Seven Moats That Still Hold in the AI Era

2 min read
startup-moatsdefensibilityai-eraproprietary-dataswitching-costsnetwork-effectscompetitive-strategy
Originally from vm.tiktok.com
View source

My notes

Watch on TikTok Tap to open video

Summary

In the AI era, code is no longer a moat because what took months to build can now be prototyped in a weekend. True startup defensibility comes from seven compounding advantages - proprietary data, infrastructure trust, regulatory permissions, structural distribution, community/brand, accumulated capital, and physical infrastructure - none of which can be replicated with money alone.

Key Insight

  • Jasper as cautionary tale: Raised at 1,5 billion USD valuation in 2022, revenue collapsed from 120 million to 55 million by 2024 once Google Docs, Notion, and ChatGPT bundled writing features. Being “just a build” means selling a temporary time advantage.
  • Fast follower advantage has flipped: The gap between first movers and copycats shrank from years to weeks. Copycats learn from your mistakes for free without paying the “innovation tax.”
  • The 7 moat sources with real examples:
    1. Proprietary data - Spotify’s Discover Weekly is defensible because of a decade of listening behaviour from hundreds of millions, not the algorithm
    2. Infrastructure trust - Stripe is woven into financial plumbing; ripping it out requires total re-architecture
    3. Permission moat - Coinbase spent years acquiring state licenses and FinCEN registrations before competitors could even start
    4. Structural distribution - Hyperliquid (11-person team, 3 trillion USD volume) made users build on their layer; leaving means leaving an ecosystem
    5. Community/brand - Notion creators earn six figures selling templates; you can clone the editor but not the community
    6. Accumulated capital/liquidity - Aave managed 26 billion USD TVL with near-zero bad debt over 7 years; you can fork code but not trust
    7. Physical infrastructure - Warehouses, sensor networks, licensed spectrum require permits and construction timelines AI cannot compress
  • The compounding pattern: Obsession with a specific user leads to compounding product insight, then compounding trust, then compounding switching costs. Moats are built from the inside out.
  • The acid test: “If a team with 50 million USD cloned you tomorrow, what would they still not be able to replicate in three years?” If the answer is nothing, you have a feature, not a company.