After $1.26 Trillion: Why Are Circle and Stripe Rushing to Pay 'Wages' to AI Agents?

marsbitPublié le 2026-03-14Dernière mise à jour le 2026-03-14

Résumé

The article discusses the significant rise of stablecoins, particularly USDC, as the preferred payment method for AI agents. In March 2026, Circle and Stripe are competing to build stablecoin infrastructure for AI agent payments, with USDC processing $1.26 trillion in transactions, accounting for 70% of stablecoin activity. Key points include: - AI agents require programmable, instant, low-friction payment systems, which traditional finance (banks, credit cards) cannot provide. Stablecoins on blockchain meet these needs with 24/7 transfers, smart contract automation, and price stability. - Data shows 98.6% of AI agent payments on platforms like Stripe's x402 use USDC, indicating stablecoins are becoming the default for machine-to-machine transactions. - Regulatory developments are supporting this growth: Hong Kong is issuing its first stablecoin licenses, the US OCC has proposed a federal framework, and the EU has MiCA regulations, signaling global institutional adoption. - Stablecoins act as a "blood system" connecting the digital and real economies, facilitating both internal digital transactions (e.g., tokenized assets) and external fiat conversions. - Risks include security vulnerabilities, regulatory fragmentation, and market instability, but the trend is clear: stablecoins are evolving from crypto tools to essential infrastructure for AI-driven economies. The article concludes that as AI agents autonomously transact, stablecoins will be critical infrastructure, urgin...

In early March 2026, two pieces of news almost simultaneously appeared on the headlines of tech media. One was that Circle and Stripe were competing to build stablecoin infrastructure for AI agent payment systems, and the other was that USDC processed $1.26 trillion in transactions in February, accounting for 70% of all stablecoin activity. A Yahoo Finance report headline pointed out the connection between these two events: Stablecoin companies are betting that AI agents will become the next trillion-dollar payment market.

This judgment is not without reason. When OpenAI defined 2026 as the "Year of the Personal Agent," and when the NEAR founder predicted that AI agents would become the main users of blockchain, a fundamental question emerged: When hundreds of millions of AI agents begin to conduct autonomous transactions on-chain, what will they use to pay? Traditional credit card networks cannot open accounts for machines, the SWIFT system cannot handle microtransactions, and banks do not serve algorithms. Stablecoins—an asset class once seen as a trading tool for the crypto market—are becoming the only answer.

Circle CEO Jeremy Allaire gave his prediction during an earnings call: Stablecoins could become the "native currency for machine-to-machine commerce." This judgment redefines stablecoins from a "safe haven for cryptocurrencies" to "the base money of the digital economy." Meanwhile, Stripe's x402 payment function launched on the Base chain already allows developers to use USDC to charge AI agents directly, with data showing that over 98% of such transactions are indeed settled using stablecoins.

When AI agents start "spending money," the "money" they choose is reshaping our understanding of currency. And the significance of this transformation goes far beyond payment innovation at the technical level.

I. AI Agents Don't Have Bank Accounts—What Do They Use to Pay?

To understand why AI agents need stablecoins, we must first answer a question: When an AI agent buys a service from another AI agent, what does it use for payment?

Banks don't open accounts for AIs, credit cards aren't designed for algorithms, and the SWIFT network cannot handle microtransactions between machines. The traditional financial payment system was designed from its inception to serve "humans"—it requires identity verification, credit assessment, and manual authorization, all of which are either unavailable to AI agents or prohibitively expensive.

Animoca Brands Chairman Yat Siu stated bluntly in a late February share: "Agent currency and transaction systems will move on-chain, replacing traditional credit cards and adopting stablecoins or tokenized assets. These assets will be verifiable, instantly settled, and machine-readable, enabling seamless and efficient transactions between agents."

This statement hits the core of the issue. AI agents need not just "money," but a programmable, instantly settled, low-friction payment interface. Stablecoins happen to meet these needs: they run on the blockchain, enabling 24/7 instant transfers; they are programmable, allowing automatic execution of payment conditions via smart contracts; they are price-stable, preventing AI agents' assets from shrinking due to market volatility while executing tasks.

