Computer Science is 'So Hot Right Now' | Grace Isford, Lux Capital

New York's Emergence as an AI Hub

• Market Drivers: NY is poised to lead due to concentrated demand (44 Fortune 500 companies headquartered there) and deep research talent pools.
• Key Academic Institutions:
  • NYU Silver Lab (led by Meta's Chief Scientist Yann LeCun) serves as a primary feeder for Meta and top AI roles.
  • Strong NLP research exists at Columbia, Cornell Tech, and Princeton (featuring Daunchi Chen and Kartik Narasimhan).
• Talent Influx Statistics:
  • Approximately 1 in 7 tech workers relocated to NY between 2019 and 2023.
  • 35 AI unicorns are currently based in NY (per SVB report), making it the clear second-largest hub behind San Francisco.
  • In the Lux portfolio alone, 57 of 179 employees moved to NY between Jan 2020 and April 2024 (more than any other location), driven by urban walkability and quality of life.
• VC Capital Distribution: New York captured 20% of AI venture capital last year, trailing San Francisco's ~40% but outperforming all other regions.

Lux Capital's New York AI Portfolio

• Headquartered Companies:
  • Hugging Face: Marketed as the "GitHub for machine learning" with a freemium model, enterprise partnerships, and revenue sharing.
  • Runway: Valued at $1.5B; reimagines filmmaker workflows with generative video tools.
  • Mosaic ML: Acquired by Databricks; research team largely based in NY.
  • Osmo AI: Focuses on "ML for all" applications.
  • Modal: Serverless inference platform with a Swedish presence; driven to NY by talent availability.
  • Together AI: AI cloud provider with a strong go-to-market and engineering office in NY; utilizes proprietary fast inference technology (Flash Attention).
• Investment Strategy:
  • Grace Isford, the youngest partner in Lux history, has sourced eight investments, including Runway, Sakana AI, Maven AGI, and Langchain.
  • Investment stages range from Seed (first check as low as $100k) to Series B, with follow-on checks up to $100M for mature winners like Hugging Face and Anduril.

Sourcing and Diligence Methodology

• Sourcing Approach:
  • Sakana AI (Japan): Identified via a Lux portfolio entrepreneur (David Ha); backed by Ha's pedigree (Goldman Sachs, Google Brain under Jeff Dean, student of Jeff Hinton) and the team's complexity research at the Santa Fe Institute.
  • Network Dependency: Relies heavily on serendipity, reputation, and deep relationships with researchers and founders; avoids cold outreach without prior thematic alignment.
  • Geographic Philosophy: Talent is universal; major metros (NY, SF, London, Paris) are critical for clustering, but global hubs like Tokyo (Sakana) and Paris (Mistral, QAIt, Hugging Face) are viable.
• Diligence Process:
  • Team Composition: Lux employs at least two PhDs and five to six Masters/Engineering degrees internally to assess technical risk.
  • Expert Network: Leverages a network of senior engineering leaders from niche fields (e.g., data infrastructure, bio-AI) for targeted due diligence.
  • Pattern Recognition: Uses repeated exposure to specific verticals (e.g., 10+ enterprise AI pitches) to identify outliers and genuine differentiators.

AI Market Trends and Valuation Logic

• Current Phase: The industry is at an inflection point where technology maturity meets enterprise adoption, but has not yet reached the "flattening" phase of the S-curve for widespread adoption.
• Valuation Multiples:
  • AI companies command high premiums (up to 100x for early revenue) driven by growth metrics, team quality, and future disruption potential rather than current revenue.
  • Risks include potential readjustment for companies failing to meet high valuation expectations or lacking clear exit paths.
• Infrastructure vs. Application:
  • Infrastructure: Requires a 10x advantage over cloud incumbents (e.g., Together AI's inference speed, Modal's developer experience); models are becoming commoditized, necessitating differentiation via fine-tuning or DX.
  • Applications: Success hinges on reimagining user workflows (e.g., Runway for filmmakers) rather than generic AI wrappers.

AI Agents and Future Trajectories

• Definition: True AI agents autonomously orchestrate complex workflows across systems (data warehouses, APIs) rather than performing single tasks.
• Adoption Status: Still in early stages; current success is limited to highly verticalized use cases like coding (Factory AI), customer support (Maven AGI), and sales.
• Barriers to Scale: Integration complexity with legacy enterprise systems and the need for robust data architecture prevent broad autonomous deployment.
• High-Potential Sectors:
  • Biotech/Science: Companies like Evolutionary Scale (spun from Meta) applying AI to material science, chemistry, and biology.
  • Implementation Strategy:
    • Start with Data: Identify repositories where AI can act as an "enterprise search" layer to traverse and retrieve data efficiently.
    • ROI Focus: Prioritize pilot programs that prove tangible value over generic implementation.
    • Build vs. Buy: Advise hiring in-house "hacker" engineers or consultants to scope strategy rather than buying off-the-shelf products that may not solve specific enterprise problems.

Transcript Production Details

• Podcast: "Sorcery," hosted by Molly O'Shea, founder of Sorcery.
• Guest: Grace Isford, Partner at Lux Capital (joined 2022), named to Forbes 30 Under 30 in 2024.
• Pre-Lux Background: Previously at Canvas Ventures (sourced $100M in deployed capital) and Handshake (Mayfield Fellowship at Stanford).
• Sponsor: Archer (urban air mobility/electric air taxis).
• Competitor Promotion: Mention of "Turpentine VC" podcast featuring Ben Horowitz and others.