analysis

Visa $7B Stablecoin Run Rate Meets AgentCard

Editorial · Jun 21, 2026 · 8 min read

Visa is now processing stablecoin payments at a $7 billion annualized run rate across more than 160 linked card programs, the company reported this week. That number, while still a fraction of Visa’s total volume, signals that stablecoin spending through existing card rails is no longer experimental. At the same time, Alchemy’s AgentCard launched with a virtual Visa card issued directly to AI agents—not humans—with programmable spending limits enforced at the API level. The two developments, taken together, point to a future where the primary customer for stablecoin settlement might not be a person tapping a phone at a point-of-sale terminal, but an autonomous agent executing a payment on behalf of a user, a business, or another agent.

The $7 Billion Number in Context

Visa’s $7 billion stablecoin run rate is not a projection. It reflects actual settlement volume flowing through Visa’s network from card programs that use stablecoins as the funding source. The 160 linked programs include crypto-native issuers and fintechs that have integrated USDC or USDT into their back-end treasury operations. What matters here is the architecture: the stablecoin is converted to fiat before it reaches the merchant. Visa’s authorization rails remain unchanged. The merchant receives dollars. The stablecoin never touches the merchant’s balance sheet. This is not native stablecoin acceptance. It is stablecoin-funded fiat spending, and it works because Visa’s network of 130 million merchant locations does not need to know or care where the funds originated.

AgentCard Makes the Agent the Cardholder

Alchemy’s AgentCard takes the same Visa rail and issues a virtual card to an AI agent. The agent is the cardholder, not the human who deployed it. Spending controls—daily limits, merchant category restrictions, per-transaction caps—are enforced at the API layer before the authorization request ever hits Visa’s network. This is a fundamentally different model from a human getting a crypto-backed debit card. The human sets the policy. The agent executes within it. The agent can pay for API calls, cloud compute, data feeds, or physical goods without a human in the loop for each transaction. Settlement still happens in fiat at the merchant endpoint, but the funding source can be a stablecoin wallet controlled by the agent’s operator.

Agents as the Distribution Layer

The convergence of Visa’s stablecoin volume and AgentCard’s agent-native architecture raises a structural question: who will be the primary user of stablecoin payment rails in five years? If agents become the dominant transaction initiators—paying for services, settling with other agents, managing microtransactions—then the distribution layer for stablecoins shifts from consumer wallets to agent orchestration platforms. Visa’s network becomes the settlement backbone, while platforms like Alchemy, Coinbase Agent Payments, and x402 handle the agent-to-agent and agent-to-merchant logic. The stablecoin itself becomes a behind-the-scenes funding mechanism, invisible to the merchant and increasingly invisible to the end user.

What’s Still Missing

The merchant side remains the bottleneck. Visa’s integration with OpenAI, which we covered last week, lets agents spend via tokenized Visa credentials, but the merchant still receives fiat. AgentCard follows the same pattern. No major merchant acquirer has announced native USDC or USDT settlement at the point of sale. Until that changes, the agent economy runs on a hybrid rail: stablecoin-native on the funding side, fiat-native on the acceptance side. That is not necessarily a problem. It lets agents access the entire Visa network today without waiting for merchant adoption. But it also means Visa captures the economics of both sides—the interchange, the foreign exchange spread on the stablecoin-to-fiat conversion, and the network fees. The stablecoin issuer gets the float. The agent operator gets programmability. The merchant gets nothing new.

Sources

E
Editorial
Related reading

Related reading