The Trillion-Dollar Agent Checkout
Brandon Teitel@brandon100
−0.2%
since Sep 30
The idea
I believe AI agents are about to start spending real money on our behalf, and within a year that could mean a trillion dollars in purchases nobody clicks through. To capture it, I built a basket across the whole stack: storefronts agents shop, the payment rails that clear their transactions, the identity and security layers that verify an agent is authorized to spend, the platforms running these agents, and the compute underneath.
What it owns
- Agents need machine-readable catalogs and checkout flows optimized for automated transactions, giving first-mover advantage to platforms that wire agent commerce into their infrastructure. · AmazonIt operates the largest merchant catalog with fastest fulfillment and runs the cloud that agents operate on, giving it leverage with both sides of the transaction., ShopifyIt is connecting millions of independent merchants to agent checkout through a single integration, becoming the default way machines reach the long tail of commerce., Booking HoldingsIt owns travel inventory and the most painful shopping experience, making it an obvious place agents will direct users to book instead of direct alternatives., WalmartIt owns the everyday household basket that people most readily delegate to agents, and opening its catalog to machine ordering keeps it the default for recurring purchases., DoorDashIt operates the delivery layer for food ordering, the smallest and most frequent purchase type agents get trusted with first.23%
- Every agent-initiated purchase still requires a payment network to clear it, and the networks that build agent-native tooling and stay compatible with machine-initiated transactions capture that flow. · VisaIt operates the largest card network that agent purchases will clear through, and its agent-payment tooling aims to make card-based machine transactions the standard., MastercardIt is one of the two networks agents cannot route around and is building agent-native payment capabilities to stay compatible with machine-initiated transactions., PayPalIt operates wallets that can hold funding sources and spending permissions, positioning itself as the account agents plug into to access consumer money., CoinbaseIt is the main regulated on-ramp for stablecoin payments, so machine-to-machine transactions that skip traditional cards flow through its platform., Circle Internet GroupIt issues USDC, the stablecoin most likely to become the money agents move at machine speed, so it benefits if a meaningful fraction of agent commerce flows in stablecoins.22%
- Machine-driven purchases at trillion-dollar scale require instant verification that an agent is legitimate and authorized, making identity infrastructure the toll booth for agent commerce. · CloudflareIt sits at the edge between the internet and merchants, positioned to verify legitimate agents and block fraud bots in a flood of machine traffic that will overwhelm human-scale security., OktaIt is building identity systems for non-human actors, so companies authenticating and authorizing agents to spend money become its customers., Palo Alto NetworksIt secures agents as they handle real money and access, turning agent security into a budget line for companies deploying autonomous purchasing systems., TwilioIt carries the verification messages that let humans approve large agent purchases, so agent spending growth that still requires human confirmation keeps its volumes high.19%
- As autonomous agents take over shopping decisions, the companies that own and operate those assistants capture value whether or not they operate their own storefronts. · MicrosoftIt owns Copilot and a stake in OpenAI, plus the Azure infrastructure running agent workloads, so it collects fees across agent platforms and compute regardless of which assistant users choose., AlphabetIt owns search, where shopping begins, and Gemini can turn that into an agent that finds and purchases, converting its distribution advantage into agent-driven transactions., SalesforceIt built Agentforce for business agents that procure and pay, capturing the first wave of agent spending if companies deploy those systems with real purchasing authority.17%
- Agents that browse, compare, and execute purchases all day consume enormous amounts of inference compute, concentrating spending on the chip and infrastructure suppliers that power them. · NVIDIAIt supplies most inference chips running agent workloads, so every expansion in autonomous shopping volume drives more demand for its processors., BroadcomIt builds custom silicon for the major platforms designing their own chips for agent workloads, so agent infrastructure expansion drives orders for its processors.10%
- Cash9%