May 8, 2025, 10:23 AM ETFireside Chat, Interview

Benchmark vs a16z: Why Stage Specific Firms Win

Benchmark executed ~63 Series A rounds over 15 years, achieving a 10% "hit rate," whereas Andreessen Horowitz executed 454 rounds with a 2% hit rate.
In absolute terms, Andreessen achieved $5B in hits (10 deals) compared to Benchmark's $6B (6 deals), illustrating the trade-off between volume/frequency and hit-rate precision.
A key dilemma identified is that mega-funds sacrifice hit-rate percentage for aggregate absolute returns, while focused funds achieve higher precision but lower absolute volume.
Josh Koppelman's "Venture Arrogance Score" analysis suggests that for mega-funds ($5B–$10B) to generate market-beating returns, they would need to capture market shares of deals historically unachievable even by top-performing focused funds.
With 5–7 mega-funds controlling 50–60% of available capital, mid-tier firms face an "option value" squeeze where mega-funds can price out Series A and B rounds as loss leaders to secure the full bundling of deals from seed to late stage.

Windsurf was acquired by OpenAI for $3 billion, described as a strategic move capturing a core developer use case for 1% of OpenAI's market cap.
Cursor, a rival AI coding assistant, recently raised at a $10 billion pre-money valuation; investors are advised to assess the risk of competing against an incumbent that can acquire the "number two" player.
Census, a previously hot data company, was acquired by FiveTran for a "nickel" price (implied significant loss relative to previous hype), reflecting a trend of high-flier seed-stage deals being acquired at distressed valuations.
Lacework, previously a leading security unicorn, was reportedly acquired for a fraction of its peak valuation after investors concluded cashing out 50–70 cents on the dollar was superior to holding for a potential IPO.
Deliveroo was acquired by DoorDash for $2.9 billion, a figure significantly higher than UK public market valuations, demonstrating US acquirers' willingness to pay premiums to absorb geographic competitors.
Olo, a public restaurant tech company, is reportedly up for sale, signaling that even mature enterprise software companies may be forced to exit if growth stalls or competitive moats erode.

Harry Stubbings predicts 50% of knowledge workers (including SaaS reps, QA engineers, and junior product managers) will be displaced by AI within 24 months, citing current hiring freezes and immediate efficiency gains at his own firm, Windsurf.
Rory argues that while AI will increase efficiency in specific tech sectors (like customer support resolution rates jumping from 35% to 70%), global GDP growth will likely remain at the historical ~2% rate, similar to the PC and Internet revolutions.
A critical distinction is drawn between "killing" a company and "maiming" it: AI may prevent IPOs by eroding margins and causing churn rather than causing total failure.
Venture firms are facing a labor paradox where high-caliber workers are unwilling to accept traditional roles, leading to a situation where AI is adopted not just for cost savings but to fill a void in human labor availability.
The "early adopter" phase of AI is larger than previous technological shifts because the tech sector itself constitutes a larger percentage of the US economy compared to the manufacturing-heavy economies of the 1980s/90s.

Investors are shifting from "intrinsic value" (stable, growing assets like Intercom at $400M ARR) to "option value" (high-growth, unproven assets like Decagon at $15M ARR), as the former offers bounded returns while the latter offers the potential for 20x multiples.
Decagon raised at a $1.5 billion valuation (100x ARR) based on the AI customer service use case, which offers the clearest ROI by replacing human support with automated resolution.
The market is experiencing a "bundling good" phenomenon where investors are forced to participate in seed rounds to access Series A/B opportunities, as mega-funds compete on price and speed rather than exclusive deal flow.
Founders prioritize "money with minimum hassle" over nuanced stage-specific strategies, making them willing to accept deals from firms that may be weaker in early stages if those firms can guarantee future capital.
The consensus is that the "game" of venture capital has shifted to a volume-based model for the next 5 years, where firms must compete against massive capital walls regardless of long-term return sustainability.

The potential removal of tax-exempt status from major endowments like Harvard could reduce capital availability for early-stage, innovative funds that rely heavily on university endowments as LPs.
There is a significant risk that the current "stability" of the 2010–2024 SaaS era is gone; enterprise AI markets may face higher churn and pricing pressure, preventing the formation of long-term oligopolies.
If AI does not generate 5–10 new trillion-dollar private companies, the current mega-fund model (assuming high market share capture) will fail to meet hurdle rates for returns.
Vinod Khosla's prediction that AI will eventually render half of current knowledge worker jobs obsolete is gaining traction among VCs, shifting the focus from "efficiency" to "replacement."
The current market environment forces firms to internalize that they are likely to be "killed or maimed" rather than survive unchanged, necessitating a shift toward "AI-first" workflows even at mature companies like HubSpot.