In March 2026, Peter Schroeder, the head of global markets at Circle, released a set of data on platform X: In the past nine months, AI agents completed 140 million payments with a total transaction value of $43 million. Among them, 98.6% were settled in USDC, with an average transaction amount of only $0.31. More importantly, the number of AI agents with purchasing power has exceeded 400,000.

This set of data illustrates the issue better 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. There is no human intervention, no bank approval, no credit card verification. Between code and code, and between protocol and protocol, the process that once required human signatures, reconciliation, and clearing has been completed.

Circle's stock price has risen from $60 to $105 in the past few trading days, a gain of 75%. The market interprets this rise as a positive response to the financial report — Circle achieved revenue of $770 million in the fourth quarter of 2025, a 77% year-on-year growth, with a net profit of $133 million. But what’s truly noteworthy is not these numbers themselves, but the structural changes behind the numbers: When AI agents become new economic entities, the logic of the entire financial infrastructure needs to be rewritten.

In this rewriting process, a deeper question is emerging: When AI agents begin to have disposable funds, when they can earn USDC by completing tasks, how will they manage these funds? Payment is the first step, asset management is the second step. The RWA (real-world assets) track needs to answer this second step.

1. From Payment Capability to Asset Holding

To understand what kind of financial services AI agents need, we first need to understand their economic activity patterns.

Deloitte's report “2026 Technology, Media, and Telecommunications Predictions” points out that if businesses and service providers can achieve efficient coordination of intelligent agents, the global agent-based AI market size is expected to reach $45 billion by 2030. The fundamental characteristic of this multi-agent collaboration model is that a complex task is broken down into multiple steps, completed by different specialized agents, with each invocation accompanied by a micropayment.

Taking API calls as an example. An AI application may need to invoke multiple large language models simultaneously, access multiple databases, and utilize several computing resources. Each invocation incurs a small charge, such as $0.01, $0.05, or $0.10. These payment amounts are tiny but occur with high frequency. Circle's data shows that in the past nine months, there have been 140 million transactions, with an average of only $0.31 each — this is precisely the typical characteristic of the micropayment market.

But the problem is, when AI agents continuously generate revenue — whether by providing services to users or by participating in distributed computing networks — funds will accumulate in their accounts. These funds cannot remain liquid indefinitely. Any rational economic entity will consider: What to do with idle funds?

This is the logical starting point for AI agents' transition from “payer” to “asset holder.”

In the traditional financial system, individuals and businesses deposit short-term idle funds into banks, purchase money market funds, or short-term government bonds to earn returns. AI agents similarly need this capability — not for speculation, but to optimize their own economic model. It is necessary to always keep a certain amount of USDC in their accounts for payments, but if the excess amount just sits, it signifies an opportunity cost loss. If the surplus funds could automatically subscribe to a tokenized fund backed by short-term US Treasury bonds and be automatically redeemed when needed for payments, then its “operational efficiency” would be enhanced.

Furthermore, if AI agents need to reserve value for long-term operations or hedge against cost uncertainties caused by gas fee fluctuations, they may develop a need to allocate assets of different risk levels. At that point, they are no longer just a “payer,” but an “investor” — even if this investor is a piece of code.

Circle addresses the issue of turning AI agents into “payers.” But to make them “investors” requires another set of infrastructure.

2. RWA and AI Agents: A "Mutual Run" That Is Happening

What Circle has been doing in the past few years can be summarized as building three layers of capabilities.

The first layer is stablecoin issuance and liquidity networks. According to Circle's official disclosures, as of the end of 2025, the circulation scale of USDC reached $75.3 billion, a year-on-year increase of 72%, with a share of nearly 50% in stablecoin trading volume. 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-level financial services. In March 2026, Circle introduced the Nanopayments system, which aggregates thousands of micropayments off-chain and then periodically packages them on-chain, reducing the transaction costs for developers to zero. The test network has supported 12 EVM chains such as Arbitrum, Arc, Avalanche, Base, and Ethereum. On the payment protocol level, the x402 protocol allows websites or APIs to directly issue HTTP 402 payment requests when returning requests, embedding payments 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 had joined, with the network's annualized transaction scale of approximately $5.7 billion. In February of this year, domestic currency and stablecoin direct payment systems were added in several regions, including Asia and the Middle East.

These three layers of capabilities constitute the “payment infrastructure” of the AI agent economy. However, a complete economic entity also needs “asset management infrastructure” — and this is exactly the area where RWA can enter.

The tokenization of RWA (real-world assets) has primarily focused on “on-chain mapping” of traditional finance in recent years. According to Defillama data, as of June 2025, the total locked value (TVL) of RWA reached $12.5 billion, growing 124% compared to 2024. Global leading banks such as Citigroup and Standard Chartered are exploring application scenarios of RWA in payment settlement, asset management, and cross-border transactions.

But to enter the economic world of AI agents, RWA needs to complete an “AI-native” transformation. This is not simply putting assets on-chain, but rather making assets “understandable by AI and tradable by AI.”

First, data standardization is needed. Leading RWA projects like Ondo Finance are pushing to transform underlying cash flows, legal terms, risk ratings, and other information into structured, machine-readable data formats. In July 2025, Ondo Finance became the first project to launch tokenized US Treasury bonds for global investors, being mentioned in a report published by the President's Digital Asset Market Working Group.

Secondly, logical programmability is required. Rules for dividends, interest payments, buybacks, and settlements are written into smart contracts and executed automatically by code. The interaction between AI agents and assets can then achieve “trustlessness” — there is no need to trust that the counterparty will perform, only to trust that the code will operate according to the established rules.

Thirdly, liquidity fragmentation can occur. After RWA is tokenized, it can theoretically be divided into very small units — $0.01 of treasury bonds, 0.1 square meters of real estate yield rights — which is critical for the small allocation needs of AI agents. Nanopayments have already proven that micropayments are technically feasible; the same logic can extend to micro-investments.

