence | 3 | 2 | 1 | 1 | Harry opens with light banter and prompts the guests on the news of Windsurf's potential acquisition. Jason and Rory offer initial reflections on inflated market valuations without friction. | |
| Cursor's Dilemma and Investor Boardroom Dynamics | 4 | 5 | 3 | 3 | Harry questions whether Cursor shareholders should worry about competing with OpenAI. Rory gently reframes the question, explaining board dynamics and why turning down an offer requires courage. | |
| The Growth Stage Shift: Loom and Quick Exits | 7 | 3 | 2 | 2 | Harry cites Brian Singerman's thesis on the value of the final double in company valuation trajectories. Rory strongly agrees and builds on the point, comparing VC outlier math to private equity. | |
| The Rise of Multi-Stage Mega Funds | 5 | 7 | 4 | 4 | Harry presents a macro thesis that multi-stage mega funds will dominate venture. Rory playfully calls Harry's agenda intellectually incoherent and walks through the multi-stage return math. | |
| The Reality of Inception Investing | 4 | 5 | 2 | 2 | Jason and Rory break down inception investing, contrasting true early-stage discovery with mega funds using seed rounds as loss-leader bundled goods. | |
| Stage Bundling and the Founder's Perspective | 5 | 6 | 3 | 3 | Harry asks if multi-stage firms can succeed without pre-seed. Rory notes that founders care about fast money rather than clean stage execution. | |
| Portfolio Concentration vs. Scale and the Hit Rate Reality | 5 | 8 | 7 | 6 | Harry attempts to push back claiming founders want low portfolio deal counts. Rory calls bullshit on Harry's premise and cites Kurt Rotman's hit rate data comparing Benchmark to Andreessen Horowitz. | |
| The Mathematics of Venture Outliers | 5 | 6 | 3 | 3 | Rory and Jason examine historical hit rates and market share of venture outcomes, while Harry suggests future trillion-dollar startups will sustain larger fund sizes. | |
| Macro Economic Limits and the Value of AI | 5 | 6 | 3 | 2 | Jason questions global GDP capacity for trillion-dollar exits. Rory grounds the discussion in US GDP and stock market capitalization metrics. | |
| The Great AI Labor Displacement Debate | 3 | 4 | 8 | 2 | Jason forcefully argues that half of knowledge workers will be displaced by AI in 24 months, offering a $100k bet and criticizing tech work ethics. Rory remains skeptical of rapid macro shifts. | |
| Is AI a Step-Function Shift or Historical Evolution? | 6 | 5 | 4 | 6 | Harry pushes back against Rory's comparison of AI to the PC and internet eras, pointing out that AI deployment lacks heavy physical installation barriers. | |
| SaaS Hiring Freezes and Baumol's Cost Disease | 5 | 6 | 4 | 2 | Rory references Baumol's cost disease to explain uneven productivity gains. Jason highlights tech hiring freezes and shifting labor dynamics. | |
| The LP Landscape and the Threat to University Endowments | 5 | 5 | 3 | 3 | Harry asks about political proposals to strip university endowments of tax exemption. Rory lightheartedly teases Harry's LP focus before outlining the importance of research universities. | |
| The Unicorn Washout and Downturn Mergers | 5 | 5 | 2 | 2 | Harry highlights Census's fire sale to Fivetran. Rory explains that out of 1,000 unicorns, most will undergo recapitalizations or discounted acquisitions. | |
| Analyzing Decagon's Valuation and Customer Support ROI | 5 | 6 | 2 | 3 | Harry questions Decagon's 100x ARR valuation. Rory defends the underlying market opportunity by citing high resolution rates in AI customer support. | |
| Option Value vs. Intrinsic Value in Venture Capital | 4 | 8 | 3 | 5 | Harry admits confusion regarding high valuations facing stiff competition. Rory delivers a lesson on why venture investors trade bounded intrinsic value for unbounded option value. | |
| The Durability and Defensibility of AI Startup Revenue | 5 | 5 | 4 | 3 | Jason argues that AI tools face 5-week disruption cycles and lack durable moats. Rory contends that enterprise integrations generate lasting stickiness. | |
| Vertical AI Moats and the Tech Productivity Explosion | 5 | 6 | 6 | 4 | Harry argues for vertical AI moats like patent law. Jason attacks VC hubris around startup durability, citing HubSpot's massive AI code velocity as a threat to incumbents. | |
| SaaS Exits, Fund Math, and Maintaining Venture Relevance | 8 | 5 | 4 | 7 | Harry pushes back on Koppelman's assertion that deal volume is required for relevance, arguing media content achieves relevance without wasting capital. Rory agrees with the strategy. |
Statements from this episode (45)
O'Driscoll: OpenAI's $3B Windsurf acquisition makes strategic sense at 1% of valuation
“It's one percent of the market cap to play In one of the largest use cases for AI and one that, you know, the core constituency of developers loves. It just makes total sense for open AI.”
O'Driscoll: Number two market players should accept acquisition offers under squeeze
“Eventually, when you're the number two and you get the heavy squeeze from the adjacent acquirer, it often makes sense to fold because, you know, because otherwise they might buy the number one and then you're done.”
