E405: Why AI Has Made Venture Capital Harder (Not Easier)

Episode Summary

Executive Summary: Michael Gilroy argues AI has made company formation cheap and fast, increasing the number of startups and founders while reducing barriers to entry. That abundance makes early-stage venture harder: more lookalike companies, lower revenue quality, and more consensus-driven investing. Marathon responds with deep diligence, human-centered sourcing, concentrated portfolios, and gross-profit/earnings-based underwriting.

Main Topics: AI has lowered barriers to company creation (Priority: 5/5): AI and modern infrastructure let founders spin up products almost overnight, dramatically increasing the number of startups in each trend and making seed selection harder. Revenue quality is weaker in the AI era (Priority: 5/5): Because buyers often adopt AI products out of fear or FOMO, reported growth can be misleading; investors must test whether usage is sticky or likely to churn quickly. Marathon’s investment framework: trend, TAM, business model (Priority: 5/5): The firm starts with trend durability, then sizes the addressable market with layered market-cap and use-case analysis, and finally assesses whether the business model can earn premium public-market multiples. Gross profit and earnings matter more than ARR (Priority: 5/5): Gilroy argues software and fintech should increasingly be valued like financial businesses: by gross profit, conversion to net income, and eventual earnings power rather than raw revenue. Human diligence remains a critical edge (Priority: 4/5): AI helps with top-of-funnel sourcing and organization, but Marathon believes edge comes from direct customer calls, back channels, body language, and relationship-driven diligence. Small, concentrated, high-touch firm strategy (Priority: 4/5): Marathon intentionally does only two to three deals a year, stays small, and uses deep pre-investment support to win allocation and build trust with founders and LPs. Culture, ego management, and founder alignment (Priority: 4/5): The firm emphasizes egoless debate, EQ/IQ/RQ balance, transparency, and selecting founders and team members with maturity, discipline, and long-term orientation.

Key Arguments: AI has expanded startup formation so much that instead of seeing 2-3 companies per trend, investors may see 20+, making early-stage picking substantially harder. Barrier to entry is nearly gone: founders can build and launch products overnight, so investors cannot rely on product availability as a differentiator. AI-driven buying is often fear-based rather than need-based, which lowers revenue quality and increases churn risk. Transcripts and automated research create no durable edge if every investor is reading the same material; direct human conversations still matter. Marathon evaluates opportunities through three lenses—trend durability, TAM size/depth, and business model quality—rather than starting with team as a differentiator. Market sizing should be layered: start with market cap or end-market analogs, then unpack the real use case, customer concentration, and achievable ACV. ARR can be misleading across differing gross-margin business models; gross profit and eventual earnings conversion are better early-stage metrics. There are two camps of AI companies: AI capital incinerators and AI capital savers, and the difference comes from founder quality, unit economics, and end-market structure. The best founders are mature, focused, and not distracted by tech Twitter; poise matters more in a rapidly changing market. Vertical AI can be durable when it solves specialized workflows that labs or horizontal incumbents would take a long time to replicate. Marathon wins by doing less, showing up deeply, and working as if it is already on the board before a round is even announced. LP trust is built through transparency, rigorous monthly reporting, and co-invest structures that align incentives rather than treating co-invest as a free option for the GP. Debate inside the firm is intense, but ego-free; being wrong is acceptable if it leads to better decisions and returns.

Data Points: Assets under management: $400 million - Marathon Management Partners capital across software and fintech Annual deal pace: 2 to 3 deals per year - How Marathon operates as a deliberately small, concentrated firm Portfolio construction at first fund: 10 to 15 names - Target portfolio size described to LPs LP fee on co-invest: 0 to 12.5% - Marathon’s incentive-aligned co-invest pricing structure Baseline co-invest fee example: 0 and 10 - Typical market structure Gilroy criticized as misaligned Potential uplift threshold: 12.5% - Fee rises only if Marathon beats its underwriting model Seed-to-Series A graduation rate: 100% - Gilroy said nine seed deals over his career all graduated to Series A Co-invest support in portfolio: Dozens and dozens of customer introductions - Example of Marathon providing direct value before a round Investment results sample: 8 of 12 deals - Gilroy and Gokul co-invested on eight of 12 prior-firm deals Capital efficiency example: Burned like 5% of that capital - An AI portfolio company raised another round after using very little of its first financing Negative gross profit example: Negative $5 million gross profit run rate - A fintech company at initial investment, later improved materially Outcome example: $300 million of net income - Reported result for the fintech example after scale and efficiency gains NDR example: 200% - Illustration of strong net dollar retention driven by expansion on a cohort LP trust quote: Most transparent firm in their entire portfolio ever - How some LPs described Marathon’s reporting style Timeline for deploy: Three-year initial deploy; possibly five years - Flexible pacing for fund deployment Historical firm experience: 2009 start in investment banking - Gilroy’s career origin and basis for lessons on failure and resilience

Pivotal Quotes: "Barriers to entry are essentially gone. Gone, completely gone." — Michael Gilroy: On how AI allows founders to create and launch products almost immediately "We segment the world today in our letters to LPs as AI capital incinerators and AI capital savers." — Michael Gilroy: Describing the two broad categories of AI companies Marathon underwrites "This job is not spreadsheets; this is a human-driven job." — Michael Gilroy: Explaining why direct customer contact and relationship-based diligence still matter

Implications: AI will likely produce more startups, more noise, and more consensus in venture. Winners will be firms that combine rigorous financial underwriting with human judgment, deep customer diligence, and concentrated, high-trust relationships.