Author: RWA Research Institute
Original Title: 99% of AI Payments Use USDC, Circle Quietly Becomes the Biggest Winner, But Where Should AI Agents' Money Be Stored?
In March 2026, Peter Schroeder, Global Market Head at Circle, shared a set of data on the X platform: over the past nine months, AI agents completed 140 million payments, with a cumulative transaction volume of $43 million. 98.6% of these were settled in USDC, with an average transaction value of just $0.31. More importantly, the number of AI agents with purchasing power has exceeded 400,000.
This data speaks louder than any financial report: AI agents are moving from concept to real economic activity.
400,000 AI agents, 140 million transactions, $43 million—this is value exchange autonomously completed between machines. No human intervention, no bank approvals, no credit card verification. Code to code, protocol to protocol, completing processes that previously required human signatures, reconciliation, and clearing.
Circle's stock price has risen from $60 to $105 over the past few trading sessions, a 75% increase. The market interpreted this surge as a positive reaction to the financial report—Circle achieved Q4 2025 revenue of $770 million, a 77% year-on-year increase, with a net profit of $133 million. But what's truly noteworthy is not these numbers themselves, but the structural change behind them: when AI agents become new economic entities, the logic of the entire financial infrastructure needs to be rewritten.
And in this rewriting process, a deeper question is emerging: when AI agents begin to possess disposable funds, when they can earn USDC by completing tasks, how will they handle these funds? Payment is the first step; asset management is the second. The RWA (Real World Asset) sector needs to answer precisely this second step.
I. From Payment Capability to Asset Holding
To understand what financial services AI agents need, one must first understand their economic activity patterns.
Deloitte's "2026 Predictions for the Technology, Media, and Telecommunications Industry" report pointed out that if businesses and service providers can achieve efficient agent collaborative scheduling, the global agent-based AI market size is expected to reach $45 billion by 2030. The basic feature of this multi-agent collaboration model is: a complex task is broken down into multiple steps, completed by different specialized Agents, with each call accompanied by a micro-payment.
Take API calls as an example. An AI application may need to call multiple large language models, access multiple databases, and use multiple computing resources simultaneously. Each call adds up to $0.01, $0.05, $0.1. These payments are extremely small in amount but very high in frequency. Circle's data shows 140 million transactions in the past nine months, averaging only $0.31 per transaction—this is a typical characteristic of the micro-payment market.
But the problem is, when AI agents continuously generate income—whether by providing services to users or by participating in distributed computing networks—funds will accumulate in their accounts. These funds cannot remain liquid forever. Any rational economic entity would consider: what to do with idle funds?
This is the logical starting point for AI agents' transition from "payers" to "asset holders."
In the traditional financial system, individuals and companies deposit short-term idle funds into banks, purchase money market funds, or short-term treasury bonds to earn returns. AI agents similarly need this capability—not for speculation, but to optimize their own economic models. It is necessary to always keep a certain amount of USDC in the account for payments, but if the amount beyond the threshold just sits there, it means a loss of opportunity cost. If excess funds can be automatically used to purchase a tokenized fund backed by short-term U.S. Treasury bonds and automatically redeemed when payment is needed, then its "operational efficiency" is improved.
Furthermore, if an AI agent needs to reserve value for long-term operation or hedge against the uncertainty of gas fee fluctuations, it may generate the need to allocate assets of different risk grades. At this point, it is no longer just a "payer," but an "investor"—even though this investor is a piece of code.
Circle solves the problem of making AI agents "payers." Making them "investors" requires another set of infrastructure.
II. RWA and AI Agents: A "Two-Way Street" Happening Now
What Circle has done over the past few years can be summarized as building three layers of capabilities.
The first layer is stablecoin issuance and liquidity network. According to Circle's official disclosure, as of the end of 2025, the circulation scale of USDC reached $75.3 billion, a year-on-year increase of 72%, with its share in stablecoin trading volume close to 50%. This provides a usable value carrier for AI payments.
The second layer is an efficient on-chain settlement network. In August 2025, Circle launched the Arc chain, specifically for institutional financial services. In March 2026, Circle introduced the Nanopayments system, which aggregates thousands of small payments off-chain and periodically batches them on-chain, reducing the transaction cost for developers to zero. The testnet already supports 12 EVM chains including Arbitrum, Arc, Avalanche, Base, and Ethereum. At the payment protocol level, the x402 protocol allows websites or APIs to directly issue HTTP 402 payment requests when returning requests, embedding payment directly into internet requests.
The third layer is the connection to the traditional financial system. The Circle Payments Network (CPN) connects banks, payment service providers, cross-border clearing institutions, and corporate clients. As of February 2026, 55 financial institutions have joined. The network's annualized transaction volume is approximately $5.7 billion. In February this year, direct payment systems linking local currencies and stablecoins in multiple regions including Asia and the Middle East were added.