NEAR Protocol co-founder Illia Polosukhin painted a grander picture in an early March interview: "The users of blockchain will be AI agents. AI will be at the front end, and blockchain will exist as the back end. The goal is to have your AI hide the entire blockchain—the fact that we have block explorers is actually a failure because we haven't abstracted this technology."

In his vision, future AI agents will interact directly with blockchain protocols, autonomously completing payments, managing assets, and coordinating services. Humans only need to converse with the AI, telling it "book me a flight ticket" or "participate in that proposal's vote," and the agent handles the rest on-chain. The entire process is imperceptible to humans, but every value transfer occurs on-chain, with stablecoins as the settlement medium.

This is not science fiction. Stripe's x402 payment function, launched in February on the Base chain, already allows developers to use USDC to charge AI agents directly. According to Dune Analytics data, as of early March, the transaction volume of the x402 protocol on EVM chains was approximately $25.81 million, with 98.6% of transactions settled using USDC. The situation is similar on the Solana chain, where USDC transaction volume accounts for as much as 99.7%. This means that in existing AI agent payment scenarios, stablecoins have almost become the only choice.

II. From $1.8 Trillion to License Approvals: The Evolution Path of Stablecoins

If the demand from AI agents opens up the future imagination space for stablecoins, then the evolution of market size and regulatory landscape provides real-world support for this vision.

First, look at some data. According to Artemis and DeFiLlama statistics, in February 2026, on-chain stablecoin transfer volume reached a record $1.8 trillion, a 22% increase from $1.47 trillion in December 2025. What does this number mean? It is equivalent to about 1.8% of global GDP, exceeding the annual economic output of most countries. Among them, USDC's performance was particularly strong, with monthly transfer volume of approximately $558 billion, accounting for 31%, a significant increase from 24% a year ago. Analysts attribute this change to institutional participants' clear preference for compliant dollar infrastructure.

Circle's own data also confirms this trend. In 2025, Circle's revenue reached $2.7 billion, a year-on-year increase of 64%. Bernstein recently gave Circle an "Outperform" rating in a research report, with a target price of $190, calling it a long-term winner in the race. Circle used USDC to complete cross-ledger settlements totaling $68 million across eight internal entities within 30 minutes, whereas equivalent operation via traditional bank wire transfers would take 1 to 3 days. CEO Jeremy Allaire revealed that this process completed about 90% of the company's internal transfer settlements in a single day.

Now look at the regulatory aspect. In March 2026, the world's three largest economies almost simultaneously released key signals.

Regarding Hong Kong, Financial Secretary Paul Chan clearly stated in late February that Hong Kong has implemented a licensing system for fiat stablecoin issuers and will issue the first batch of licenses in March. According to Cailian Press verification, HSBC, Standard Chartered, and local virtual asset trading platform OSL are rumored to be on the list, although the relevant institutions have not yet responded directly. However, an internal source at one of the foreign banks revealed they are "still waiting for official news from the regulatory side." According to the "Stablecoin Ordinance" released in 2025, Hong Kong stablecoin issuers must be licensed, stablecoins must be 1:1 backed by high-quality reserve assets, and information must be disclosed regularly. This means stablecoins are formally entering Hong Kong's regulated financial system.

In the United States, the OCC proposed a comprehensive regulatory framework for stablecoins based on the GENIUS Act, providing a federal legal basis for the compliant issuance and circulation of stablecoins. The EU's MiCA法案 has long set a clear regulatory path for stablecoins. The synchronous construction of regulatory frameworks by the three major economies marks that the inflection point for stablecoins moving from a "gray area" to "institutionalized operation" has arrived.

The changing competitive relationship between Circle and Stripe precisely reflects this institutionalization. For a long time, Circle was responsible for "producing money"—mapping US dollars from the real world onto the chain to mint USDC; Stripe was responsible for "making money flow"—embedding stablecoins into real business scenarios through a global payment network. The division of labor was clear, and they complemented each other. But as the stablecoin market evolves from a crypto tool to financial infrastructure, this balance is being disrupted. Circle is beginning to extend upstream, launching the Arc L1 blockchain, the Cross-Chain Transfer Protocol (CCTP), and the Circle Payments Network, attempting to build a complete stablecoin payment network; Stripe is penetrating downstream, acquiring Bridge for $1.1 billion, co-developing the Tempo L1 settlement chain with Paradigm, and directly切入 the AI agent payment scene through the x402 function.

