Circle 不只是 USDC？Arc、AI Agent，以及下一步

I was fascinated by this idea that if you have programmable money and a programmable dollar, you could transform the way finance works. Circle was very well-positioned because it was doing, in some ways, what Twilio had done in the past: take the existing dollar system, which is very solid and built over years of regulation and banking, use that as a backbone, and make the dollar digital-native. That was the incarnation of USDC. When you have a regulated dollar backing a real stablecoin, customers can bank their business on it, and novel use cases can emerge.

The way you think about Circle is that it’s a neutral market infrastructure company building the primitives for modern finance that is native to the internet. The customers are enterprises, developers, and agents.

I’m doing great. Thank you for having me.

Nomadic

We’ve actually never had anybody on from Circle before, which is a bit surprising to me. We wanted to get you on to give us the holistic view of where Circle’s going. We all know it’s in stablecoins, but what are the other bits and pieces people might be missing from the story? Some things we want to discuss today are where Circle is positioned in today’s market, how it’s set up to succeed in the future, and, more broadly, what people might be missing about the Circle story.

Why don’t we start with your background? Who are you? What did you do before Circle, and what is your day-to-day role there?

1. The Programmable Money Origin

I’m an engineer by background, and I transitioned into product management almost as a natural instinct because I was very interested in what problems we should solve, apart from how to solve them. What brought me to Circle was a belief in programmable money. That belief was fostered through one of my roles about a decade ago, when I was working at a company called Twilio.

If you know Twilio, it pioneered programmable communications. The idea was simple: SMS and voice go over telco carriers, and if builders could write a few lines of software to trigger an SMS or a voice call, they could program communication into applications. I happened to be one of the product managers who worked there when Twilio was still a private company.

I was fascinated by this idea. I wanted to create an application to organize my sister’s wedding in India, with about 2,000 people, and I didn’t know how to do that. I came across this product, entered some code, and was able to trigger broadcast messages and organize the event. I was blown away by the possibilities.

Fast-forward 5 years: I had done product management at Uber and Facebook. I was always thinking about what major shift could happen. I was curious about blockchains and programmable money: what major shift could happen if money were programmable the way communications had become programmable?

In communications, that gave rise to numerous industries, like two-factor authentication. You enter your password, an SMS shows up on your phone, or you get a receipt or notifications. All of this happens through lines of code. Programmable money is really powerful because money isn’t being reinvented or changed for the internet. There is no internet-native money. Since the advent of blockchains, that idea became possible.

I was fascinated by the idea that if you have programmable money and a programmable dollar, you could transform the way finance works. Circle was well-positioned because it was doing, in some ways, what Twilio had done in the past: take the existing dollar system, which is very solid and built over years of regulation and banking, use that as a backbone, and make the dollar digital-native. That was the incarnation of USDC.

When you do that and have a regulated dollar backing a real stablecoin, customers can bank their business on it, and novel use cases can emerge. My pitch to Jeremy and Nikhil was to let me come and focus on the developer side of the programmable side of that business. That’s how I joined Circle.

I have a very similar origin story with first learning about Circle through USDC. I think it was probably the first stablecoin I used regularly, especially with the support of Coinbase. The programmable dollar you described—the ability to access it permissionlessly and use it in composable DeFi—really resonated with me.

I remember likening it to some other peers in Silicon Valley, like Salesforce. I was thinking about the Salesforce stack and all the different applications being built on it, and how much more powerful that was. Then I looked at DeFi, at really primitive DeFi, and thought, “Holy cow, this is the same sort of thing. You’re going to have all these advanced financial applications, but you need a foundation of this programmable dollar. You need stablecoins.”

USDC was probably the first one I became reliant on, and I think to this day I use it in lots of different applications. I store cash in it as well. But let’s talk about Circle today. From a product standpoint, where is Circle as a company? We know USDC, but it sounds like you’ve expanded to lots of different product lines.

2. Circle’s Financial Infrastructure

Circle’s mission is to enable frictionless exchange of value globally. That has been the company’s mission for more than a decade. We think about all our products as fitting within that mission.

The stablecoin was the first product that reached widespread scale. As a side note, last week USDC crossed $100 trillion in transaction volume. That’s a monumental milestone. USDC is the most transacted dollar stablecoin.

