Fireside Chat, Conference Presentation
Why VC Today is Worse than 2021
- Everett Randall joins Benchmark as a General Partner, having previously worked at Kleiner Perkins and Founders Fund, replacing an departing partner without disrupting the firm's operations or portfolio momentum.
- Benchmark secured Randall's recruitment through an "equal partner" pitch and a highly attractive carry pool, including participation in funds backed by major portfolio companies like Fireworks, McCaw, Lagora, and Manus.
- Rory Fenton and Harry Stebbings debate VC compensation versus tech talent, concluding that while venture carries offer long-term wealth over 20-30 years, top AI engineers at firms like Meta can secure $1B+ in fully liquid restricted stock over four years, creating a more immediate and liquid wealth event.
- Revolut raised $3B at a $75B valuation (up from $45B in 2024), continuing the trend of private capital outperforming public markets for late-stage growth companies, with Revolut generating ~$3B revenue, $1B profit, and 60% growth.
- Jason Calacanis expresses deep concern over TAM (Total Addressable Market) exhaustion in AI verticals, arguing that the current euphoria is leading to over-valuation of niches that cannot sustain the velocity required to reach $1B+ ARR exits.
- Harry Stebbings counters that the best founders expand their TAM sequentially from point solutions to broader adjacent markets, citing Spotify, Deel, and Revolut as examples where founders successfully grew their addressable market over time.
- Rory Fenton challenges the "Great Man Theory," arguing that while founder skill determines the winner, the market size must be inherently large at inception; he attributes Spotify's success to specific geopolitical licensing advantages in Europe rather than just founder execution.
- A key thesis shift is identified regarding AI adoption: CIOs and enterprises are currently "in market" due to exogenous pressure to adopt AI, mirroring the 2020 pandemic surge, which risks over-extrapolating growth rates if adoption slows once the initial wave of purchasing is complete.
- Harry Stebbings warns that confusing a temporary "everyone is in market" phase with a permanent structural shift could lead to catastrophic valuation errors if growth rates revert from 5-10x to 2-4x after the initial saturation period.
- Poolside AI announced plans to build its own 2-gigawatt AI data center, signaling that competing at the software model layer now requires owning massive physical infrastructure, a decision driven by the inability to procure sufficient compute capacity from hyperscalers like CoreWeave.
- The industry is pivoting toward extreme capital intensity, with smart founders recognizing that building proprietary data centers is now a necessity to secure GPU supply, raising the cash flow break-even point for AI startups from hundreds of millions to billions of dollars.
- OpenAI's shifting infrastructure strategy involves spending potentially more with Oracle than Microsoft, effectively offloading the balance sheet risk of capital-intensive data center build-outs to Oracle while retaining the upside of AI model development.
- Rory Fenton argues that Microsoft made an economically rational decision to step back from OpenAI's capital needs, allowing Oracle to take on the risk of high leverage (4.6x debt-to-equity) in exchange for potential market access, whereas Microsoft avoids "irrationally" funding a subsidiary that requires infinite capital.
- A debate on venture capital strategy highlights the "temporal diversification" problem, where funds raising every 18-24 months fail to spread risk across market cycles, leaving them exposed to the inevitable correction when the "most aggressive" players face the "crash."
- Harry Stebbings suggests that the optimal strategy is to be aggressive enough to capture boom returns but calibrated to survive the downturn, noting that current market aggression often exceeds the risk tolerance required for long-term survival.
- Jason Calacanis argues that early-stage investing has never been easier for "checking" (meeting entrepreneurs) but is becoming harder due to high entry valuations, low ownership percentages, and compressed margins, despite the presence of high-quality founders and technology.
- The conversation touches on the risks of AI content moderation, specifically the potential for generative AI to facilitate erotica or harmful content, with Harry noting that the "hard seat" of content moderation will be the primary operational challenge for LLM providers in the next five years.
- A "Agree or Disagree" segment features a bet on whether Lovable (a coding agent startup) will hit $1B ARR by the end of next year, with Harry disagreeing due to concerns over market size and churn, while Jason and Rory express skepticism about the timeline for such rapid scale in the current economic environment.
- Harry Stebbings identifies the "vibe coding" phenomenon (using AI to generate code instantly) as a disruptive force that invalidates traditional early-stage due diligence metrics, as product quality can now be achieved at the seed stage by non-technical founders, making differentiation harder.
- In a comparison of Rippling and Deel, Jason selects Rippling due to the "sticky" nature of domestic US payroll and its established installed base, while Harry leans toward Deel, citing the larger, more fragmented global payroll market and the "Wild West" opportunity of international employment laws.
- Both investors acknowledge the growing absurdity of the venture landscape, where funds simultaneously invest in companies at $1M ARR and $1B ARR, blurring the lines between early-stage VC and late-stage private equity/public-style investing.
- The overarching trend identified is a consolidation of capital toward the "anointed winners" (OpenAI, Anthropic, Revolut, Deel) and infrastructure plays, with less confidence in funding middle-tier vertical AI tools that lack clear paths to massive scale.