The Pivot Was Public, the Number Was Specific
Animoca Brands chairman Yat Siu argued at Consensus Miami 2026 that the metaverse’s next phase will be driven by fleets of AI agents transacting on blockchains rather than humans in immersive virtual worlds. The claim was not buried in a panel. It was a keynote. Siu said there could eventually be “50 to 100 billion agents roaming essentially on the internet.” That is not a rounding error. That is a thesis statement — and one with money behind it.
As part of the pivot, Animoca announced a $10 million initiative for developers building AI agent applications through its Animoca Minds platform, framing agents as its next major investment category after the metaverse era closes. The Minds platform itself is not new — Animoca Brands and CryptoSlam’s Ethoswarm announced a strategic partnership in February 2026 to launch Animoca Minds, a service that enables users to deploy and operate persistent AI agents, powered by the Ethoswarm protocol to democratize access to the agentic web. But the Consensus keynote packaged these moves inside a new label. The agentic web, or Web4, marks an ongoing evolution of the internet as it shifts from a human-centric model to a decentralized ecosystem of autonomous AI agents — primary actors with persistent memory and the capacity to negotiate, collaborate, and transact independently. Animoca is now staking its next positioning cycle on that label.
Siu did not casually stumble into this language. He has been explicit: “We’ve entered this era of Web4, which is the agentic web. But it’s only really possible because of Web3, because you have digital property rights.” The argument has internal logic. The gap is in the infrastructure Animoca controls — or more precisely, does not yet control — to make that argument machine-readable.
What Exists Onchain Under the Animoca Name — and What Doesn’t
Animoca Brands is developing an identity and reputation framework for both humans and agents through Moca Network, its digital identity-based Web3 ecosystem. That is not nothing. Moca Chain is an EVM-compatible Layer-1 blockchain specifically engineered for decentralized digital identity and credential management, announced on June 25, 2025, to address the systemic fragmentation of digital identities by enabling users to manage a single, interoperable, and chain-agnostic digital identity through Moca ID. The architecture is real. Moca Chain’s credential infrastructure is designed around privacy-preserved verification, using zero-knowledge proofs and encrypted onchain storage to enable trust-minimized access without revealing underlying user data.
But none of that is .animoca. There is no brand-owned, root-level onchain TLD under the string “animoca.” What exists onchain adjacent to the brand name are third-party registrations. .animoca-brands is a top-level domain secured at the root layer, described as a bold bet on IP, interoperability, and identity in the open metaverse — but it was registered independently, not by Animoca Brands the company. The distinction matters. A third-party TLD under a brand-adjacent string does not give the brand cryptographic authority over the namespace. It gives the registrant that authority. Animoca controls its Moca ID architecture. It controls Animoca Minds. It does not, as of this writing, control a machine-resolvable .animoca root from which it can issue, authenticate, or namespace its own agents.
This is the gap. Siu’s thesis depends on agents having machine-readable identity at scale. He has said: “We are designing a structure that verifies human identity to confirm who owns an agent, and then extends that to an agent’s credentials, reputation and trust.” Credentials and trust require anchors. Anchors require controlled namespaces. The Moca Chain DID framework is one layer. A brand-sovereign .animoca TLD would be a different, more direct one — and it does not exist.
The Use Case That Requires the Address
Here is the concrete problem. Through Animoca Minds, the platform is designed so anyone can create an agent with just an email address, and those agents come with crypto wallets and can make payments and execute transactions. That is the supply side. Now consider the demand side: a third-party AI agent — built by a developer outside the Animoca ecosystem — trying to discover, authenticate, and transact with an Animoca Minds agent programmatically, without a human in the loop.
There is currently no stable, brand-sovereign endpoint for that interaction. No agent.animoca resolves to a capability manifest. No registry.animoca exposes a list of available Minds agents with onchain-verifiable credentials. A third-party agent has no cryptographic way to confirm it is talking to a legitimate Animoca Minds instance rather than a spoofed service. That is not a UX problem. That is an agent-economy problem.
The protocol infrastructure to solve it already exists. Developed by Coinbase, x402 revives HTTP’s long-dormant 402 Payment Required status code and transforms it into a programmable payment rail for autonomous AI systems, natively making payments possible between clients and servers. The mechanics are settled. x402 is an HTTP-native payment protocol enabling autonomous agents and APIs to execute micropayments per request, without human intervention or account setup — when an AI agent encounters the need for payment, x402 allows the agent to pay instantly with stablecoins for API access, no accounts created, no human approval required. The volume is real. As of March 2026, total transactions across all chains exceed 119 million on Base alone, with daily on-chain volume around $28,000, and the protocol handling roughly $600 million in annualized payment volume across the ecosystem.