But to make it work in different applications, you need a lot of infrastructure around it. Our Circle Payments Network, for instance, is another product line. It ensures that regulated, compliant cross-border payments can be made easily using stablecoins. You need an on-ramp from any fiat currency into a stablecoin, and then it needs to be settled in another or similar fiat currency on the other side of the border. To make that happen, you need infrastructure to connect both legs of the payment. You need RFQ systems that match currencies, so you need an FX system. All of that is our payments stack.

Over the last 5 years, we’ve expanded USDC to more than 30 networks and worked with numerous enterprises, giving them infrastructure to enable stablecoin-based payments. Every time, we would come back to one question: which network should we use to settle these payments?

USDC is on Ethereum, Base, Solana, and many other networks. In enterprise-level discussions, the questions would be: How are risk and compliance teams weighing the risk of who the validators are that settle these transactions? When is a transaction deemed final? What is the cost of holding tokens?

When you’re holding Ether, Solana, or any other token, it’s volatile. There’s accounting complexity associated with it because these tokens fluctuate in value. These questions kept coming up and became a set of problem statements for us to solve. Solving those problems is where the design for Arc was born.

Arc is our latest Layer 1 blockchain. It’s designed to have distributed control from global institutions. When you put these pieces together, Arc is the foundational layer that is supposed to be the economic OS for the internet, and I can describe what that means.

On top of that, we have digital assets, including USDC, stablecoins, EURC, other partner stablecoins, Bitcoin, and many more. Payments are one application, through Circle Payments Network. The way you think about Circle is that it’s a neutral market infrastructure company building the primitives for modern finance that is native to the internet.

Nomadic

All right. I want to get into something we were talking about offline, Gagan. We have a lot of investors who listen to this podcast, and I think this is a question I want to ask when thinking about Circle.

You just outlined a whole stack, and I think a lot of people think of Circle as a stablecoin company, but there’s a lot more. There’s a bigger vision going on, I think, and I want you to try to connect the dots for people. What is Circle up to? What are you trying to solve over the next few years? Pull together the big picture of Circle and where it’s going.

The big picture is that we want every single person connected to the internet to have access to modern financial tools. If you think about how people think about money, it’s very simple. They want to receive payments in a currency that doesn’t inflate over time, and they want their savings to be protected. When they have money in a savings account, they want to be able to earn on it. When they want to buy assets that grow in value, they want to be able to invest wisely.

These opportunities aren’t equally distributed around the world. Take communication before and after the internet. Anyone connected to the internet around the world can send and receive email. That experience is equal. Anyone around the world can instantly send a message on an application like WhatsApp. That experience is equal. But anyone connected to the internet around the world can’t send, receive, and hold money the same way. They can’t invest in assets the same way.

Our North Star is to build financial infrastructure on the internet so builders can create local applications that give people access to the best-in-class global assets. A person sitting in any country should be able to buy US equities, US Treasuries, or representations of those on the blockchain. Does that help? That’s the North Star we’re trying to achieve, and everything we’re building is in service of that.

3. Stablecoin Network Effects

Nomadic

Yeah, totally. When I’m talking to people who aren’t in our little crypto bubble and trying to explain how big stablecoins have gotten, one thing that really helps them understand is comparing transfer volume to Visa and Mastercard, the things they know.

I just pulled this up. This might not be exactly right; it’s what my favorite LLM is telling me. For 2025, USDC on-chain transfers were $18.3 trillion, Visa had $14.2 trillion, and Mastercard had $10.6 trillion. Visa and Mastercard combined were $24.8 trillion, which means USDC transfers are very much competing with what people consider the global standard for financial transactions. When I explain this to people, they’re like, “No way.” They’re almost still in disbelief. It’s funny how this movement has been happening almost in silence for most people.

Yeah, it’s so interesting that stablecoin transaction volume has already leapfrogged Visa and Mastercard’s annual transaction volume, but stablecoins don’t seem that prevalent around the world. I think part of it is a perception issue, because Visa and Mastercard have distribution at retail endpoints around the world. In the physical world, of course, they’re much more reachable. These networks have been built over multiple decades, while the stablecoin network is probably still in its first decade.

The reason the transaction volume has leapfrogged is that Visa and Mastercard are endpoint networks. They solve endpoint merchant payment problems. They help a user who has a different bank account from a merchant make a payment to that merchant. That is primarily the problem they solve.

