Sarah Guo: On Her New $101M Fund; How AI Impacts Inequality; AI Startups vs Incumbents | E1007

Detailed Summary

Introduction (0:00 - 0:55)

The discussion opens with Harry Stebbings welcoming Sarah Guo, noting their first interview was in 2017. Sarah reflects on the early timing of their previous conversation about conversational AI and introduces her new venture, Conviction Capital, a $100M fund.

Why Sarah Left Greylock (0:55 - 3:02)

• Sarah Guo spent 10 years at Greylock, a firm she deeply respects, but felt a strong desire to focus on early-stage, zero-to-one investing.
• She wanted to operate differently, with a small team, and change the founder experience, driven by her belief that AI represents a "breaking change."
• Transitioning from a GP to a fund manager and founder involved more existential dread and administrative tasks, such as running payroll and managing an office.
• She highlights that a fund is a business with two sets of customers: entrepreneurs and LPs, requiring strategic thinking and recruiting efforts.

Conviction’s Thesis (3:02 - 3:51)

• Conviction Capital's core thesis is that AI is the biggest value creation opportunity of our lifetimes.
• Sarah is confident that small teams (10-20 people) will build billion-dollar businesses using AI.
• The fund aims to be central to the AI community, deeply understand the technology, and become experts at identifying and partnering with founders building these companies.

AI & Wealth Inequality (3:51 - 5:08)

• Sarah shares concerns about AI potentially exacerbating wealth inequality, agreeing with Harry's worry about the centralization of wealth.
• She acknowledges that technology drives abundance, and while AI offers incredible productivity benefits, society must address the unequal distribution of this abundance.
• Despite the distribution challenges, she believes progress should not be halted, as people rarely want to give up new technologies once adopted.

AI Regulation (5:08 - 6:43)

• Sarah expresses concern about the significant gap between regulatory bodies' knowledge and the actual technology of AI.
• She notes that this isn't structurally different from other tech areas (e.g., cybersecurity) but highlights the unprecedented speed of AI's evolution.
• It is incumbent upon those developing and enabling AI to partner with policymakers for education and to build new regulatory muscles.

Investing in AI (6:43 - 23:00)

• Sarah explains why Conviction is a vertically-centric AI fund, viewing AI as a horizontal technology that will eventually permeate all software.
• Specialization allows Conviction to invest in the AI community, act as a matchmaker for teams, and provide strategic relationships for model access, data, GPUs, and design partners.
• She emphasizes the need for investors to understand the technical and dynamic nature of AI, as well as strategic issues like safety and alignment.
• Sarah acknowledges that many VCs lack deep understanding of AI, but different approaches to value creation exist, including focusing on founder quality.
• She addresses the "craziness" in AI funding, noting that media often amplifies a small number of large rounds, while most founders face more constrained realities.
• Sarah clarifies that most AI companies do not need to raise massive sums to train models from scratch; instead, they focus on applying, fine-tuning, or building on existing models.
• Conviction Capital's $100M fund size is a deliberate choice to foster discipline and focus on early-stage (seed and Series A) investments, typically $1M-$10M.
• She believes the bulk of the opportunity lies in capital-efficient companies and that larger funds often face pressure to grow, which doesn't necessarily generate better returns.
• Conviction avoids investments requiring $100M-$200M upfront, as they are unlikely to impact a fund of its size.
• Sarah emphasizes that brand in VC is built on returns, important companies, and reputation with founders, not just participating in every hot deal.
• She identifies low-hanging fruit for AI applications, such as legal tech (Harvey.ai), code generation beyond simple autocomplete, RPA, tooling for LLM retrieval feedback, and multimodal models for creative and marketing use cases.
• Ideal AI founders often combine deep technical knowledge (e.g., AI research) with strong domain expertise (e.g., law), though over time, product-oriented software engineers will increasingly leverage AI tools.
• Sarah predicts that in five years, AI will move beyond code autocomplete to more iterative, end-to-end code generation, allowing users to communicate preferences more naturally to AI systems.
• She discusses the startup vs. incumbent dynamic, stating that startups' primary advantage is speed, which is more critical than ever in the rapidly evolving AI landscape.
• While data moats are often discussed, entrepreneurs are creative in collecting and generating data, making founder quality and execution paramount.
• Microsoft is highlighted as the incumbent best navigating the AI transition due to Satya Nadella and Kevin Scott's championing of the technology and strategic bets like OpenAI.
• Amazon and Apple are seen as needing to make more investments in AI, lacking the cutting-edge labs of other tech giants.

