How can encrypted payments support autonomous AI business? Google and PayPal provide the answer

In May 2026, at the Consensus conference in Miami, Richard Widmann, Head of Web3 Strategy at Google Cloud, and May Zabaneh, Senior Vice President of PayPal’s crypto business, jointly delivered a clear message: next-generation internet commerce powered by AI agents will run on a crypto payment track. This claim is not mere marketing language—it is the result of the combined effects of structural obstacles encountered by AI agents within traditional financial systems and the technical properties of crypto infrastructure that are natively suited to autonomous machines. As millions of AI agents begin to take on complex economic tasks, they need a payment system—not designed for people, but for software services.

Why Traditional Bank Accounts Are Fundamentally Unable to Serve AI Agents

AI agents cannot directly open bank accounts. This is not because the technology is immature, but because traditional financial systems are incompatible with autonomous software at the underlying-architecture level. Widmann stated plainly at Consensus: “AI agents cannot open bank accounts. This is not just difficult—it is fundamentally impossible.” The structural reasons come from both technical and regulatory layers. Technically, traditional payment APIs and network authentication processes rely on human identities and interaction entry points, and they do not provide programmatic interfaces designed for autonomous agents. From a regulatory perspective, opening a bank account must meet KYC (Know Your Customer) requirements, but AI agents do not have legal or natural person identities and therefore cannot pass identity verification. By contrast, crypto wallets are generated from private keys, require no account-opening approvals, and are naturally suited as settlement tools for machines. Widmann summarized the contrast as: crypto payments are “an excellent machine-readable payment interface.”

How Much AI Agent Traffic Has Already Appeared on the Merchant Side

Whether commercial activity can migrate to AI agents first depends on how prepared merchants are. PayPal recently released survey data covering a large number of merchants: at present, 95% of merchant websites have AI agent traffic, but only about 20% of merchants have machine-readable product catalogs. This data reveals a structural gap between supply and demand—AI agents have already begun actively entering commercial scenarios, while most merchants still present products in a way designed for human consumers. Zabaneh noted that the current transition logic is similar to the early shift from physical stores to online shops: merchants need to present products in agent-readable formats as quickly as possible, or they will miss the next wave of business upgrades. Compared with the offline-to-online transition cycle, the penetration speed of AI agent commerce may be faster, because the demand side is already operating autonomously.

How Crypto Payment Infrastructure Adapts to AI-Native Commerce

Crypto payment tracks inherently have four key architectural attributes that support autonomous AI transactions. First, no identity verification is required—agents sign transactions directly with private keys, without account-opening processes blocking the way. Second, real-time final settlement: blockchain transactions are typically confirmed within seconds with non-reversible finality, so there is no “IOU-style” delay like that seen in credit card networks within banking systems. Third, programmability: smart contracts can predefine spending rules, enforce limits, and automate execution logic—this is exactly the fund-management model that AI agents need to “operate autonomously according to rules.” Fourth, globalization and instant clearing: when AI agents perform cross-border tasks across different jurisdictions, they do not need to rely on the lengthy linkages of multi-country banking systems. Zabaneh positioned PYUSD as “a very natural programmable payment layer,” and emphasized the inherent fit of stablecoins for globalization, AI-native experiences, and the direction of tokenized assets.

How Multi-Party Custody Solves AI Agent Fund Security

AI agents have independent payment capability, which requires corresponding security mechanisms to prevent funds from being misused. Widmann explicitly proposed that multi-party custody will become the core solution for AI agent fund management. He pointed out that AI agents should not control the full private key, but instead should hold only one of the two or three key shares. The core logic of this design is: the agent can initiate transaction requests, but it cannot unilaterally and unconditionally transfer funds—it must obtain the signatures of other authorized parties. In consumer scenarios, this means users or authorized entities can monitor and intercept the agent’s payment actions in real time; in profit scenarios, the revenue earned by the agent still needs to be distributed through a custody framework. Google has expanded its Cloud KMS platform into the crypto custody domain, attempting to deploy multi-party custody mechanisms within cloud infrastructure to provide an enterprise-grade fund security solution for large-scale AI agent deployments.

Why Stablecoins Become the Settlement Medium of Choice for AI Agent Commerce

Market data for 2026 has already validated the dominant position of stablecoins in AI agent payments. According to disclosures by Circle, over the past nine months AI agents completed 140 million payments totaling 43 million US dollars, with 98.6% settled in USDC and an average value per transaction of only 0.31 US dollars. This means that more than 98% of autonomous machine economic activities have already chosen stablecoins as the settlement tool rather than traditional payment methods. The technical foundation supporting this trend is evolving rapidly. In May 2026, Circle launched a USDC nano payment system for AI agents, reducing the minimum transaction amount to 0.000001 US dollars, with no Gas fees charged to users, and covering 11 mainstream public chains. The system uses off-chain micro-payment aggregation and batch on-chain settlement, reducing Gas costs to an ignorable level and directly removing the economic barrier to small high-frequency payments. Experts predict that the volume of autonomous commerce transactions executed by AI agents could reach 3 trillion to 5 trillion US dollars by 2030, and the strategic value of stablecoins as the underlying settlement layer will continue to be released throughout this process.