These three layers of capabilities constitute the "payment infrastructure" for the AI agent economy. But a complete economy also needs "asset management infrastructure"—and this is precisely the field where RWA can enter.
RWA (Real World Asset) tokenization exploration in recent years has mainly focused on the "on-chain mapping" of traditional finance. According to Defillama data, as of June 2025, the total value locked (TVL) in RWA reached $12.5 billion, a 124% increase from 2024. Global leading banks like Citigroup and Standard Chartered are exploring application scenarios for RWA in payment settlement, asset management, and cross-border transactions.
But to enter the economic world of AI agents, RWA needs to undergo an "AI-native" transformation. This is not simply putting assets on-chain, but making assets "understandable and tradable by AI."
First is data standardization. Leading RWA projects like Ondo Finance are promoting the transformation of underlying cash flows, legal terms, risk ratings, and other information into structured, machine-readable data formats. In July 2025, Ondo Finance, as the first project to launch tokenized U.S. Treasury bonds for global investors, was included in the White House report released by the U.S. Presidential Working Group on Financial Markets.
Second is programmable logic. Rules for dividends, interest payments, repurchases, liquidation, etc., are written into smart contracts and automatically executed by code. Only then can the interaction between AI agents and assets achieve "trustlessness"—not needing to trust the counterparty to perform, only needing to trust that the code will run according to the established rules.
Third is liquidity fragmentation. After RWA tokenization, it can theoretically be divided into extremely small units—$0.01 of treasury bonds, 0.1 square meters of real estate收益权 (收益权:收益权收益权收益权收益权收益权收益权收益权收益权收益权收益权收益权收益权收益权收益权收益权收益权收益权收益权收益权收益权收益权收益权收益权收益权收益权收益权收益权收益权收益权收益权收益权收益权收益权收益权收益权收益权收益权收益权收益权收益权收益权收益权收益权收益权收益权收益权收益权收益权收益权收益权收益权收益权收益权收益权收益权收益权收益权收益权收益权收益权收益权收益权收益权收益权收益权收益权收益权收益权收益权收益权收益权收益权收益权收益权收益权收益权收益权收益权收益权收益权收益权收益权收益权收益权收益权收益权收益权收益权收益权收益权收益权收益权收益权收益权收益权收益权收益权收益权收益权收益权收益权收益权收益权收益权收益权收益权收益权收益权收益权收益权收益权收益权收益权收益权收益权收益权收益权收益权收益权收益权收益权收益权收益权收益权收益权收益权收益权收益权收益权收益权收益权收益权收益权收益权收益权收益权收益权收益权收益权收益权收益权收益权收益权收益权收益权收益权收益权收益权收益权收益权收益权收益权收益权收益权收益权收益权收益权收益权收益权收益权收益权收益权收益权收益权收益权收益权收益权收益权收益权收益权收益权收益权收益权收益权收益权收益权收益权收益权收益权收益权收益权收益权收益权收益权收益权收益权收益权收益权收益权收益权收益权收益权收益权收益权收益权收益权收益权收益权收益权收益权收益权收益权收益权收益权收益权收益权收益权收益权收益权收益权收益权收益权收益权收益权收益权收益权收益权收益权收益权收益权收益权收益权收益权收益权收益权收益权收益权收益权收益权收益权收益权收益权收益权收益权收益权收益权收益权收益权收益权收益权收益极小的单位—这对于AI代理的小额配置需求至关重要。Nanopayments已经证明微支付在技术上可行,同样逻辑可以延伸到微投资。(收益权: usufructuary right) —this is crucial for the small-scale allocation needs of AI agents. Nanopayments has proven that micro-payments are technically feasible, and the same logic can extend to micro-investments.
J.P. Morgan's Kinexys department provides a reference case. In May 2025, Kinexys completed the first public transaction of tokenized U.S. Treasury bonds on the Ondo Chain testnet, using Ondo Finance's tokenized U.S. Treasury bond fund (OUSG), and settled via Chainlink's cross-chain infrastructure. The transaction followed the "Delivery versus Payment" (DvP) model, achieving simultaneous exchange of assets and payment. J.P. Morgan's Kinexys department currently processes over $2 billion in transactions daily and has facilitated over $1.5 trillion in notional value transactions since its inception.
The value of this case lies in: it demonstrates the combination of RWA and institutional-grade payment settlement networks. In the future AI agent economy, the trading entity might change from J.P. Morgan to an AI agent, the transaction scale from millions of dollars to a few dollars, but the underlying logic is the same—value transfer and value storage need to be seamlessly connected.