Stablecoins, on the other hand, solve the problem of being able to pay, but also to collateralize and have a collateral asset, and to create programmatic conditions where traders can trigger transactions if they receive money. In software, each dollar on the internet makes multiple turns, unlike a physical payment made via Visa or Mastercard. You can’t program payments against something where the settlement lag is 48 or 72 hours. Your credit card transactions stay pending for a while before they settle.

With dollar stablecoins, or stablecoins in general, the minute a transaction settles, activity can happen right at that settlement, when it’s triggered. That’s why transaction volume keeps growing much faster: developers are building applications where both humans and, increasingly, AI bots can transact much more quickly.

One of the questions that’s been hard for us to answer as investors, as stablecoins have grown in adoption, is how you place a bet on the growth of stablecoins. I’ve got friends who have nothing to do with crypto, and they know about my involvement and my focus on DeFi, stablecoins, and Ethereum specifically. One of the companies I bring up is Circle. It’s one of—I think it’s the only publicly traded company in the world that we know is working specifically on stablecoins.

The name Circle is always one of two names that come up. Tether as well, and then you’ve got this whole other competitive landscape that you’re probably beginning to see evolve more rapidly with more friendly regulators. So I’m wondering: how do you win the next chapter of stablecoin adoption? You’ve got all this first-mover advantage. You’re publicly traded, so folks can actually buy Circle stock. I believe you have the second-largest amount of stablecoin liquidity in the world, behind Tether. How do you win that next chapter of adoption?

We continue to do what we’re working on, and we welcome the competition. Competition is good because it means the market is expanding, and that’s why many more entrants are coming in. That’s expected. As stablecoins are regulated and regulations become clearer, it makes sense for more market participants to create new stablecoins, both for their countries and within the United States.

We’re very focused on our strategy and the network we’ve built with USDC. Money is a network business, and that network is very strong. As you were describing earlier, stablecoin transaction volume has surpassed Visa and Mastercard. USDC has a third of Tether’s circulation, but in terms of utility and utilization in transactions, it actually transacts much more than Tether. That’s because people trust it as a programmable stablecoin with solid reserves that are audited.

The trust people have in USDC and this trusted network give us a lot of staying power. There’s also the programmability of USDC on 30-plus networks.

That is something builders are continuing to build novel applications with, which reinforces the network effect. For instance, I’ll give two examples of very interesting use cases. You have novel perpetual trading applications like Hyperliquid, which are using USDC. Then you have prediction markets like Polymarket that are also using USDC.

These two are on very different networks, but the experience of using USDC is managed in such a way that people can buy into and use these applications very easily. The reason that happens is because we have something called the Cross-Chain Transfer Protocol, which is like TCP/IP for moving money. It is the most used interoperability protocol for moving stablecoins between networks, and it moves USDC across 30 networks.

Developers can build applications on any of these networks and have their users use a currency that is familiar to them. This kind of network effect is very hard to build. It comes from years of understanding regulations and blockchains, figuring out how to keep the reserves safe, and maintaining the one-to-one peg: a dollar in the reserve is represented as a dollar on the network. This infrastructure is very resilient, so we are continuing to double down on it and make sure we can fulfill as many use cases as possible.

Nomadic

Gagan, there’s a lot in what you just walked through. I don’t know your business intimately, but I think the work you’ve done to get on all these different chains and create deep liquidity for USDC is very important. From the outside looking in, I’d call that a bit of a moat. Liquidity begets liquidity, and it’s not easy. You can’t just have that set up on day zero, so I think that’s definitely working for you guys.

Something else I’ve been thinking about is that there are chains that die and chains that are showing promise. I’d assume you have to put your chips where you think the budding ecosystems are going to be. One we’ve seen recently is Hyperliquid. I was looking this up earlier: about 63% of USDC is on Ethereum mainnet, but Hyperliquid is a growing beast in the sector, and I think something around 9% of USDC is on Hyperliquid.

A lot of people in crypto have been talking about this aligned quote asset V2 thing, this deal you’ve done with Hyperliquid. I don’t know if you can get into the specifics of that deal, but maybe you could walk through the high-level thinking. Circle almost has to be a bit of a speculator, I would think, saying, “Hey, this looks like a partner for the long term that we should be hitching our wagon to in some ways.” Does that thinking make sense?

