Halo launches P2P AI inference marketplace on Virtuals Protocol, processing 8 billion tokens in early weeks

Warden Protocol just launched Halo, a peer-to-peer marketplace where participants can earn USDC by handling AI inference requests. The public alpha went live on June 30 on the Base blockchain, and within its first weeks, the network has processed over 8 billion AI tokens across more than 200 supported models.

How Halo actually works

Someone (or some AI agent) needs a language model to process a request. Instead of routing that through OpenAI or Google Cloud, Halo lets it bounce to a decentralized network of providers who compete to fulfill the job and get paid in USDC for their trouble.

The team describes it as “BitTorrent for inferences.” What makes Halo particularly interesting is its integration with Virtuals Protocol, the leading launchpad for tokenized AI agents on Base. Virtuals has tokenized over 18,000 AI agents to date, and previously those agents relied on inference costs denominated in $VIRTUAL. Halo opens up a new lane: distributed, permissionless inference access that doesn’t depend on a single centralized provider.

The integration creates a revenue-sharing mechanism that benefits both agent token holders and the underlying infrastructure. Inference fees can be paid in an agent’s own token, $VIRTUAL, or stablecoins.

The players and the economics

Founding inference contributors on the network include AskVenice and 0G Labs. Warden Protocol maintains its own token ecosystem centered around WARD, which features buyback mechanics tied to network activity.

By denominating earnings in USDC rather than a volatile protocol token, Halo removes a significant friction point in decentralized compute marketplaces. The dual payment option — stablecoins or $VIRTUAL — means the protocol can still capture value for token holders while keeping the on-ramp accessible.

Why this matters for the broader market

Platforms like Virtuals Protocol have been building infrastructure for agent-to-agent commerce, and Halo slots in as the compute layer that makes it work without relying on AWS or Google Cloud.

The $VIRTUAL token could see increased demand if Halo successfully becomes the default inference layer for Virtuals Protocol’s 18,000-plus tokenized agents. Processing 8 billion AI tokens across 200 models in the early weeks is a notable early metric, but sustainability matters more than launch momentum, and whether Halo can convert early momentum into a self-sustaining marketplace will determine if this is a real infrastructure play.

Disclosure: This article was edited by Editorial Team. For more information on how we create and review content, see our Editorial Policy.