Mastercard Says Cards Will Prevail in an Agentic World: The Logic Behind More Transactions and What It Means for Merchants
Mastercard's CEO says cards will prevail in an agentic world. We break down the logic behind more transactions, how Agent Pay and the BVNK deal back it up, and what merchants should prepare for.
Key Takeaways
- On its June-quarter earnings call, Mastercard framed agentic commerce as the next evolution in payments, and CEO Michael Miebach stated plainly that cards will prevail in an agentic world. His reasons were the breadth of the acceptance network, risk management, and the fact that a way to dispute a purchase already exists.
- Behind the declaration sits a stack of shipped products: Agent Pay from 2025, Verifiable Intent in March 2026, and the machine-to-machine rail Agent Pay for Machines in June. On stablecoins, the BVNK acquisition brings the fiat conversion layer in house.
- Consumer acceptance has not caught up. A Gartner survey found only 11% of U.S. consumers willing to let AI make purchase decisions. For merchants, the realistic starting points are an authentication setup that does not blanket-block agent traffic, and evidence that holds up in a dispute.
A strong quarter became a stage for an agentic declaration
As payments executives wrestle with fast-moving trends in AI-driven commerce and in cryptocurrency, card networks are gearing up to win more business in both markets.
www.digitaltransactions.netMastercard reported its June-quarter results on July 30, 2026. The numbers were solid. Net revenue reached $9.3 billion, up 14% year over year, according to the official release. Full-year guidance was raised as well.
Even so, a large share of the analyst discussion went to agentic commerce. Miebach opened by calling agentic commerce the next evolution in payments and a significant opportunity for Mastercard, then added a short sentence. It means more transactions.
| Metric | June quarter 2026 | Year over year |
|---|---|---|
| Net revenue | $9.3 billion | +14% (+12% currency-neutral) |
| Net income | $4.4 billion | +19% (+16% currency-neutral) |
| Diluted EPS | $4.97 | +22% from $4.07 |
| Gross dollar volume | $2.881 trillion | +9.2% (+8.0% in local currency) |
| U.S. volume | $858 billion | +5.5% |
| Cards in circulation (worldwide) | 3.46 billion | +7.2% |
| Cards in circulation (U.S.) | 739 million | +4.8% |
| Value-added services revenue | Growth rate only | +20% (+18% currency-neutral) |
That line drew more attention than the results themselves.
We really believe that cards will prevail in that world. The card infrastructure, the card ecosystem, and the Mastercard proposition within that is unique.
Taking apart the claim that transactions will increase
When a card network calls AI agents an opportunity rather than a threat, it reads at first like the expected posture. Follow the argument on the call, though, and the structure is clear.
The line Miebach drew separates where the decision to buy is made from the machinery that makes a purchase work. AI changes how products get found, compared and chosen. What a payment needs in order to complete does not change. Merchants need broad acceptance, a predictable payment experience and protections. Consumers need a way to challenge a purchase made by autonomous software.
That last point sits at the center of the argument. When an agent buys the wrong thing, who does the consumer complain to, and how? Direct bank transfers and direct stablecoin payments have no standard answer yet. Card payments have chargebacks, a circuit that already exists. Verifiable Intent, which Mastercard announced in March 2026, adds a proof layer on top: it ties the consumer's identity, the specific instructions they gave, and the actual outcome of the transaction into a single tamper-resistant record.
Miebach put it in plainer terms on the call. Verifiable Intent lets you challenge a transaction and say you never wanted to buy this. Existing chargeback processes take it from there.
The foundation is already in place. CFO Sachin Mehra said tokenized transactions exceeded 40% of all switched transactions during the quarter. Agentic Tokens sit on the same tokenization base that has carried mobile contactless payments and card-on-file. The idea is not to lay new wiring but to write permission conditions onto wiring that is already run.
So why would transactions increase? Because a single human button press no longer produces a single transaction. In Mastercard's own example, an entrepreneur opening a flower shop tells an agent to launch the store, and purchases of a domain name, hosting, images and checkout pages execute in a chain within a set budget. One request decomposes into several transactions. In the machine-to-machine space, where systems buy data and APIs from each other, the company expects payments of fractions of a cent to occur continuously.
For the network, in other words, agents are expected to lower the value of each transaction while sharply raising the count. That fits a business model that earns on volume.
How far the supporting products have actually shipped
Whether the earnings-call language is a promissory note can be judged from what has already been released.
Agent Pay, the framework introduced in 2025, defined how AI agents get registered and authenticated on the network before they transact. Agentic Tokens are the core of it. Verifiable Intent arrived in March 2026 as the proof layer on top, designed as open source and agnostic to other existing protocols. At launch, chief digital officer Pablo Fourez explained that as autonomy increases, trust cannot be implied and must be proven.
Then on June 10, 2026, Agent Pay for Machines arrived. It handles high-frequency, low-latency, low-value payments between machines end to end, from permissioning through settlement. More than 30 companies including Adyen, Cloudflare, Coinbase and Stripe were named as early supporters.
There is one caveat here that should not be missed. Agent Pay for Machines is designed to span payment types from cards to stablecoins. Mastercard itself, in other words, is building on the assumption that non-card rails could lead in machine-to-machine payments. The claim that cards will prevail is precisely a claim about consumer purchasing, not about the entire field including machine payments.
