Agent Economy Supply Moats
TJ Stein@tj_stein
+2.2%
since Oct 6
The idea
I believe personal agents will reshape shopping: software won't fall for branding the way people do, so pricing power shifts from ads to whoever owns real supply, merchant integrations, and proprietary data. This basket leans heaviest into supply-owning platforms agents must route through, then the plumbing that makes businesses machine-readable, checkout rails that collect when agents transact, and data and underwriting moats, with pure AI and compute kept deliberately small since the model layer should commoditize.
What it owns
- Agents optimize transactions but must route through whoever controls the underlying supply: warehouses, drivers, merchants, and logistics networks that can't be easily replicated. · Amazon.comIt owns integrated logistics, warehouses, and inventory that agents optimizing for price and speed must route through, giving its supply chain the role search ads once played., Uber TechnologiesIt owns the driver and courier supply network that an agent cannot summon independently, making the operational moat durable even as the AI layer commoditizes., Booking HoldingsIt negotiates proprietary rates and holds exclusive room inventory that travel agents cannot access elsewhere, keeping hotels dependent on its demand., AirbnbIt owns exclusive listings that exist only on its platform, forcing agents planning stays to route through it rather than access inventory through open channels., DoorDashIt has integrated hundreds of thousands of restaurants, stores, and couriers into one network that agents must use rather than rebuild independently.28%
- Machine-readable pricing, verified inventory, APIs, and reputation data become the new optimization layer as agents replace keyword search, and companies selling this infrastructure benefit. · ShopifyIt is the platform millions of merchants use to expose machine-readable pricing, inventory, and APIs to agents in a single update rather than building custom integrations., CloudflareIt sits in front of internet traffic as a gateway layer, positioned to authenticate agents and let businesses charge or control them, creating a tollbooth between agents and commerce., TwilioIt provides the API layer businesses already use to communicate with customers, which can be repointed at agent-to-business communication., YelpIt owns two decades of local reputation data that agents need to choose between plumbers and restaurants, though this depends on agents licensing rather than replacing this data source., SalesforceIt is the system of record where much of corporate America stores policies, service areas, and negotiation rules that businesses must expose to agents.19%
- As agents handle more transactions, payment networks that process agent-initiated commerce collect on every deal regardless of which platform or model wins. · VisaIt collects on every agent-initiated transaction that settles over card rails, creating a persistent revenue stream independent of which platform or model wins., MastercardIt takes a cut on every agent-completed purchase that flows through card networks, collecting regardless of which platform or model initiated the transaction., PayPal HoldingsIt provides a wallet and merchant checkout layer that agents can trust with stored credentials and buyer protection, offering an alternative to direct card network settlement.15%
- Structured datasets like claims history, pricing patterns, and credit records cannot be scraped or commoditized, so companies controlling them retain pricing power when agents automate underwriting and comparison. · Verisk AnalyticsIt owns the insurance industry's claims database, which agents cannot build or scrape, making it indispensable for any agent pricing or underwriting insurance., ProgressiveIt has decades of proprietary pricing and telematics data that survives agent comparison-shopping, allowing it to win share through better risk accuracy rather than margin compression., EquifaxIt owns credit and income records that cannot be rebuilt, making it essential for any agent arranging financing as part of a purchase.12%
- A modest slice of companies positioned at the agent layer itself, sized conservatively since the AI model layer is expected to commoditize over time. · MicrosoftIt monetizes the agent layer through its OpenAI stake, Azure cloud services, and enterprise distribution while staying insulated from model commoditization., AppleIt controls the device layer where personal agents are most likely to live and retains the ability to collect commission on agent transactions, maintaining gatekeeper leverage.10%
- Agents running continuously for billions of users require sustained inference capacity, so semiconductor and foundry companies benefit from growth in total compute demand. · NVIDIAIt supplies the chips running constant agent inference across billions of users, benefiting from growth in total compute demand even if no single AI player dominates., Taiwan SemiconductorIt manufactures chips regardless of which design wins, making it the compute position least dependent on picking a single AI winner.9%
- Cash7%