Moats That Survive AI
Chirag Narang@chiragnarang
+2.5%
since Oct 6
The idea
I believe big tech's cash-rich platforms keep compounding, but only where AI widens the moat instead of eroding it. So this owns the compute landlords renting out AI infrastructure, device and attention ecosystems AI makes stickier, marketplaces and payment rails agents can't route around, and enterprise systems of record that are hard to rip out. I skip seat-priced software and ad models an agent could disintermediate. Google gets a small, deliberate hedge position, and I keep some cash on the side for a broad, large-cap-anchored take meant to be held for years.
What it owns
- Massive installed bases and daily user habits create distribution networks that AI tools make more valuable, not weaker. · Meta PlatformsAI-driven targeting and content recommendations make the attention network stickier while improving ad monetization per user., AppleTwo billion devices represent the primary interface where AI reaches users, a distribution point no competitor controls., NetflixSubscriber scale and content investment create an attention network that AI tools reinforce without disintermediating the subscription model., SpotifyDaily audio engagement and playlist curation form an attention network that AI enhancements strengthen rather than displace.26%
- Every AI model, regardless of capability, requires computing infrastructure that commands recurring fees from the builders that rent it. · MicrosoftAzure infrastructure powers AI models regardless of origin, and the enterprise software stack deepens lock-in as AI integration expands., AmazonAWS remains a top compute provider for any AI infrastructure need, and retail operations control real-world supply agents cannot circumvent., OracleOracle built a compute infrastructure business on top of databases enterprises cannot replace, creating dual dependencies for AI workloads.25%
- Platforms controlling physical supply, inventory, and logistics networks that no algorithm can generate or route around. · Booking HoldingsHotel inventory relationships mean any AI travel assistant must route reservations through the platform that owns the supply., Uber TechnologiesA physical network of drivers and logistics infrastructure exists nowhere else, making it essential to any agent handling transportation or delivery., MercadoLibreOwnership of marketplace, logistics, and payment rails across Latin America creates multiple reinforcing moats on a high-growth platform., AirbnbIndividual home inventory available nowhere else makes the platform mandatory for any AI agent planning or booking accommodations.17%
- Payment settlement rails that process transactions for AI agents just as they do for humans, with no alternative path at scale. · MastercardPayment settlement flows through the card network regardless of whether transactions are initiated by humans or AI agents., VisaAI-driven commerce increases transaction volume flowing across the payment network, multiplying toll collection on a larger base.12%
- Enterprise databases holding the authoritative data that AI agents depend on to act, making the moat deeper rather than shallower. · SAPEnterprise resource planning systems hold the financial and operational data that AI agents require to execute meaningful transactions., IntuitSmall business financial data locked into QuickBooks gives AI assistants access to the source of truth they cannot replace.9%
- A position sized small to hedge on the strongest AI research lab against the risk that search revenue erodes faster than new capabilities monetize. · AlphabetThe company operates the leading AI research capability and cloud infrastructure while facing the highest risk that search revenue erodes.3%
- Cash8%