The metaverse was supposed to be for humans. The AI agents coming for it do not care about your digital real estate.
The metaverse was supposed to be for humans. The AI agents coming for it do not care about your digital real estate.
When Yat Siu took the stage at Consensus 2026 and projected 70 to 140 billion AI agents inhabiting digital infrastructure by decade's end, he was describing a future where the real estate agents fight over isn't yours. Not because someone took it. Because they don't play by your rules.
The metaverse was marketed to humans as a new frontier of digital ownership. Scarce virtual land. Limited-edition digital goods. Blockchain-verified property rights. The pitch was that as more of human life migrated online, ownership would matter more, not less. Platform-agnostic NFTs. Interoperable identities. Your digital self, properly titled.
The agents coming don't need any of that.
Christian Catalini, the MIT economist who helped design the Libra blockchain, put it plainly at the same conference: most agents today function as little more than large language models with a credit card bolted on. That is assisted checkout, not autonomous commerce. The agent doesn't browse your digital wardrobe and think "I want that." It executes a task. Price, availability, API response time. That's what drives its decisions. (CoinDesk Opinion, Apr 28 2026)
The distinction matters because the entire economic architecture built for the metaverse assumes human psychology. Scarcity creates value. Ownership signals status. Limited editions drive demand. AI agents don't have FOMO. They don't collect virtual real estate. They don't care if their digital sneakers are numbered 1 of 10 or 10,000 of 10,000. What they do care about is whether the endpoint responds and the cost per call.
This is already creating visible stress fractures in the infrastructure built to support digital ownership. The x402 protocol, which lets agents pay HTTP servers directly rather than routing through platforms, processed roughly $28,000 per day according to Artemis data from March 2026. But the wash trading picture tells a more complicated story. Approximately half of x402 volume appears gamified or synthetic. The a16z crypto team's analysis from April 2026 put actual protocol revenue at around $1.6 million monthly after filtering wash activity, against headline figures suggesting $24 million. The gap between reported and real volume isn't fraud. It's a sign that the infrastructure is being stress-tested by use cases the original designers didn't fully anticipate. (CoinDesk Markets, Mar 11 2026) (a16z crypto, Apr 2026)
Coinbase's developer documentation describes how autonomous agent payments are meant to work in principle. An agent maintains a crypto balance, initiates transactions without human approval, and operates based on its own value assessment. When that works, it's genuinely new infrastructure. Humans need bank accounts and cards and fraud detection and customer service. Agents need a protocol. x402 is building that protocol. The question is what economic assumptions are baked into it. (Coinbase Developer Documentation)
Animoca Brands has been one of the most aggressive investors in this space. The company's FY2024 investor update showed $314 million in bookings, much of it directed at blockchain gaming and digital asset infrastructure. In early May 2026, Animoca launched a $10 million investment program targeting the Minds platform, which is explicitly positioned as an agent-native social layer. (Animoca Brands FY2024 Update) (Animoca Brands Press Release, May 5 2026) The thesis is coherent: if agents become the dominant users of digital infrastructure, the platforms they operate on become the new real estate. But that reframes the investment entirely. The metaverse humans were sold wasn't just a different kind of virtual space. It was a different kind of ownership economy. What Animoca may actually be buying is machine-to-machine commerce infrastructure, not human digital real estate.
The tension emerges when you ask what scarcity means in a machine economy. NFTs worked, briefly, because humans speculated on digital property. When speculation collapsed, so did the value proposition. If AI agents become the primary economic actors in digital spaces, they will optimize for properties humans never designed ownership around. Speed. Throughput. Cost per transaction. Latency. A truly agentic economy doesn't need NFTs or limited editions or blockchain-verified ownership of anything except the ability to execute transactions reliably. The entire human-facing metaverse economic premise might be a transitional architecture, valuable during the migration from physical to digital life, obsolete once the migration is complete and machines are the ones doing the migrating.
This doesn't mean the infrastructure disappears. Blockchain-based ownership verification could remain relevant for agent coordination, settlement, and value transfer between systems that do care about verified state. But the meaning changes. Virtual land in Decentraland was never valuable because of its coordinates. It was valuable because humans agreed coordinates mattered. Agents will never share that agreement. They might use the same blockchain to coordinate entirely different activities, settle entirely different transactions, value entirely different properties.
The metaverse's infrastructure was built to solve a human problem. The agents moving in are not human. This is not a failure of technology. It's a category error in the original investment thesis. The real infrastructure war isn't about who controls virtual real estate. It's about which payment and ownership protocols agents actually adopt. Yat Siu's trillion-agent vision is compelling because it describes an economy operating at scales and speeds no human could navigate. Whether that economy respects the ownership conventions humans designed is a separate question.
The answer, increasingly, looks like no. The agents are coming. They don't care about your digital real estate. They care about the protocol.