Idea Generation (26:11 - 28:45)

• Sarah advises founders against generically searching for ideas, as it leads to generic outcomes.
• She advocates for "high-resolution customer conversations" to identify real problems, either by solving personal problems or actively seeking out customer pain points.
• Open research questions attached to interesting markets (e.g., generating usable 3D models) also present significant opportunities.
• Founders should spend more time picking ideas and gaining deep understanding and confidence before embarking on a venture.
• True progress comes from direct customer contact, understanding their problems, and assessing the real importance of those problems, rather than just receiving polite interest.

Founder vs Market (28:45 - 30:58)

• Sarah identifies as a "founder-first investor" but acknowledges the common VC adage: "great founder meets bad market, market wins."
• Her nuance is to assess if founders can navigate to a better market or possess unique insights that make the market more attractive than it appears.
• She believes founders can sometimes change a market, especially if they recognize industry problems and have a unique approach to breaking through structural challenges like distribution or low margins.

Sarah’s Biggest Investing Mistake (30:58 - 32:49)

• Sarah's biggest regret is not making investments in companies like Benchling and Rippling, which were led by exceptional founders (Sachi and Parker Conrad).
• Her mistake was underestimating the ability of a founder to change a market or overcome perceived risks, such as Parker Conrad's controversial exit from Zenefits.
• This experience strengthened her orientation towards taking risks for truly special founders, even when the market or circumstances seem challenging.

Startup Defensibility (32:49 - 33:55)

• Sarah believes that "defensibility doesn't exist" at the seed stage, as startups begin with nothing.
• She argues that investors are wrong to look for it, as they are investing in a team's trajectory and ability to navigate a market and thesis.
• At the early stage, the market's future is often unknowable, and seeking defensibility in a nascent company is a mistake.

Multi-Stage Firms vs Boutique Firms (33:55 - 37:42)

• Sarah discusses the bifurcation of venture capital into large multi-stage firms and smaller, boutique firms.
• She explains that firm structure and incentives determine strategy; large funds with big fees incentivize hiring and broad coverage, but it's hard for a $5B fund to have "skin in the game" on a $5M investment.
• She advises founders to educate themselves on how different firms operate, their incentives, and to make decisions based on the type of help they need.
• While large firms offer advantages like coverage and reach, returns are often dominated by a few partners, and large groups can suffer from groupthink, seniority overriding decisions, and internal politics.
• Sarah prefers the simpler, more focused approach of smaller firms, which she believes is easier for founders to understand.
• She notes that many founders fail to do investor references, which is a crucial step in understanding a firm's dynamics.

The Death of the Generalist VC (37:42 - 41:11)

• Sarah addresses Hunter Walk's argument that the generalist seed VC is dying due to the increasing size of the tech market, the scale of networks, and the importance of technical innovation.
• Conviction Capital leans into this by being highly specialized in AI.
• However, she acknowledges that there are multiple ways to be a successful investor, citing examples of specialized investors (Eric Vishria in infrastructure) and generalists (Jim Goetz, Elad Gil) who excel.
• She remains skeptical of conclusive statements about VC strategy, as it's a dynamic market.
• Sarah shares lessons from her podcast with Elad Gil, "No Priors," highlighting that media is a business and that high-quality content requires significant effort, not low effort.

Will the economy improve by the end of 2023? (41:11 - 42:04)

• Sarah believes the macro economy will still be "ugly" by the end of 2023, with most of the pain yet to come.
• She anticipates a "digestion period" for tech startups that overcapitalized during the "fat startup" era, as many lack the efficiency to build durable businesses.

AI Software vs Services (42:04 - 44:10)

• Sarah identifies a trend that others might be missing: a significant part of the AI opportunity lies in services, not just software markets.
• Traditionally, software investors focus on the tech stack (chips, infra, tools, apps, consumer).
• However, AI can do work that is currently labor-intensive, opening up massive productivity opportunities, even if it leads to labor displacement.
• She uses the legal profession as an example: currently a services market, AI can automate low-level legal work, creating a larger pie than just software sold to law firms.