Yeah, that’s a very interesting question. There was a time when we started with USDC deployments on only 4 or 5 networks. That was about 2 and a half years ago. If you roll back the clock about 3 years, USDC was on Algorand, Arbitrum, Ethereum, Solana, and a couple of others.

We had so much demand for stablecoins from novel emerging networks that we had to figure out how we were going to fulfill it. Every network had a very different pitch. There were applications taking off, and as they did, developers wanted to platformize them and create a new network. In other cases, networks were solving very specific problems for payments or trading.

We had a list of about 40 to 50 networks that wanted stablecoins, so we had a choice. We could start with bridged USDC. We first tried to say that we had to make a choice about deploying USDC on only some networks, as you said. So we released an open-source standard called Bridge to Native. We said, “Networks can develop their own wrapped stablecoin.” Imagine USDC is wrapped by another asset: you can reserve USDC in a contract, then have wrapped USDC. But that experience isn’t great for users.

We did a lot of hard engineering work to figure out how we could scalably deploy USDC on as many networks as possible. We created our blockchain partnerships product line. Our approach was that every network had to pass certain criteria: a compliance bar, a user safety bar, and a security bar. We had a checklist for every network to go through. If it passed that checklist and our partnership incentives were aligned, we would deploy USDC on that network.

We didn’t want to be in a position where we were trying to pick winners. We’re a neutral company. We want as many builders as possible to try novel use cases with stablecoins and succeed. It just wouldn’t be right for Circle to pick winners. We came up with this construct, and we were able to scale USDC to 30-plus networks. One of them was Hyperliquid.

As you said, Hyperliquid created a fantastic perpetual trading product stack and grew it over time. At the time, they were securing a lot of USDC on Arbitrum, and they wanted to migrate it over to their own chain, HyperEVM, for a better user experience, more security, and to unlock new experiences on their network. We worked with them for almost a year to make sure the migration happened smoothly and users had a good experience. That partnership was very fruitful for both teams.

4. Branded Stablecoins for Neobanks

Guys, I want to switch gears and talk about the trend of neobanks, banks in general, and fintechs creating their own stablecoins. Very few have the first-mover advantage Circle has from pioneering stablecoins and then expanding from there to create what you could call a neobank. What are your thoughts on that? I feel like it’s very difficult to dislodge USDC’s advantage.

I think about how dependent I am on it, how it sits in all the different liquidity pools I use, and the fact that it’s used as a unit of account in all the leading crypto apps like Coinbase. What are your thoughts on the difference between your expansion and everyone else’s—the Stripes of the world, the Revoluts, even our friends at Ether.fi, which are more of a neobank? I think they’re also starting to work on a stablecoin offering.

A lot of neobanks approach us as well, and they start with, “We want our own stablecoin.” Our first question is, “What is your goal?” Is your goal to earn interest on a reserve asset, like a dollar in the bank that earns you the Treasury rate? You don’t need to build your own stablecoin for that. We have distribution partnerships with so many companies.

Is your goal to give your customers a stablecoin and let them experience its benefits? That unlocks new business for you. I think a lot of the exploration around creating a stablecoin breaks down very quickly once you go deeper into the goal. If the goal of the neobank is to have a branded stablecoin for its own users, we have programs that let neobanks create a branded stablecoin on top of USDC. That product exists today, and you can do that while still getting the economic benefits.

The magic of stablecoins for a neobank happens when users deposit money into their account. If that money starts doing more for users, they can earn more. They can invest that money in different accounts or DeFi wallets, for instance, to earn a higher interest rate. That’s when the value of stablecoins emerges, or when they help a neobank expand to new markets or regions on the internet very quickly.

Creating the stablecoin is only the starting point. How will you get liquidity and distribution for it? How will you get a Morpho vault where a user can earn 6% or 7% interest by lending into that vault? DeFi on the backend with a simplified user experience on the frontend requires the network effects we were talking about with developers. It requires different developers to use the stablecoin for borrowing and lending.

It requires the stablecoin to be a collateral asset for lending and borrowing use cases. This kind of thing doesn’t happen overnight with a new stablecoin. Builders need to trust it, and there needs to be enough liquidity to borrow it for trading. That’s where I think the concept of every neobank having its own stablecoin starts breaking down.