The identity layer needed to complete the loop is also being built — by competitors. World, an identity project cofounded by Sam Altman, has launched AgentKit, a toolkit that lets AI agents carry cryptographic proof they are backed by a unique human via the World ID system, integrating with Coinbase and Cloudflare’s x402 protocol to make AI agents verifiable economic participants. That combination — World ID for human-backing, x402 for payment rails — describes exactly the kind of primitive Animoca is publicly building toward with Moca Network and Animoca Minds. As Erik Reppel, head of engineering at Coinbase Developer Platform and founder of x402, put it: “Payments are the ‘how’ of agentic commerce, but identity is the ‘who.’” Animoca has pieces of both. What it lacks is a namespace endpoint that ties them together under a brand-sovereign address.
Imagine the full flow. A third-party AI agent calls agent.animoca. The TLD resolves — onchain, verifiably — to a registry of available Animoca Minds agents. The requesting agent pulls a capability manifest from research.agent.animoca, confirms the onchain identity signature matches the Moca Chain DID for that agent, and fires an x402 micropayment to initiate a delegated task. No human login. No OAuth flow. No API key rotating on a server somewhere. The entire handshake — discovery, authentication, payment, execution — runs between machines on open protocols. The pattern is already emerging: software paying for software, automatically, without a human in the loop. Animoca’s own Minds platform is positioned for exactly this pattern. The namespace primitive to expose it machine-first is absent.
This is not a speculative future architecture. Animoca has positioned Minds at the intersection of AI and blockchain, arguing that future AI-agent ecosystems will require decentralized settlement layers, low-cost transaction rails, and on-chain identity systems to operate at scale. That argument is correct and complete — except that it describes a system that needs a root-level resolvable name to function as described. In Siu’s own framework, wallets, tokens, decentralized identity systems, and on-chain payments become machine infrastructure powering an emerging “agent economy.” Machine infrastructure needs machine-readable addresses. The Moca ID system handles human-linked identity. An .animoca namespace would handle agent-facing discovery at the brand level — the layer where an external agent first asks the question: is this an Animoca agent, and what can it do?
The identity problem in the agentic web is precisely this: most websites treat automated traffic as suspicious and even block bots outright — a design approach that is increasingly at odds with a world in which legitimate software agents are gradually acting on users’ behalf. A brand-sovereign TLD, resolving to an authenticated registry, short-circuits that friction for any agent trying to reach into the Animoca ecosystem. Without it, every third-party integration requires bespoke trust negotiation — the kind of friction that kills network effects in an agent economy operating at the scale Siu described in Miami.
The Architecture Points Somewhere
Animoca Brands has framed Minds as part of a wider shift toward what it describes as the “agentic web,” or Web4, where AI systems move beyond information retrieval into autonomous task execution on behalf of users. The framing is credible. The company, co-founded in 2014 by Yat Siu and David Kim, has evolved into one of the most active investors in Web3, with a portfolio spanning more than 600 companies across gaming, DeFi, AI, and real-world assets. The $10M Minds initiative, the Moca Chain identity layer, the stablecoin infrastructure through Anchorpoint — these are real commitments, not slide-deck vision. Animoca Brands has secured a Hong Kong stablecoin license through Anchorpoint, a joint venture established with Standard Chartered Bank (Hong Kong) Ltd. and HKT.
But a keynote that projects 50 to 100 billion autonomous agents transacting onchain — and then does not control the brand’s own onchain namespace — is a thesis that outruns its own infrastructure. The Moca Network handles credentials. x402 handles payments. Animoca Minds handles agent deployment. The one address that ties external discovery to internal capability — the thing a machine would call first — has no brand-sovereign answer. Siu himself has said the most important thing for builders now is to design services around how AI agents will use and interact with them, and that AI agents could eventually become more dominant users than humans. Designing for agents means designing machine-first endpoints. That starts with a name.
The architecture Animoca is describing publicly is coherent. The namespace layer it requires is not yet theirs.
The author holds onchain positions related to this topic. This post reflects independent editorial judgment.