JPMorgan's Kinexys division provides a case reference. In May 2025, Kinexys completed the first public transaction of tokenized US Treasury bonds on the Ondo Chain test network, using Ondo Finance's tokenized US Treasury bond fund (OUSG) and settling through Chainlink's cross-chain infrastructure. This transaction followed the “delivery versus payment” (DvP) model, achieving simultaneous exchange of assets and payments. JPMorgan's Kinexys division currently handles over $2 billion in transactions daily and has facilitated over $15 trillion in nominal value since its inception.

The value of this case lies in it showcasing the combination of RWA and institutional-level payment settlement networks. In the future AI agent economy, the trading entity may shift from JPMorgan to an AI agent, the transaction scale may shrink from millions of dollars to a few dollars, but the underlying logic remains the same — the seamless connection of value transfer and value storage is essential.

3. Beyond the Payment Network, There Is Another Layer of Imagination

If we connect the aforementioned logic, a complete closed loop begins to emerge:

An AI content generation agent accumulates a considerable USDC balance by providing services to multiple clients. Its underlying protocol sets funding management rules: Any balance exceeding 1000 USDC is automatically allocated through a RWA aggregator, evenly distributed among three tokenized short-term treasury bond funds and one tokenized green energy fund. When demand from clients decreases one month and the account balance needs to be replenished, the protocol automatically redeems a portion of the RWA shares in exchange for USDC for daily operations.

In this process, the actions completed by the AI agent include: monitoring account balances, assessing the risk-return characteristics of different assets, executing subscriptions and redemptions, and recording transaction flows for subsequent audits. All actions are accomplished automatically by code, with no human intervention needed.

For instance, an AI travel planner, after booking flights and hotels for a user, receives a transfer of USDC to its account as a budget. While waiting for the flight, the AI agent notices a RWA insurance product for sale based on flight delay data. It automatically subscribes a small share of this insurance using a portion of the temporarily idle USDC in its account. A few hours later, when the flight is delayed, the RWA insurance product triggers a payout according to the rules, increasing the AI agent's account balance.

Each technological module constituting these scenarios already exists: USDC provides the value carrier, Nanopayments solves the micropayment cost problem, the x402 protocol allows payments to be directly embedded in internet requests, tokenized treasury bonds are already functioning on platforms like Ondo Chain, and the DvP settlement mechanism has been validated by JPMorgan. The remaining work is integration — connecting the payment layer, asset layer, and trading layer to enable AI agents to utilize these financial functions just as they would call an API.

Li Ming, the executive president of the Hong Kong Web3.0 Standardization Association, noted in evaluating RWA development, “We hope to find a standardized entry point for Web3.0 that can connect the RWA ecosystem.” For the AI agent economy, this entry point may be precisely the connection between payments and assets.

4. Old Problems in a New World: Risk and Responsibility

Of course, there are still many obstacles to overcome from today's AI payments to tomorrow's AI asset management.

The first is the issue of data authenticity. The underlying assets of RWA exist off-chain, and their state, value, and risk information need to be reliably transmitted on-chain. If AI agents depend on incorrect or tampered data, their “investment decisions” will be flawed. The research report on RWA industrial development released by organizations such as the Hong Kong Web3.0 Standardization Association points out that successfully achieving large-scale implementation of assets requires meeting three thresholds: value stability, legal clarity of rights, and verifiability of off-chain data.

The second is the model risk of AI agents. Even if data is accurate, the investment decision logic of AI agents could still err. Who will be responsible for the erroneous decisions made by AI agents? Is it the person, the protocol, or the AI agent itself? This responsibility attribution issue currently lacks answers at the legal and regulatory levels.

The third is liquidity risk. The on-chain trading depth of RWA is far inferior to 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 the transactions can proceed smoothly is uncertain.

The fourth is regulatory differences. Countries exhibit different regulatory attitudes towards RWA; the legal status of the same asset may vary greatly across different jurisdictions. AI agents need to be capable of recognizing and managing this complexity, placing higher demands on current AI capabilities.

Lastly, there is technical security. Risks such as smart contract vulnerabilities, cross-chain bridge attacks, and private key leaks will not disappear just because the trading entity is AI. On the contrary, when AI agents implement automated trading, the speed and scale at which vulnerabilities may be exploited could far exceed human operations.

Conclusion

Returning to the initial set of data: 400,000 AI agents, 140 million transactions, $43 million.

The significance of these numbers does not lie in the scale itself — compared to the tens of trillions of dollars in payments made by humans annually, $43 million is insignificant. Their true significance lies in revealing a direction: machines are becoming independent economic entities, possessing their own income, their own accounts, and their own payment capabilities.

And when machines have income, they will soon have a need for asset management. This is not a distant imagination, but a natural path in the evolution of the AI agent economy.

Circle is laying the groundwork for this future “payment neural system” — enabling AI agents to transfer value efficiently and at low cost. What RWA needs to do is to become the “energy storage system” of this economy — enabling AI agents to manage their assets as they do their own code.

If this judgment holds, then the question that today's RWA practitioners need to ponder is: when 400,000 AI agents begin searching for configurable assets, and when asset management needs arise after 140 million payments, are the RWA products you hold ready to be evaluated, selected, held, and traded by AI agents?

(This article is based on Circle's official financial statements and announcements, Deloitte's “2026 Technology, Media, and Telecommunications Predictions,” Defillama data, publicly available information from Ondo Finance, JPMorgan Kinexys official disclosures, and the Hong Kong Web3.0 Standardization Association's “RWA Industrial Development Research Report,” and does not constitute any investment advice. The market carries risks; investments should be approached with caution.)