O'Driscoll: Board members evaluate M&A offers based on personal fund returns
“Everyone comes to these M&A discussions around the boardroom table, mentally running their eternal cap table, and saying, what does it mean for me?”
Lemkin: Andreessen Horowitz received a 1x return on its Loom investment
“When Loom sold and Andreessen invested at one And they got their one ex back in a year.”
O'Driscoll: Thrive Capital doubled its Instagram investment in just days
“Thrive invested in Instagram, and like, literally four or five days later, it sold for a two X, right?”
O'Driscoll: Private equity structurally beats venture capital except for massive outliers
“Almost everything about PE is better in terms of making money than what we do. And the only thing we have in our favor is that every once in a while, maybe once every five years, you find yourself with a 10% ownership in a thing that's already worth billions o…”
O'Driscoll: Doubling late-stage winners beats grinding early ARR growth
“That double from there. So, so much easier than grunting it out from a million to five million in ARR. Jason, because you did it as a CEO to ten million in ARR. Then the shitty year where you only go to 16. Then you re-accelerate a little. Oh my God. And now y…”
Stebbings: Multi-stage mega-funds will dominate venture capital over next decade
“I think multi-stage win The next 10 years.”
O'Driscoll: Multi-stage mega-funds control 50% to 60% of venture capital
“They have between 50 and 60% of the capital, provided they don't under index The rest of the industry, they're going to have 60% of the wins.”
O'Driscoll: Mega-fund capital will make mid-tier venture investing difficult
“So my operating assumption as a mid-tier firm in terms of size is for the next three to five years, there's great big walls of capital that will make a lot of investing very difficult.”
Lemkin: Refuses meetings with seed-stage founders seeking $100M valuations
“Whenever I get an email from a founder that has what appears to be those metrics, right? That on their seed, they're going to get ten million at a hundred, right? 1015. I just email them back. I'm like, I can't compete. I don't take the meeting.”
Stebbings: Lost a $10M on $50M seed deal to mega-funds
“I actually lost an inception investing check this week to the two big ones, 10 on 50 for some good people out of a good company.”
O'Driscoll: Mega-funds treat Seed and Series A deals as loss leaders
“And the seed and even the series A is a loss leading product. It's like milk at the grocery store. Come on in, buy your cheap milk, but we're going to upsell you all the strawberries you can buy, baby, right? You know, wait till you see the series C and D, rig…”
O'Driscoll: Founders care about capital and help, not VC stage strategy
“To the founder, they don't give a damn about your nuanced stage specific strategy. A founder wants two things from his venture investor. He wants money, money, lots of it with the minimum amount of hassle and perhaps the maximum amount of help, right?”
O'Driscoll: Multi-stage VC return performance will take 5-7 years to evaluate
“I don't think this bundling thing is not going to stop because the founders don't like it. They're going to love it. It's only going to stop if the returns from it are subpar relative to the returns of people who are more specialized and even, and I think that…”
O'Driscoll: Averaged two venture deals per year over 30-year career
“I do two deals a year. Shockingly, I've been in this business 30 years and I've done 60 deals.”
O'Driscoll: Benchmark achieved 10% hit rate versus a16z's 2%
“Benchmark did something like 63 series A's and had a 10% hit rate across 15 years. Andreessen did a 454 series A's and had a two percent hit rate across the same period of time. Now in absolute numbers, they had 10 five billion hits and Benchmark had six.”
O'Driscoll: Focus funds win on hit rate; mega-funds win on absolute hits
“The focus fund is better at hit rate, has more as a percentage, and lower as an absolute number than the guys cranking through 454 A's.”
O'Driscoll: Historically, a16z captured 10% of top VC outcomes, Benchmark 6%
“In a less competitive time, the best firm in terms of percentage of total outcomes was Andreessen. They got 10% of the good outcomes. Benchmark got six, and everyone else bunches in the two, three, four, five percent rate, right? In other words, when there was…”
O'Driscoll: Six or seven $8B venture funds cannot all succeed
“It's going to be really hard. It might be possible for one eight billion dollar fund to be freaking amazing. It's going to be very damn hard for six or seven eight billion dollar funds to be freaking amazing at the early states level to the same, to the extent…”
O'Driscoll: SpaceX and OpenAI are the only $100B+ private compounders
“It all boils, the deciding between does this strategy work or not, is other companies that compound from a hundred billion to four hundred billion in the private markets before they go public. And right now there's two. There's SpaceX, There's open AI.”
O'Driscoll: Roughly $1 trillion in new value is created per decade
“Somewhere every decade, there's roughly, I mean, it boils down to this, there's roughly a trillion dollars per decade of new value created plus or minus, right?”
O'Driscoll: Five of six trillion-dollar companies were VC-backed
“There are six companies with a trillion dollar market cap. And I think all of them, but Berkshire were founded By VCs and founded within my lifetime, at least.”