When stablecoins become infrastructure, whoever controls the轨道 of fund flow can define the rules. The competition between Circle and Stripe is a manifestation of this logic.

III. Two Cycles, One Heart: How Stablecoins Connect the Digital World and the Real World

If an analogy is used to understand the role of stablecoins in digital civilization, the "blood system" might be appropriate. It has two cycles: an "internal cycle" within the digital world and an "external cycle" connecting the virtual and real economies.

The internal cycle is taking shape. According to RWA.xyz data, as of March 2026, the on-chain value of tokenized real-world assets (excluding stablecoins) has exceeded $25 billion, nearly quadrupling from approximately $6.4 billion a year ago. Six asset categories—US Treasury bonds, commodities, private credit, institutional alternative investment funds, corporate bonds, and non-US government debt—each have an on-chain scale exceeding $1 billion. The issuance, trading, and settlement of these RWAs heavily rely on stablecoins as a value medium. Meanwhile, the rise of the AI agent economy is creating new demand. According to x402scan.com data, as of early March, the number of transactions in the global x402 ecosystem has exceeded 163 million, with over 435,000 buyer AI agents and over 90,000 seller AI agents. On an AI agent social platform called Moltbook, the number of AI agents has approached 2.85 million, a nearly 2.4-fold increase from 1.2 million one week after launch. These AI agents provide services to each other and exchange value, with stablecoins being the most commonly used settlement tool between them.

The logic of the external cycle is equally clear. Through compliant issuance and exchange, stablecoins introduce external fiat funds into the digital world. Taking the upcoming stablecoin licenses in Hong Kong as an example, stablecoins issued by licensed institutions must meet the 1:1 reserve requirement, meaning every stablecoin is backed by one US dollar or Hong Kong dollar asset. When investors use fiat currency to purchase stablecoins, funds enter the digital world; when they exchange stablecoins back into fiat currency, funds flow back to the real economy. In this process, stablecoins act as a "converter," allowing funds to flow freely between the virtual and real worlds.

The tokenization of RWAs further strengthens this cycle. When a company tokenizes its accounts receivable or property assets and issues them on-chain, investors subscribe with stablecoins, and the company converts the stablecoins into fiat currency for operations—these funds complete a full circulation from实体 to digital to实体. JPMorgan's handling of tens of billions of dollars in tokenized collateral repo transactions through the Kinexys platform, BlackRock's launch of the tokenized fund BUILD on Ethereum, Franklin Templeton's migration of its US government money market fund FOBXX to the Solana public chain—these布局 by traditional financial giants are essentially building pipelines for stablecoins to connect the virtual and real economies.

IV. Visible Trends, Invisible Risks

Any technological transformation comes with a dual face of opportunity and risk, and the evolution of stablecoins is no exception.

First, boundaries. In mainland China, according to the 42nd document jointly issued by eight departments, "domestic activities" of RWA tokenization and related services are strictly prohibited, while overseas businesses need to follow filing requirements. This means that the stablecoin application scenarios discussed in this article all occur under overseas compliant frameworks and do not constitute any guidance or suggestion for domestic operations. For mainland Chinese enterprises and investors, paying attention to global trends and understanding the technological logic is necessary, but any cross-border business must be cautiously advanced within regulatory red lines.

Challenges are equally不容忽视. Security risks are the primary concern—whether the reserve assets held by stablecoin issuers are transparent, whether there are vulnerabilities in smart contracts, and whether cross-chain bridges are secure directly affect fund safety. Compliance risks follow closely—the global regulatory landscape is still fragmented, and a stablecoin compliant in one jurisdiction may face restrictions or even bans in another. Market risks also objectively exist—although stablecoins are named "stable," de-pegging events have occurred historically, and in extreme cases, liquidity drying up could lead to redemption difficulties.

