Visa is not known for moving fast, but when it moves, it signals where the plumbing of global payments is heading. The card network is now testing stablecoin settlement with Brale, a firm that builds privacy-enabled stablecoin infrastructure, on the Canton Network—a permissioned blockchain purpose-built for regulated financial institutions. The pilot, reported on June 9, is not a consumer-facing crypto product. It is a back-end experiment in whether a bank-grade distributed ledger can settle obligations between institutions using digital dollars without leaking transaction data to the entire network. For an industry watching AI agents begin to spend money autonomously, the timing is not coincidental. Visa is laying rail for a world where the entity initiating a payment is not a human with a card, but a piece of software that needs finality in under a second and a settlement cost measured in fractions of a cent.
Why Canton and Not a Public Chain?
The choice of the Canton Network tells you everything about Visa’s priorities. Canton is not a general-purpose blockchain like Ethereum or Solana. It is a Daml-based distributed ledger designed by Digital Asset that gives each participant strict control over who sees their data. In a Canton transaction, only the counterparties and any explicitly designated observers can view the details—a model that mirrors the confidentiality of existing interbank messaging but with the atomic settlement that blockchain offers. For Visa, this solves the single biggest objection to putting settlement on-chain: exposing proprietary transaction volumes, counterparty relationships, and pricing to competitors running chain-analysis software. Brale’s role is to issue and redeem the stablecoins used in these test transactions, likely tokenized deposits or regulated stablecoins like USDC, within that privacy-preserving envelope.
The Real Target Is Machine-to-Machine Settlement
Visa’s pilot should be read alongside the broader industry push toward agentic payments. We have written repeatedly about the gap between AI agents that can make decisions and the payment rails they are forced to use—rails built for human-initiated, card-based transactions that settle days later and cost 2-3% per swipe. A network like Visa’s, if it can settle stablecoin transactions with sub-second finality and near-zero marginal cost, becomes viable infrastructure for the high-frequency, low-value payments that AI agents will generate. Think of an agent paying $0.003 for a real-time data stream, or $0.01 for a micro-API call. Card networks cannot support that economics. Stablecoin rails can, but only if they offer the privacy and compliance guarantees that enterprises demand. Visa’s Canton test is an attempt to bridge that gap without ceding control to permissionless protocols.
The Tradeoff: Privacy vs. Open Access
There is a cost to the Canton approach, and it is the same tradeoff that every permissioned blockchain makes: you gain confidentiality and throughput at the expense of open access and composability. An AI agent running on a protocol like Coinbase’s x402 can permissionlessly settle in USDC on Base, interacting with any smart contract in the Ethereum ecosystem. An agent trying to settle through Visa’s Canton-based system would need to be onboarded, credentialed, and granted access to a private network. That is fine for a corporate treasury paying a known counterparty. It is a non-starter for an open marketplace where any agent should be able to pay any other agent without a pre-existing business relationship. Visa is building a settlement layer for the institutional side of the agentic economy—the part that looks like today’s B2B payments, just faster and cheaper. The open, permissionless side will run on different rails.
What to Watch Next
The pilot’s significance will depend on whether Visa moves from testing to production, and whether it attempts to connect the Canton-based settlement system to public stablecoin ecosystems. A bridge that lets an agent pay via x402 on Base and have the recipient settle through Visa’s private rail would be a meaningful convergence. Without it, the agentic payment landscape will bifurcate into a permissioned institutional track and a permissionless crypto-native track, with stablecoins as the common unit of value but very different rulebooks governing who can participate. Visa’s experiment also puts pressure on Mastercard, which has been quieter on stablecoin settlement but faces the same structural pressure as software begins to spend money on its own behalf.