III. Beyond the Payment Network, Another Layer of Imagination
If we connect the above logic, a complete closed loop begins to emerge:
An AI content generation agent, by providing services to multiple clients, accumulates a considerable USDC balance in its account. Its underlying protocol sets fund management rules: the portion of the balance exceeding 1000 USDC is automatically allocated, via an RWA aggregator, evenly into three tokenized short-term treasury bond funds and one tokenized green energy fund. When client demand drops in a certain month and the account balance needs replenishing, the protocol automatically redeems some RWA shares, exchanging them back into USDC for daily operations.
In this process, the AI agent completes actions including: monitoring account balance, evaluating the risk-return characteristics of different assets, executing subscriptions and redemptions, recording transaction logs for subsequent audits. All actions are automatically completed by code, without human intervention.
Another example: an AI travel planner, after booking flights and hotels for a user, receives a transfer of USDC into its account as a budget. While waiting for the flight, the AI agent detects an RWA insurance product based on flight delay data being offered. It uses a portion of the temporarily idle USDC in its account to automatically purchase a micro-share of this insurance. A few hours later, the flight is delayed, the RWA insurance product automatically triggers a payout according to the rules, and the AI agent's account balance increases.
Every technical module constituting these scenarios already exists: USDC provides the value carrier, Nanopayments solves the micro-payment cost issue, the x402 protocol allows payments to be directly embedded in internet requests, tokenized treasury bonds are already operating on platforms like Ondo Chain, and the DvP settlement mechanism has been validated by J.P. Morgan. The remaining work is integration—connecting the payment layer, asset layer, and trading layer, enabling AI agents to call these financial functions like calling an API.
Li Ming, Executive President of the Hong Kong Web3.0 Standardization Association, commented on RWA development, "We hope to find a standardized entry point for Web3.0, to open up the RWA ecosystem." For the AI agent economy, this entry point might precisely be the connection point between payments and assets.
IV. Old Problems in a New World: Risk and Responsibility
Of course, from today's AI payments to tomorrow's AI asset management, there are still many obstacles to overcome.
First is the issue of data reliability. The underlying assets of RWA are off-chain, and their status, value, and risk information need to be reliably transmitted on-chain. If AI agents rely on erroneous or tampered data, their "investment decisions" will be problematic. The "RWA Industry Development Research Report" jointly released by the Hong Kong Web3.0 Standardization Association and other units pointed out that assets successfully scaled need to meet three thresholds: value stability, clarity of legal ownership, and verifiability of off-chain data.
Second is the model risk of AI agents. Even if the data is accurate, the investment decision logic of AI agents can be wrong. Who is responsible for the wrong decisions of AI agents? Is it the person, the protocol, or the AI agent itself? This issue of responsibility attribution has no answer yet in terms of law and regulation.
Third is liquidity risk. The on-chain trading depth of RWA is far less than that of mainstream cryptocurrencies, and some assets may have poor liquidity. When a large number of AI agents need to redeem the same RWA fund at the same time, whether it can be traded smoothly is uncertain.
Fourth is regulatory differences. Countries have different regulatory attitudes towards RWA, and the legal status of the same asset can vary drastically across jurisdictions. AI agents need to be able to identify and handle this complexity, which places high demands on current AI capabilities.
Finally, technical security. Risks such as smart contract vulnerabilities, cross-chain bridge attacks, and private key leakage will not disappear just because the trading entity is AI. On the contrary, when AI agents achieve automated trading, the speed and scale of vulnerability exploitation may far exceed manual operations.
Conclusion
Returning to the initial set of data: 400,000 AI agents, 140 million transactions, $43 million.
The significance of these numbers is not the scale itself—compared to the tens of trillions of dollars in annual human payment volume, $43 million is negligible. Their true significance lies in revealing a direction: machines are becoming independent economic entities, with their own income, their own accounts, and their own payment capabilities.
And when machines have income, they will soon have asset management needs. This is not a distant imagination, but a natural evolutionary path of the AI agent economy.
Circle is laying the "payment nervous system" for this future—enabling AI agents to transfer value efficiently and at low cost. What the RWA sector needs to do is to become the "energy storage system" of this economy—enabling AI agents to manage their assets as they manage their code.
If this judgment holds, then the question RWA practitioners need to think about today is: when 400,000 AI agents start looking for assets to allocate, when 140 million payments start generating demand for asset management, is the RWA product in your hand ready to be evaluated, selected, held, and traded by AI agents?
Twitter:https://twitter.com/BitpushNewsCN
Bitpush TG Discussion Group:https://t.me/BitPushCommunity
Bitpush TG Subscription: https://t.me/bitpush