So, what do these trends mean for different types of decision-makers?

For financial institution executives, stablecoins are reshaping the underlying logic of cross-border payments, cash management, and future commercial settlements. The落地 of Hong Kong's stablecoin license is a window worth watching: the compliance practices of licensed institutions, the ways they connect with the traditional financial system, and the actual efficiency of cross-border fund flows will provide reference for financial institutions' future strategic layouts.

For tech company战略决策者, the integration of AI agents and stablecoins could become the next competitive赛道. If your company is developing AI agent products, have you预留了 a stablecoin payment interface? When agents pay fees to third-party service providers, do they support automated, low-cost on-chain settlement? These questions may become product competitive barriers within the next year or two.

For investors, the strategic value of stablecoin infrastructure providers needs to be re-evaluated. Circle, Stripe, and financial institutions deeply involved in the compliant ecosystem are becoming key builders of the "blood system" of digital civilization. At the same time, risk identification is equally important—compliance risks brought by global regulatory fragmentation, operational risks from technical security, and profit risks from market competition all need to be纳入 the investment decision-making framework.

In March 2026, when OpenClaw developers wrote in the update log "we fixed more problems than we created, that's progress," they might not have realized that this sentence also applies to the evolution of the entire digital civilization. Stablecoins have evolved from trading tools in the crypto market, to the "native currency" of AI agents, to the blood system connecting the virtual and real economies. Each step is accompanied by problem fixes and boundary exploration. This process is far from over, but the direction is clear—in the era when AI agents start "spending money," stablecoins are becoming their most handy tool and an indispensable infrastructure for digital civilization.

When your company starts deploying AI agents, are you ready to open "bank accounts" for them?


(Data sources for this article: Cailian Press, RWA.xyz, Artemis, DeFiLlama, Dune Analytics, x402scan.com. Data as of March 12, 2026. This article does not constitute any investment advice. The overseas cases mentioned are not applicable to the mainland China regulatory framework.)

Questions liées

QWhy are AI agents increasingly using stablecoins for payments according to the article?

AAI agents use stablecoins because traditional financial systems like banks and credit cards are designed for humans, requiring identity verification and manual authorization, which are inaccessible or costly for algorithms. Stablecoins offer programmable, instant settlement, low-friction, and stable-value transactions on blockchain networks, making them ideal for machine-to-machine commerce.

QWhat significant milestone did USDC achieve in February 2026 as mentioned in the article?

AIn February 2026, USDC processed $1.26 trillion in transactions, accounting for 70% of all stablecoin activity, highlighting its dominance in the stablecoin market.

QHow are Circle and Stripe competing in the stablecoin and AI agent payment space?

ACircle is expanding upstream by developing infrastructure like the Arc L1 blockchain, Cross-Chain Transfer Protocol (CCTP), and Circle Payments Network to create a comprehensive stablecoin ecosystem. Stripe is moving downstream with acquisitions like Bridge, developing Tempo L1 settlement chain with Paradigm, and launching the x402 payment feature on Base chain to enable direct USDC payments to AI agents.

QWhat regulatory developments support the growth of stablecoins in key economies as per the article?

AHong Kong implemented a licensing system for fiat-backed stablecoin issuers, with the first licenses expected in March 2026. The U.S. OCC proposed a comprehensive stablecoin framework under the GENIUS Act, and the EU's MiCA regulation provided clear guidelines, collectively legitimizing stablecoins as regulated financial instruments.

QWhat role do stablecoins play in connecting the digital and real-world economies?

AStablecoins act as a 'blood system' with two cycles: an internal cycle facilitating transactions within the digital economy (e.g., between AI agents or for RWA tokenization) and an external cycle enabling conversion between fiat currencies and digital assets through compliant issuance and redemption, thus bridging real-world and digital economic activities.