Buy or Sell: Multi-Stage & Boutique Firms (44:10 - 46:15)

• Harry asks which multi-stage and seed firms Sarah would "buy" (invest in).
• Sarah strategically chooses Sequoia as the multi-stage firm, citing its execution and culture.
• For a seed fund, she humorously states she would "buy my fund" (Conviction Capital), emphasizing her focus.
• When pressed to "short" (bet against) firms, she predicts a hard time for sub-scale seed funds without differentiated strategies, as LPs become more careful.
• For multi-stage firms, she suggests that early-stage and growth investing are more different than they appear, and firms that tried a "blanket-based approach" to early-stage investing (like Tiger Global) might not fare well.

Dangers of AI Code Generation (46:15 - 46:56)

• Sarah expresses concern about near-term abuses of AI, specifically malicious code generation.
• She warns that nation-states and hackers will use AI tooling to write code, and doing so at scale is dangerous.
• She emphasizes that this is an immediate issue that needs addressing and investment in defenses, distinct from long-term AGI safety concerns.

Sarah’s New Mindset for Investing (46:56 - 48:14)

• Sarah's mindset has evolved to be more comfortable with the unknowable nature of market outcomes at the very beginning of a company's journey.
• She believes that market structures are determined by actors with agency, not just static advantages.
• She is now more comfortable without knowing exactly how things will play out, focusing instead on the quality and agency of the founders.

Reserve Management (48:14 - 49:05)

• Sarah argues that the concept of "reserves" in venture funds is flawed because it assumes investors can accurately predict winners within an 18-month timeframe.
• She points out that many winners are slow burners, not overnight successes.
• Conviction Capital, as an early-stage fund, explicitly chooses to invest in early rounds and leave money on the table in later rounds, focusing on initial alignment with entrepreneurs rather than constantly underwriting their own portfolio.

The World of LPs (49:05 - 50:42)

• Sarah expresses gratitude for her LP base, which consists of long-term oriented individuals and founder/CEO friends.
• She finds the LP landscape to be "clubby" and notes that many LPs lack individual conviction, often following bigger brands.
• She believes more independent thinking among LPs would benefit the ecosystem.
• She also criticizes the practice of GP commits, where GPs often fund their commitment through fees, which she sees as a box-ticking exercise rather than genuine alignment.

Will Trump win in 2024? (50:42 - 52:03)

• Sarah believes it's as likely for Trump to be in prison as to win the election, noting his current criminal sentences.
• She acknowledges that she might have been insufficiently pessimistic about such outcomes in the past.

What does success look like for Conviction? (52:03 - 54:28)

• Sarah outlines four key measures of success for Conviction Capital over the next 20 years:
  1. Returns: Achieving "Best in Class Venture multiples" on a multiple basis, not absolute dollars.
  2. Relevance: Being part of very important companies and picking well.
  3. Partnership: Building a small, collaborative partnership that makes better decisions and has better access as a team.
  4. Beloved by Entrepreneurs: Being cherished by a set of the most important entrepreneurs of the next generation.
• She also mentions a broader mission: to nudge the use of AI in the world to be productive, aligned, and helpful.
• The interview concludes with Harry expressing gratitude and hope for another interview sooner than six years.

Notable Quotes

"AI is the biggest value creation opportunity in our lifetimes." — Sarah Guo

"I'm really worried about wealth inequality... I agree with you." — Sarah Guo

"I think the thing that is different today may be the speed of change... I don't think we have decades to adjust to these capabilities." — Sarah Guo

"If you can do code generation, you can do malicious code generation." — Sarah Guo

"Defensibility doesn't exist [at the seed stage]." — Sarah Guo

"I am increasingly convinced that like it's not knowable what the outcomes are from com for companies like at the very beginning." — Sarah Guo

"A significant part of the opportunity for AI is services not software markets." — Sarah Guo

"I think it will be a hard time for sub-scale seed stage funds without a differentiated strategy to persist." — Sarah Guo

"Returns in most firms are dominated by a few good in Masters." — Sarah Guo

"More independent thinking would be good... for the ecosystem." — Sarah Guo