5. Payments for AI Agents

Nomadic

And speaking of users, there’s a new user archetype emerging that I think Circle is quite opinionated on: the emergent agentic AI economy. We could probably spend a whole podcast talking about this, but succinctly, what is Circle’s positioning and plan of attack? What tooling are you making available for this wave of new users?

It’s a fascinating new user archetype. It’s a machine, and when you give an agent some goals, it needs access to tools. Today, it gets stuck accessing those tools because they’re gated behind having a payment on file, like subscription payments that happen over cards or bank rails.

The beauty of stablecoins is that there’s this protocol, x402. With that, a machine can instantly make a payment to another machine without having to go through all the onboarding, and that payment settles instantly. We have an x402 facilitator, which means you can settle payments more quickly on a specific network, like Arc. There are x402 facilitators on Solana and other networks as well.

USDC is the stablecoin. It turns out that 99.6% of all agentic transactions are with USDC. The machines are, in a way, choosing USDC because of its liquidity, programmability, and existing network effects.

There are still a lot of problems to solve. We’ve invested heavily in a product called nanopayments. It ensures that agents can make very small payments, at the microcent level. You can batch these payments together, because you want to ensure that payments still go through and settle instantly with stablecoins, while the cost of payments stays absolutely minuscule. If you’re making a payment of a microcent and paying a cent in network fees, that doesn’t work.

There are still problems to solve around reducing costs and making payments even faster. Throughput should be much faster so that, at scale, machines make 100,000 times as many payments as humans. That touches on the transaction volume we were talking about earlier. We expect that to grow another 100x from here, so we’re investing in the full agent stack for those use cases.

6. Why Circle Built Arc

As you scale and build for the agentic economy, and take over more of the legacy payments infrastructure, I want to go back to Arc. Arc is its own L1. Why not build an Ethereum L2 and start with the security inherited through Ethereum instead of starting with a new L1? More broadly, you’re across so many different networks, not just Ethereum L1 and L2s. Do you see everything related to USDC collapsing to just Arc, or will Arc be one of many chains where USDC lives? I’ve been trying to make sense of those two visions: the multichain world you already live in versus this new infrastructure, this new L1 that Circle fully owns.

Let’s bring that together. Arc solves very specific problems that other networks don’t, starting with having a validator set of global institutions. Imagine you’re an enterprise and you think stablecoins are great. We were talking about neobanks. You want to move your payment settlement to blockchains. Your enterprise team and risk team are going to ask, “How are these transactions getting validated? When are they deemed settled?” Those are questions any financial risk and compliance team will ask before they greenlight your stablecoin program.

On other networks, you can’t make the same claim. There’s a difference between saying, “The chain is fully decentralized,” and saying, “This network has BlackRock, SBI, DTCC, Mastercard, and Visa as validators.” These are global institutions, and when they’re securing the network, your enterprise risk team will instantly understand who they are and where the trust assumptions lie. If you’re banking your business on a network over the next decade and longer, you want to know that the validators are trusted institutions. That’s something no other network was able to offer.

The compounding effects extend to other problems. You have to hold a volatile network token, and you have to account for it and budget for it. USDC as gas means dollar-based accounting with a regulated stablecoin, which is much simpler for an enterprise and its CFO to understand and account for.

Then you think about when a payment is deemed settled. On all the L2s, they’re either optimistic rollups or ZK rollups. Basically, they settle in batches periodically, so if there’s a reorg, the payment isn’t settled. On L1s with a BFT consensus mechanism, the payment is settled; these reorg problems don’t happen. We chose the architecture that ensures payments are settled within a second. We call that subsecond finality.

This collection of foundational building blocks and features makes Arc unique and suitable for businesses globally to literally bank on it. That’s why, over time, we did contemplate an L2 architecture, then migrated toward building an L1.

Nomadic

I’m trying to get a mental picture of where Arc will sit in the broader blockchain ecosystem. Is the end goal for Arc to win all the transaction volume and all the USDC TVL, essentially creating an island structure? Or should I think of Arc as a new blockchain sitting among all the other blockchains? We were talking offline: USDC really grew up on Ethereum, and we mentioned how much TVL still sits on Ethereum. Will there still be cross-pollination or cohabitation between these blockchains? What’s the architectural vision for how this will all play out?