How Open-Standard Protocols Drive Coordinated Transformation Across the Industry Chain

The large-scale adoption of AI agent payments requires a protocol layer that is interoperable and open—not closed systems operating in isolation. Google took a key step in this direction—launching the Agentic Payments Protocol (AP2) and donating it to the FIDO Foundation. The protocol currently has more than 120 partners, including PayPal. Widmann compared this initiative to donating the x402 network-native payment standard to the Linux Foundation, emphasizing that “open dialogue and open standards are the foundation of infrastructure building.” Industry-leading companies have also sent clear signals of collaboration. AWS has integrated the x402 payment protocol into its AI infrastructure through Amazon Bedrock AgentCore Payments, enabling AI agents to automatically complete service payments in USDC. In May 2026, Exodus launched XO Cash, a stablecoin designed specifically for AI agents, built on Solana, allowing developers to configure independent wallets and spending functions for agents without handing over private keys. From payment protocols to custody solutions and dedicated stablecoins, the industry chain is advancing in sync on multiple levels, and this openness is accelerating the overall formation of the AI agent commerce ecosystem.

What Deep Impacts Will the AI Agent Business Era Bring

The rise of the AI agent economy is driving a fundamental shift from “human-designed payments” to “economic infrastructure designed for machines.” This transformation requires the industry to address several key questions. First is responsibility allocation: if an AI agent makes an incorrect purchasing decision, who should bear responsibility? Zabaneh stated candidly that “this is absolutely a question that our entire industry must seriously think about.” Second is regulatory adaptation: traditional payment regulatory frameworks (such as PSD2’s requirements for strong customer authentication) were not designed with AI agents as payment initiators in mind; whether an AI agent’s cryptographic signatures can be legally recognized as “valid authorization” still has no clear answer. Third is trust mechanisms: in open networks, AI agents rely on the aggregation of identity verification and reputation signals to safely conduct economic activities, but the lack of identity systems today and the risk of complex composite vulnerabilities are major concerns. The last unresolved question is “how to integrate agents into existing capital markets and infrastructure systems,” as Widmann put it. While crypto payment tracks provide launch conditions for AI agent commerce, pushing into a broader scope of asset management and capital allocation still requires integrating agents into the wider financial markets. From payments to assets, from trust to regulation, the AI agent business era has only just begun.

Summary

The remarks by PayPal and Google Cloud executives at Consensus are not isolated statements, but a sharp recognition of the structural changes in AI agent commerce. Traditional bank accounts’ natural exclusion of AI agents forces a brand-new logic: crypto payment tracks, with machine-readable interfaces, real-time settlement, programmability, and global circulation, are deeply aligned with the architecture of autonomous AI commerce. From Widmann’s proposed multi-party custody security scheme to Zabaneh’s emphasis on merchant readiness, from building the open AP2 protocol ecosystem to coordinated progress by industry participants such as AWS and Exodus, the underlying payment infrastructure for the AI agent economy is taking shape as a multi-layered development landscape. When 98.6% of AI agent transactions have already chosen stablecoins as the settlement tool, this self-reinforcing trend is accelerating—and it also means that crypto payment tracks will play a key role in the future of the AI economy.

Frequently Asked Questions

Q: Why can’t AI agents complete payments using traditional credit cards or bank transfers?

AI agents lack legal person or natural person identities, so they cannot meet the KYC authentication requirements of banking systems, and they also cannot autonomously apply for multi-currency bank accounts in cross-border scenarios. Most traditional payment interfaces are designed for human interaction processes and lack standardized programmatic interfaces for autonomous agents. Crypto wallets are managed via private keys, require no account-opening approvals, and naturally fit scenarios where machines operate autonomously.

Q: What is Google’s AP2 protocol?

AP2 (Agentic Payments Protocol) is an open payment protocol launched by Google for AI agents, and it has been donated to the FIDO Foundation. It currently involves more than 120 partners, including PayPal. The protocol is intended to provide a unified machine-readable payment standard for autonomous AI agent business activities, enabling agents to complete transactions seamlessly across different platforms.

Q: How is the security of AI agent payments ensured?

Multi-party custody is the industry’s current consensus security solution. AI agents do not hold the full private key; they only hold key shares. Transactions require signatures from multiple authorized parties in order to execute. Google has extended its cloud-based key management system into the crypto custody domain and deployed this architecture accordingly.

Q: How large is the expected scale of AI agent-driven commerce?

Industry estimates suggest that by 2030, the scale of agent-based commerce led by AI could reach 3 trillion to 5 trillion US dollars. Data that has already been implemented also shows that in just the past nine months, AI agents have autonomously completed 140 million crypto payments totaling 43 million US dollars.

Q: What regulatory and compliance challenges exist for AI agent payments?

At present, there is no special regulatory framework specifically for AI agent payments. In the European Union, AI agent payments must comply with PSD2’s requirements for strong customer authentication; whether an AI agent’s cryptographic signature can be recognized as valid authorization under current law still needs clarification. In addition, if an agent makes an incorrect purchasing decision, responsibility allocation is not yet covered by legal frameworks, so the entire industry must explore solutions together.

Q: How can merchants prepare for the AI agent commerce era?

According to PayPal’s survey, currently only about 20% of merchants have machine-readable product catalogs. Merchants need to adjust product information into agent-readable data structures (such as structured API interfaces and standardized product catalogs) so that AI agents can autonomously discover, evaluate, and purchase products. In addition, merchants also need to consider how to implement automated reconciliation, payment verification, and integration with backend systems within transaction workflows to support high-frequency agent transactions.