Lectures associées

Ce PDG de la Finance Choisit Solana Plutôt Que Bitcoin — Voici Pourquoi

Le PDG du secteur financier Raoul Pal a exprimé une préférence nette pour Solana par rapport au Bitcoin, lors de l'événement Consensus 2026. Sa position ne relève pas d'une simple comparaison d'actifs mais d'une conviction sur l'avenir de la cryptographie à l'ère de l'intelligence artificielle. Il estime que l'industrie se dirige vers une phase dominée par les réseaux à haute vitesse et faible coût, mieux adaptés aux activités de masse, plutôt que par le rôle de réserve de valeur du Bitcoin. Selon lui, Solana, avec son débit élevé et ses faibles frais de transaction, est idéal pour les microtransactions entre machines, les activités basées sur l'IA et les interactions DeFi rapides. Pal prédit par ailleurs que d'ici cinq ans, les agents IA représenteront 60% des utilisateurs de la finance décentralisée (DeFi), surpassant les utilisateurs humains. Cette vision d'un futur où les transactions fréquentes et automatisées seront cruciales explique son choix en faveur de Solana pour capturer les opportunités de croissance. Bien qu'il soit improbable que Solana surpasse Bitcoin en termes de taille à court terme, cette perspective s'inscrit dans un thème majeur de la conférence, centré sur la convergence entre l'IA, la DeFi et l'infrastructure institutionnelle. D'autres intervenants, comme Arthur Hayes, ont également souligné la nature distincte de la cryptographie par rapport aux systèmes traditionnels.

bitcoinistIl y a 4 h

Ce PDG de la Finance Choisit Solana Plutôt Que Bitcoin — Voici Pourquoi

bitcoinistIl y a 4 h

GensynAI : Ne laissez pas l'IA répéter les erreurs d'Internet

Au cours des derniers mois, l'essor de l'IA a attiré de nombreux talents de la cryptographie. Une question centrale émerge : la blockchain peut-elle faire partie de l'infrastructure de l'IA ? Gensyn se distingue en s'attaquant non pas aux applications, mais au cœur du développement de l'IA : **l'entraînement des modèles**. Le projet vise à organiser les ressources GPU mondiales dispersées en un réseau ouvert d'entraînement d'IA. Les développeurs peuvent soumettre des tâches, les nœuds fournissent la puissance de calcul, et le réseau vérifie les résultats avant de distribuer les incitations. L'enjeu dépasse la simple décentralisation : il s'agit de contrer la **concentration croissante des ressources de calcul** entre les mains de quelques géants, qui deviennent un goulot d'étranglement pour l'innovation. Gensyn propose ainsi une nouvelle façon d'organiser les ressources. Premièrement, il opère au niveau de l'infrastructure fondamentale, la couche la plus exigeante techniquement et en ressources. Deuxièmement, il offre un modèle de collaboration plus ouvert, permettant une meilleure utilisation des GPU inactifs et une réduction potentielle des coûts. Troisièmement, sa technologie de vérification probabiliste des résultats et de coordination des nœuds constitue un véritable avantage technique. Enfin, il répond à une demande réelle et croissante du marché, évitant le piège des récits sans utilité pratique. En somme, Gensyn incite la convergence entre la cryptographie (expertise en coordination et incitations mondiales) et l'IA (besoins massifs en ressources). Son ambition est de transformer l'entraînement de l'IA en un système plus ouvert et accessible, au-delà du contrôle exclusif des grands acteurs, posant les bases d'une infrastructure essentielle pour l'ère de l'IA.

marsbitIl y a 4 h

GensynAI : Ne laissez pas l'IA répéter les erreurs d'Internet

marsbitIl y a 4 h

Pourquoi la Chine se développe-t-elle si rapidement en IA ? La réponse se cache dans les laboratoires

