Phase raises $45m to build blockchain infrastructure for AI agents

Phase, a San Francisco-based infrastructure company, has launched its Layer 1 blockchain mainnet backed by $45 million in founder funding from Kirill Eves, the entrepreneur behind global payments processor Unlimit. The bet is that autonomous AI agents will soon need their own financial and computational rails, and that no single company should own them.

The timing is deliberate. As agentic AI moves from prototype to production, the question of how agents pay for services, access the open web, and route tasks across competing models is becoming a live infrastructure problem. Phase positions itself as the operating layer beneath all of that: a decentralised network of everyday consumer devices, smartphones first, that collectively provide connectivity, compute, and on-chain settlement for AI workloads.

From payments rails to agent infrastructure

Eves brings credible cross-sector pedigree. Unlimit processes over $40 billion in payments annually across more than 180 countries, giving Phase's founder direct experience of the regulatory, liquidity, and interoperability challenges that any new payments layer must navigate. Phase is not a payments company in the traditional sense, but payments are central to its design. Agents transact using stablecoins or the network's native PHASE token, with support for the x402 payment standard and EVM-compatible smart contracts, meaning autonomous workflows can settle value without triggering a human approval loop for every micro-transaction.

The launch includes Phase Proxy, which routes AI agent web requests through residential and mobile internet connections across the decentralised network, enabling access to sites that would otherwise block automated traffic. The company prices entry at 0.5 USDT per 1,000 requests and 1 GB of traffic, compared with what it says is an average of over $5 per GB among mobile proxy providers, citing Proxyway's 2026 market research. Two further services are scheduled for November: Phase Router, which gives agents a unified access point across leading AI models, optimising for quality, cost and performance; and Phase Compute, which harnesses participating devices to run open-weight AI models for decentralised inference.

"Soon, almost everyone will have AI agents acting on their behalf," said Kirill Eves. "Whoever controls the infrastructure behind AI agents will control what people can achieve. That power shouldn't belong to any single company."

The convergence capital question

Phase sits at the intersection of three forces that Disrupts readers will recognise as structurally linked: the rise of agentic AI, the maturation of stablecoin-based settlement rails, and a growing investor thesis that decentralised physical infrastructure networks, so-called DePIN plays, can compete with hyperscaler cloud services on cost and resilience. The $45 million commitment is founder capital rather than institutional VC, which sidesteps the current caution from mainstream venture funds around pure blockchain infrastructure raises. That structure also gives Eves longer time horizons than a typical Series A would afford, though it concentrates execution risk.

The broader competitive landscape is crowded. Centralised hyperscalers (AWS, Azure, Google Cloud) already offer AI inference and model-routing services at scale, and agentic payment standards such as x402 are still nascent. Whether a decentralised device network can match hyperscaler latency and uptime at commercial scale remains the critical open variable. For enterprise buyers allocating AI infrastructure budgets, the cost proposition is compelling on paper; the reliability question will determine whether Phase graduates beyond early adopters.

For cross-sector investors tracking the agentic economy, Phase is an early signal of where the infrastructure spend could fragment: away from consolidated cloud providers and towards tokenised, distributed networks that monetise idle consumer hardware. If that model gains traction, the implications ripple into cloud-infrastructure valuations, stablecoin adoption curves, and the regulatory perimeter that financial watchdogs are still drawing around autonomous on-chain transactions.