The company behind the world’s most widely used regulated dollar stablecoin has spent the past month making the case that the next dominant users of blockchain will not be humans at all. Circle CEO Jeremy Allaire said over the weekend that he expects AI agents — autonomous software programs that transact on their own — to become the primary users of Arc, the company’s new institutional blockchain, according to Yahoo Finance.
The comments cap a five-week stretch in which Circle has systematically stacked building blocks onto Arc: a mainnet launch in mid-September, an agent marketplace with USDC payment tools days later, identity-verification integrations, and now a public narrative that ties the chain directly to the AI economy. TokenPost described Arc this week as a blockchain built for institutional finance and the AI economy — a pairing that sounds like conference-stage marketing until you trace who is actually building on it.
What Arc actually is
Arc is Circle’s purpose-built layer-1 blockchain, launched on mainnet in mid-September in what the company called an “economic operating system for the internet.” CoinDesk, covering the debut, framed it as Circle’s biggest bet beyond the USDC stablecoin business — a business that at the time stood at roughly 74 billion US dollars in circulation, per CoinDesk’s reporting. Where USDC made Circle a payments company, Arc is an attempt to make it an infrastructure company: the railroad on which tokenized funds, deposits and machine-to-machine payments run.
The design choices reflect the customer. Institutions want permissioned entry points, compliance hooks and predictable fees, not memecoin casinos. To that end, Circle shipped an Agent Marketplace in late September that packages USDC payment rails and live data tools for developers building autonomous agents on Arc, as reported by Bitget. Around the same time, identity firm Socure brought AI-powered identity checks to the chain, tackling the awkward question of how you verify a counterparty that is a piece of software rather than a person.
Why AI agents need their own financial plumbing
The bullish case rests on a simple observation: software agents cannot open bank accounts, swipe cards or wait three days for a wire to clear. If an AI agent booking travel, procuring compute or paying freelancers needs to settle instantly, around the clock, programmatically, stablecoins on public blockchains are effectively the only existing rails that fit. BlackRock made the same argument in September, saying it expects AI agents to pay each other in stablecoins, with Ethereum and Circle’s Arc cited among the candidate rails in coverage of the remarks.
There is already a small but concrete demonstration economy forming. In one example reported by Pluang, AI agents on Arc hired humans to capture 3D scans of physical spaces and paid them in USDC — machines contracting labor, humans cashing out. Anecdotes like that are easy to dismiss as stunts, and at today’s volumes they mostly are. But infrastructure eras in crypto have consistently started exactly this way: a toy use case, a dedicated chain, and a payments giant willing to subsidize the rails until the volumes arrive.
It is also the second major institutional-grade blockchain push this quarter that treats AI agents as a first-class design constraint rather than a buzzword. Tether’s backing of the agentic-payments startup AIsa, which this newsroom covered earlier today, is the same thesis from the other side of the stablecoin market: machine-to-machine payments as the next volume driver for dollar tokens.
The competitive landscape is not waiting
Circle is not operating in a vacuum. Coinbase has been pushing Base as the default chain for agent wallets, OpenAI has experimented with agents holding funds in test networks, and the crypto-native world spent much of this year anointing various “agentic finance” tokens as the next narrative. What distinguishes Arc is the boring stuff: Circle already holds the compliance relationships, the banking partnerships and the reserve infrastructure that make USDC the default institutional stablecoin in most Western markets. Arc extends that moat from the token to the settlement layer underneath it.
The risks are equally structural. Arc’s success depends on institutional adoption of onchain finance generally, which remains hostage to regulation — Circle itself is preparing for enhanced supervision under Europe’s MiCA regime after USDC’s designation as a significant e-money token, as covered previously on this site. And a chain optimized for permissioned institutional flows may struggle to attract the grassroots developer activity that made Ethereum and Solana resilient. Network effects cut both ways.
What it means for everyday investors
Direct exposure to Arc is limited — it is infrastructure, not a consumer token launch, and Circle has been careful not to frame it as a speculative vehicle. The investable signal is second-order. If Allaire’s forecast is even partially right, demand for stablecoin settlement grows in step with agent activity, which strengthens the issuers and the chains that host the flows. Ethereum, which supplies the bulk of USDC settlement today, remains the most liquid proxy for that thesis; Solana’s payments push is another.
The more honest framing is that Circle is making an infrastructure bet on a future most people have not priced: a world where a meaningful share of digital commerce is initiated by software, settled in dollars on public blockchains, and audited through identity systems designed for machines. That future may arrive slowly, then suddenly. When the chief executive of the largest regulated stablecoin issuer says his new chain’s main users will be AI agents, it is less a prediction than a product roadmap — and roadmaps, unlike forecasts, tend to get built.
allaire saying the quiet part loud, no human does 400 microtransactions a day but agents will. if arc grabs even a slice of machine to machine settlement the l1 bet pays for itself. just needs real agent volume now, not demos
five weeks from mid september mainnet to agents are our primary users is a wild pivot speedrun. could be right tbh, but the narrative arrived suspiciously ready made for conference season
machines paying each other in usdc while humans just watch the mempool, cool cool cool. the agent marketplace with payments baked in days after mainnet is the detail everyone is glossing over imo
the marketplace with usdc tools shipping days after mainnet means it was built long before launch. this was always the plan, rails for machines first, humans get the explorer
right? people dunked on the economic operating system line but arc being purpose built for agent to agent settlement is the one use case where a new L1 actually makes sense
agreed a purpose built L1 makes sense here but socure doing identity checks on software counterparties is the weird part for me. kyc for an agent, what does that even verify
apparently it verifies the company behind the agent, not the code that runs it. so a compromised verified agent still sails through. feels like security theater until they can sign the actual runtime
verified entity, unverified behavior. the first exploited agent with a clean badge makes every SOCURE integration look decorative
the 3d scan example is the detail that got me. agents hiring humans and paying usdc sounds like a stunt until you remember every rail era started as a toy use case
Credit to Allaire for saying primary users and not hedging into some AI-plus-crypto word salad. The identity verification layer for agents is the unglamorous part that decides whether institutions actually show up.
74 billion usdc in circulation buys a lot of patience to subsidize agent rails until volumes show up. blackrock pushing the same thesis the same month is the real tell here
exactly, circle is not betting on agents, they are paying to make Arc the default rail before someone else does. the 74B float is the moat
agents as primary users five weeks post mainnet is either the most honest roadmap in crypto or the most expensive narrative. no in between