We believe there will be multiple blockchain networks in the future, and Arc will be one of them. We’re continuing to invest in and support other networks. One of the things we announced: when we think of Arc as an economic OS, one of its layers is the interoperability stack we built. Yesterday, we announced that we’ve opened up that stack for every asset issuer to connect to all the existing networks.

For instance, USDC currently moves with CCTP between Ethereum, Solana, Base, Arc, and 20 other networks. We’ve opened that up to Euro Coin and third-party asset issuers. We’re continuing to invest in this multichain future. We believe that all the world’s use cases—all the world’s financial use cases—aren’t going to be on just one network.

There are purpose-built networks for specific use cases, and we want to make sure they’re successful. What matters is that app builders can build applications and benefit from the utility stablecoins offer. Networks are in service of the benefits users and builders get from using these stablecoins, right?

On Arc, we have this concept of sectors. For payments, for instance, there’s a payment sector. For agents, there’s an agent sector, which has slightly different requirements. But that doesn’t mean all the Ethereum liquidity will migrate into Arc. We absolutely don’t see or envision that, because Ethereum has distributed security, a distributed architecture, and security that’s battle-tested. It serves a very important role in the whole ecosystem.

We’ll continue to build our protocols on Ethereum alongside Arc. We’ll continue to connect the two and support partner assets across multiple networks. Arc is one of the networks that solves very specific problems for builders on the internet and for the emerging new class of customers, AKA agents. I think it’s different from what Ethereum does. I see them as coexisting and interconnected networks.

7. Arc’s Next Milestones

What are some of the milestones we should be looking forward to over the next few quarters? Are there any major products that are public and due to be released? I know Arc was a big one, with Mainnet live, and that’s still very early. What else is on the horizon?

There are 3 milestones, or product areas, I’m really looking forward to. One is the privacy solution on Arc. Until now, it’s been very hard to achieve a good user experience with privacy on blockchains. Once an address is doxxed, users and analysts can track it and know exactly what the holdings are.

Yesterday, we showed a sneak peek of Arc’s privacy solution, which enables asset issuers and customers to have confidentiality without breaking smart contract structures or EVM compatibility. I’m really excited about that because I think it unlocks payroll and novel trading use cases that don’t happen in public.

The second area I’m excited about is the progression of our agent stack. We think about agents as workers: I ask my agent to do something or give me a report, and it does it for me. But we still hear about security issues and hacks, and people hesitate to give agents their personal or bank information because they don’t know where it might get indexed or leak.

Can you create trust guarantees that an agent will protect your data and process it without letting it leave a secure enclave, a trusted zone? We’re investing in that trust stack, where agents publish logs on the blockchain and those audit logs are hashed so their work can be audited. The more audit history an agent has, the stronger it can be as a worker.

It’s similar to how, when you’re looking for an Airbnb, you search for a property online and see that the host has a hundred thousand reviews and five stars, so you think it must be a fantastic place to stay. Or when you hire someone on the internet, or look at GitHub stars, the number of reviews and stars gives us confidence. We’re working on a chain of trust, where we think we can have similar trust in agents and their ability to work if we can produce tamper-proof guarantees on the network.

It’s an interesting bet and idea that we’re excited to develop. Our goal is to facilitate trust in AI on the open internet. We’ve been good at instilling trust in money through open software: people have trusted our open dollar protocols for over a hundred trillion dollars in USDC volume. We believe we can extend that trust stack to agents as well. That’s the second area. The third is continuing to progress our proof of stake. Right now, we have a set of foundational institutions as validators. We want to go through a full proof-of-stake launch with our token and onboard more validators, continuing to make the network more resilient over the next year. Those are the 3 areas I’m most excited about.

So interesting. I think this is a good place for us to start wrapping up. Gagan, thank you so much for your time. It was so great to meet you. This is the first time we’ve ever interviewed anyone at Circle. Oddly enough, I’ve been using USDC since 2017 or 2018, so it’s nice to connect the dots here with someone who’s been powering all that utility. I want to give you the final word before we go.

Thank you so much, guys. It was an absolute pleasure to meet you as well, and I hope we have another conversation. We’ll report progress to users on how Arc and agents are doing. Hopefully, we’ll still be in a world where we can all delegate important tasks to agents and have three-day work weeks.