Lemkin: AI will replace half of knowledge workers in 24 months
“Half of these knowledge workers are going to be gone in 24 months”
O'Driscoll: AI will not cause mass unemployment within 12 to 24 months
“I don't buy the mass unemployment in 12, 24 months.”
Lemkin: SaaStr eliminated five jobs in 90 days due to AI
“Literally, we, even at Sastra, we've gotten rid of five people in our team in the last 90 days due to AI.”
O'Driscoll: Klarna's AI adoption will not deliver step-function profit growth
“Take Klarna as the public example. I mean, they talked boldly and they probably did save those people, right? But I think in the end you'll discover it'll have roughly the same profits as other lenders. It'll be marginally more efficient. It won't be a step fu…”
O'Driscoll: GDP and productivity growth will remain at 2% for 20 years
“I think GDP growth and productivity growth has been roughly two percent since the dawn of the industrial revolution. And I'm willing to lean into the fact that it'll be two percent for the next 20 years.”
Lemkin: Hands-off tech middle managers will be eliminated within a year
“That in our industry, it's going to lead to massive human disruption. Just as many founders, just as many VCs, but man, if you're a hands-off keyboard middle manager, you're going to be gone in a year.”
Lemkin: Virtually all growth-scale tech CEOs currently have hiring freezes
“Every CI CEO I know that's at growth scale has that isn't windsurf has some version of a hiring freeze going on, right? Even no matter unless the growth is insane and it's all you can hire, but you got to get rid of somebody, right? It's all AI first. And so i…”
O'Driscoll: Marginal return on college over the last decade is profoundly negative
“The marginal return on the marginal entrance to college in the last 10 years is profoundly negative, right? You've got the set of skills that don't have a market value and you owe 150 grand.”
O'Driscoll: University endowments fund $150M early-stage VCs, not $1B mega-funds
“Those are the LP of choice for the small early innovative funds. It's not a great trend because, you know, if you're raising a billion, you've long since stopped talking to Harvard in a meaningful way. You're actually talking to pick your sovereign wealth fund…”
O'Driscoll: Only 200 of 1,000 current unicorns will IPO
“Honestly, I didn't even notice because it's going to be one of five, 600 of these that's going to have to happen. You know, there's somewhere between 800 and a thousand unicorns and a couple of hundred of them are going to get public and the rest of them are g…”
O'Driscoll: Top AI use cases are personal chat, coding, and customer service
“If you look at the top two or three use cases of AI, the number one is just personal chat, the number two is coding, and the number three is customer service.”
O'Driscoll: Generative AI doubles customer service automated resolution rates
“In the sense that, you know, we talked to people who said, and I had an investment in this space pre-Gen AI, right, where the resolution rate was roughly 30, 35%. In other words, one in three calls got solved. We did a bunch of references around this space and…”
Stebbings: Intercom is valued at $2B after 17 years
“Intercom is, You know, bluntly, an amazing story to not a huge enterprise value today for 17 years. I know two billions a lot.”
O'Driscoll: Intercom is the best incumbent adapting to AI
“Intercom is by far the best of those. They've done an amazing job of adding AI, their little fin charging. I mean, if there's anything that you could, if there's anything that a pre-gen AI company could do to get relevant in-gen AI for customer success, I thin…”
O'Driscoll: VCs prioritize high option value over intrinsic value
“Venture guys love new shit with option value over all shit with intrinsic value. It's as simple as that. We don't do intrinsic value. You know why? There's no upside in intrinsic value. We are junkie, upside junkies, right?”
O'Driscoll: Scale Venture Partners debated and passed on Decagon's $1.5B round
“And in a way, and you know, look, we had some internal discussions. Should we be one of the hundred people pleading to put money into Decagon at one and a half? And we had some interesting discussions.”
Lemkin: AI startup revenue is less durable than traditional SaaS revenue
“What I worry is, is just that when there's so much competition, I think everyone's going to be less durable. It's not your ten-year-old SaaS company that is seeing less durability. I think everyone, I think this is new, this less durable revenue”
Lemkin: Disruption in AI software is occurring every five weeks
“And then these spaces, it's not once a decade disruption or once every five years. Now it's once every five weeks, literally every five week disruption, right? So this lack of stability is where I think that the DecaCon revenue growth justifies a hundred X. It…”
O'Driscoll: AI markets will consolidate into 3-4 dominant players
“I think that what a successful set of products at AI starts to gel over the next couple of years, and the people who are in the lead at that point in time get a similar ten-year run to, you know, the ten-year runs you and I both benefited from in SaaS, Jason, …”
Lemkin: HubSpot engineers using Cursor are pushing out too many features
“Literally I talked with Yamini wrong on from HubSpot last week. Okay. This is HubSpot. She said now at HubSpot there with cursor, they are pushing out so many features. They can't put them into production anymore. She wasn't kidding. They're there. It said the…”
Stebbings: DoorDash bought Deliveroo for $2.9B at a $1.4B premium
“Delivery is also getting bought by door dash for 2.9 billion. Just to put that in context for you guys, and I don't mean that rudely, but I'm sure you're not aware of public markets in the UK. It was valued at between 1.4 and 1.5 billion. That is a 1.4 billion…”