Sarah Guo - Funding the Frontier - [Invest Like the Best, EP.489]

Episode Summary

Executive Summary: The conversation centers on how Conviction’s investor thinks about AI investing amid a frenetic technological and geopolitical moment. Themes include backing extraordinary people, identifying real-world application fit, compute and energy constraints, open-source proliferation, robotics and biology as emerging winners, and a hands-on, conviction-driven investment process built around first principles, deep learning, and close founder support.

Main Topics: Investing in a high-velocity AI era (Priority: 5/5): The guest describes the current period as unusually chaotic and hard to backtest, forcing investors to make judgments without historical anchors while trying to avoid missing major technology shifts. People, judgment, and the 'great person' theory (Priority: 5/5): A central belief is that exceptional founders and researchers can change outcomes when given the right capital and environment; the firm aims to be close to a small number of frontier people and support them well. How Conviction makes decisions (Priority: 5/5): The guest explains an instinct-led but structured process: initial strong conviction about people/ideas, then deep diligence, memo writing, second reads, and pressure-testing weak spots before committing. Compute, energy, and infrastructure as bottlenecks (Priority: 5/5): The discussion argues that compute supply, power, data-center buildout, and regulatory alignment are major constraints on AI progress and U.S. competitiveness, not a lack of technical capability. Open source AI and diffusion of capability (Priority: 4/5): The guest believes open-source frontier models are already broadly diffusing capability and that attempting to suppress them would mainly slow law-abiding U.S. users; instead, rigorous safety testing is needed. Robotics and biology as real AI application frontiers (Priority: 4/5): Examples such as Sunday Robotics and Chai Discovery illustrate the belief that models can create large value in embodied labor and biology, with practical constraints and customer adoption driving investment theses. Firm-building, LPs, and culture (Priority: 3/5): The guest reflects on fundraising, early LP trust, firm culture, and the importance of being honest about uncertainty rather than forcing a polished narrative before reality is tested.

Key Arguments: AI investing today requires judgment under uncertainty because the environment is moving too quickly to rely on backtests or historical analogies. Exceptional founders and researchers matter enormously; a small set of high-agency people can meaningfully alter outcomes if supported with capital and networks. Conviction’s edge comes from being close to frontier builders, understanding the technology deeply, and focusing on a relatively small number of the most interesting people. The firm’s decision-making starts with strong instinct on people and ideas, then uses memos, peer feedback, and domain research to uncover blind spots. Compute, energy, and supply-chain constraints are the major physical bottlenecks to AI scaling, and these are as much regulatory/political issues as technical ones. Open-source AI models will continue to proliferate; trying to block them domestically would mostly handicap compliant users while doing little to stop adversaries. There is meaningful venture opportunity in robotics, biotech, semiconductors, data-center infrastructure, and other capital-intensive markets if AI changes the underlying economics. In biology, empirical evidence has shifted the guest toward believing that AI can create and capture enormous value, not just improve drug discovery marginally. A key investment filter is whether the market is actually friendly to a venture-backed company, not merely whether the technology is exciting. The guest values honest, unscripted fundraising and firm-building over polished storytelling; credibility comes from execution and genuine support for founders.

Data Points: Businesses using Ramp: 70,000+ - Mentioned in sponsor copy describing Ramp’s customer base. Revenue growth of Ramp customers: 3.2x faster - Sponsor claim about companies running on Ramp. Annual savings from Ramp: 5% on average - Sponsor claim about business savings. WorkOS core enterprise capabilities: SSO, SCIM, RBAC, audit logs - Sponsor copy explaining enterprise-readiness features. Vendor assessment time reduction with Vanta: up to 50% - Sponsor copy about Vanta’s agentic trust platform. Security/compliance customers of Vanta: 16,000+ - Sponsor copy describing Vanta’s customer base. New companies seen per week: 4 to 6 - Guest describes current pace of sourcing and meetings. Companies seen in first couple months at previous firm: 500 - Guest contrasts earlier high-volume sourcing with current more selective approach. People in the frontier circle: ~250 - Mike’s estimate of entrepreneurs and researchers pushing the frontier. OpenAI headcount referenced: 200 people - Used to argue that individual researchers still can matter despite scale. Compute spending referenced by researchers: $750 billion - Illustrative scale of resources some think are required for progress. Observed backing impact on rounds: 5% to 10% max - Guest estimates how often portfolio companies might not have raised without their firm’s help. Sunday Robotics company age: under 2 years - Guest notes the company’s speed from formation to hardware development. Sunday Robotics beta timeline: beta end of this year - Team expectation described by the guest. Chai Discovery customer references: top 10 pharma companies - Guest cites early evidence of AI value in biology. Portfolio/company examples: Harvey, Sunday, Chai, Suno, Notion, Rippling, Sigma - Used as examples of companies and research bets discussed throughout the interview.

Pivotal Quotes: "I believe in like the great man and great woman theories of history." — Guest: Explaining the importance of high-agency individuals in shaping outcomes in AI and investing. "Nothing is true until it is shipped." — Guest: Discussing rapid iteration in robotics and the importance of real-world delivery over theory. "If we don't have it, we're naturally not competitive or we're at least not independent." — Guest: Arguing that compute independence is essential to U.S. competitiveness and national security.

Implications: Investors should focus on frontier talent, concrete product-market fit, and physical constraints like compute and energy. For AI broadly, open-source diffusion and infrastructure buildout will shape who captures value and whether the U.S. stays competitive.