The Next Uber Will Be an AI Agent Marketplace
Andreessen Horowitz’s crypto arm (a16z Crypto) has published a thesis arguing that the next major consumer internet platform will be built around AI agents that transact on behalf of users. The firm believes that the next Uber-scale business will emerge from an agentic marketplace that connects machines, not just people, and that crypto’s “intents” and order-flow auction mechanisms provide the blueprint for organizing complex, multi-party transactions.
From Ride-Hailing to Dinner Plans
Today’s marketplaces are optimized for single, well-defined tasks like booking a ride or a hotel room. AI agents, however, will handle multi-step, context-rich goals such as organizing a group dinner, planning a business trip, or procuring supplies for a small business. These tasks require coordinating multiple service providers, negotiating prices, and managing fulfillment — a complexity that traditional two-sided platforms struggle to handle efficiently.
The firm suggests borrowing from decentralized finance, where “intents” let users declare a desired outcome and let a competitive market of solvers find the best execution path. Applied to real-world commerce, an AI agent could broadcast a request (e.g., “book a table for six at a mid-priced Italian restaurant in downtown Austin on Friday at 8pm”) and let restaurants, reservation platforms, and payment providers bid to fulfill it.
Order-Flow Auctions as a Revenue Model
a16z Crypto proposes that the marketplace operator monetize through order-flow payments — charging participants for the right to fulfill agent-initiated requests. This model aligns incentives: merchants pay for high-intent demand, while users get competitive pricing and faster fulfillment. The firm recommends starting with a narrow vertical, such as single-city restaurant reservations, to prove out the mechanics before expanding to travel, procurement, and other categories.
Long-term defensibility, will come from fulfillment reputation and trust infrastructure. Agents will need reliable signals about which merchants actually deliver on their promises, and a decentralized reputation system could become a moat that is hard for centralized incumbents to replicate.
Implications for Crypto and AI
The thesis adds to a growing narrative that AI agents will need crypto rails for payments, identity, and dispute resolution. If agents are to transact autonomously, they will require programmable money, verifiable credentials, and escrow mechanisms — areas where blockchain infrastructure already has a head start. a16z Crypto’s argument suggests that the convergence of AI and crypto may first manifest in commerce, not in trading or speculation.
For builders, the message is clear: the next platform giant may not be a consumer app but an agentic coordination layer. Start small, solve a real coordination problem, and build trust infrastructure that scales.