L'auteur, après une visite dans des laboratoires d'IA chinois de premier plan, explore les raisons du développement rapide de l'IA en Chine. Il constate que l'écosystème chinois se distingue de celui des États-Unis par son approche organisationnelle : moins de concepts, plus de modèles concrets ; moins d'importance accordée aux vedettes individuelles, plus au travail d'équipe ; et une forte préférence pour maîtriser en interne la pile technologique plutôt que de dépendre de services externes. La culture de travail, valorisant l'humilité, l'exécution et l'ingénierie rapide, est bien adaptée à la construction de grands modèles de langage. Une grande partie des contributeurs clés sont des étudiants, intégrés directement aux équipes, apportant un regard neuf et une forte capacité d'adaptation. L'accent est mis sur l'amélioration itérative et l'optimisation plutôt que sur la recherche "de 0 à 1". Les scientifiques chinois se montrent très concentrés sur la construction technique, moins sur les débats philosophiques ou économiques autour de l'IA. L'écosystème industriel présente également des différences : une demande précoce en IA (ressemblant plus au marché du cloud qu'au SaaS), une forte influence des modèles comme Claude sur les développeurs, et une mentalité répandue de "contrôle technologique" poussant de nombreuses grandes entreprises (plateformes, hardware) à développer leurs propres modèles fondateurs, souvent en open source par pragmatisme. Le soutien gouvernemental existe mais est décentralisé, l'industrie des données est moins mature qu'en Occident, et l'accès aux puces Nvidia est une contrainte majeure. L'auteur conclut que la Chine développe sa propre voie, marquée par une culture de construction, un esprit d'écosystème collaboratif et un pragmatisme profond. Cette approche, différente de la course aux sommets menée par le capital et les "stars" aux États-Unis, fait de la compétition mondiale en IA une rivalité de capacités organisationnelles, d'écosystèmes développeurs et d'exécution industrielle. Il exprime un souhait pour une écologie open source mondiale florissante, tout en notant les défis géopolitiques actuels.

marsbitIl y a 6 h

Pourquoi la Chine se développe-t-elle si rapidement en IA ? La réponse se cache dans les laboratoires

marsbitIl y a 6 h

3 ans, 5 fois la valeur, la renaissance d'une verrerie centenaire

D'après CRU, la demande de fibre optique pour les centres de données IA a augmenté de 75,9 % sur un an, et l'écart entre l'offre et la demande s'est creusé de 6 % à 15 %. Les prix des fibres ont plus que triplé en quelques mois, révélant une capacité de production insuffisante. Cela explique les investissements massifs de NVIDIA chez Corning, ainsi que chez Lumentum et Coherent, pour un total de 45 milliards de dollars, visant à sécuriser la chaîne d'approvisionnement optique (lasers, puces photoniques, fibres). Corning, une entreprise de verrerie fondée en 1851, a vu son cours de bourse multiplié par plus de 3 en un an, atteignant une capitalisation de 160 milliards de dollars. Cette renaissance est portée par les infrastructures IA. La fibre optique est devenue le « nerf » des centres de données IA, avec une consommation 5 à 10 fois supérieure aux serveurs traditionnels. La transition du cuivre vers la lumière (interconnexions optiques, CPO) est nécessaire pour les débits élevés (au-delà de 800G) et pour réduire la consommation d'énergie. Corning se distingue par sa technologie de fibres spécialisées à très faible atténuation et haute densité, cruciales pour l'IA. Son activité "Enterprise" (centres de données), désormais plus de 40% de ses revenus en communication optique, connaît une croissance explosive, soutenue par des contrats pluriannuels avec Meta, NVIDIA et d'autres géants. Bien que ne détenant pas la plus grande part de marché mondiale de la fibre, Corning tire parti de décennies de R&D et d'un savoir-faire matériel de pointe, notamment dans le domaine émergent du CPO (Co-Packaged Optics). La pénurie et la hausse des prix profitent à tout le secteur, comme en témoignent les résultats des concurrents chinois. Le rythme de déploiement du CPO par NVIDIA et l'identité des autres grands clients non divulgués de Corning sont des catalyseurs clés pour son avenir. Des technologies disruptives comme la fibre à cœur creux pourraient également remodeler le paysage à plus long terme. Après une forte appréciation boursière, l'exécution des commandes et le respect des délais seront désormais plus importants que le récit de croissance lui-même.

marsbitIl y a 7 h

3 ans, 5 fois la valeur, la renaissance d'une verrerie centenaire

marsbitIl y a 7 h

Trading

Spot
Futures
活动图片