The 7 Most Powerful Moats For AI Startups

YC’s Lightcone hosts break down Hamilton Helmer’s Seven Powers framework for AI startups, arguing speed beats all early moats and process power wins most vertical battles.

• Cursor ran one-day sprint cycles in 2023–2024; no big company can ship at that cadence, making speed the #1 early moat.
• Vertical AI SaaS agents can capture 4–10% of customer wallet share vs. ~1% for traditional SaaS, making the market 10x larger.
• Per-seat SaaS incumbents (Zendesk, Intercom) face structural counter-positioning trap: AI success reduces headcount, directly shrinking their own revenue.
• Process power (deep, painstaking AI agent tuning) is the dominant defensibility for vertical AI companies — demo versions are weekend hacks; production-grade reliability takes 100x more effort.
• Legora counter-positioned against legal AI leader Harvey by betting on the application layer over fine-tuning, which Harvey over-invested in early.
• ChatGPT still outpaces Google Gemini in daily users despite Gemini being a comparable model and Google having the world’s largest consumer brand.
• DeepSeek’s RL efficiency breakthrough was real but still required expensive foundation model pre-training — media overstated how much it eroded lab-scale economies.
• Memory is already a switching cost for consumer AI: ChatGPT’s personalization creates lock-in that compounds over time.

2025-10-03 · Watch on YouTube