What the BVNK acquisition changes at checkout
Mastercard's position on stablecoins is addition, not replacement. The company already handles stablecoin settlement on its own network, and Miebach said crypto volume has tripled over the past two years. He singled out peer-to-peer transfers as a focus area.
The bigger step is the acquisition of London-based BVNK. The price is $1.5 billion with a contingent earnout of up to $300 million, and the deal is expected to close by year's end. BVNK lets merchants accept stablecoins at checkout and convert them into local fiat currency. Founded in 2021, it operates as a money services business in the U.S. and holds multiple licenses.
We still need conversion to fiat, and that's what BVNK will do for us.
What makes this matter for agentic payments is that merchant operations stay unchanged. Even when an agent pays in stablecoins, the merchant receives local fiat as before. Accounting, inventory and payment operations can accept a new payment method without being rebuilt. Chief product officer Jorn Lambert explained at the time of the deal that fragmentation across multiple chains is exactly what creates the need for a secure and predictable transaction experience.
Note that Mastercard has not disclosed financial projections for the BVNK business, and the same goes for pricing. Miebach himself cautioned that a stablecoin capability is not the answer to everything.
Where Visa's approach diverges
A day earlier, on July 29, Visa also positioned agentic commerce as a growth driver on its third-quarter call. The two companies say similar things, but they lean in different directions.
| Dimension | Mastercard | Visa |
|---|---|---|
| Earnings date | July 30, 2026 (June quarter) | July 29, 2026 (FY26 Q3) |
| Net revenue | $9.3 billion (+14%) | $11.6 billion (+14%) |
| Agentic payments foundation | Agent Pay, Agentic Tokens | Visa Intelligent Commerce |
| Where trust is anchored | Verifiable Intent, proving the shopper's intent | Agent Score and Agent Directory, rating and registering the agent itself |
| Tokenization penetration | Over 40% of switched transactions | Roughly 60% of global e-commerce transactions |
| Stablecoins | Buying BVNK to own the fiat conversion layer | Visa Stablecoin Platform, multi-coin and multi-chain |
| External connections | 30-plus partners on Agent Pay for Machines including Stripe, Coinbase and Cloudflare | Payment tie-ups with OpenAI and Meta |
The sharpest difference is where trust gets anchored. Visa is working to manage the identity and reputation of agents themselves through Trusted Agent Protocol, Agent Score and Agent Directory. Mastercard puts its weight less on the agent's credibility and more on proving that the purchase intent belonged to the actual person. On distribution, Visa is securing consumer touchpoints such as OpenAI and Meta, while Mastercard is going after the plumbing: machine payments and fiat conversion.
The counterargument that cards get disintermediated
Lining up only the promoters' claims does not show the real picture. The view that agents will route around card networks exists in several concrete forms.
The most prominent is x402, a path where agents pay each other directly in stablecoins using the HTTP 402 status code, now operated as the x402 Foundation under the Linux Foundation. Pay-by-bank, which debits a bank account directly, is also being developed as a card alternative.
What makes it interesting is that both Visa and Mastercard have joined the x402 Foundation. That can be read as evidence that the networks themselves judge non-card paths capable of reaching a scale they cannot ignore.
The other caveat is that the market has not started yet. In a Gartner survey of 322 U.S. consumers in January 2026, willingness to let AI make purchase decisions topped out at 11% even in lower-stakes categories. Narrowing choices fares better, at 31% for household supplies and 28% for personal electronics. People want help shopping while keeping the decision themselves.
In a separate study of 1,009 U.S. consumers by Exploding Topics, 77.6% had used AI to shop in the past six months, yet the most common amount consumers would let AI spend autonomously was $0. The median cap was $50. A wide gap still separates using AI from delegating to it.
The forecast that transactions will increase is a story about what happens after that gap closes.
What merchants should settle first
If agent-initiated transactions do grow, the levers merchants can pull in advance are limited. In order of priority:
Authentication comes first. Agent-initiated traffic often looks bot-like. Leaving rules that block it wholesale will also drop legitimate agent transactions. Agentic Tokens are designed to arrive through the merchant's payment processor, so there is not much to implement in house. Checking whether existing fraud rules are wrongly rejecting agent traffic comes before anything else.
Chargeback readiness is next. As mechanisms like Verifiable Intent spread, proof of whether the shopper intended the purchase will ride along with the transaction data. What merchants get asked for is the evidence on the other side: delivery records, the accuracy of the product information shown at order time, consistency between price and inventory. If what the agent read differs from what arrived, disputes will resolve against the merchant. The Gartner finding that 54% of AI shopping users had to double-check the accuracy of the information reads directly as a product-data quality problem.
On fees there is nothing to work with. Neither the fee structure for agent-initiated transactions nor the treatment of interchange was disclosed in this quarter's results. Investment decisions that assume a cost base cannot be made yet.
Conclusion
A card network can claim it will prevail in the agentic era because it already holds three things: authentication, tokenization and a dispute path. Mastercard's bet is that even when the place where agents choose products changes, the conditions that make a payment trustworthy do not.
What decides that bet is the consumer side, not the technology. Plenty of people will let AI help them shop; few will hand over the wallet. The next things worth watching are how Verifiable Intent gets built into actual chargeback processing, and how much of the merchant-facing terms get disclosed once the BVNK deal closes. When the fees for agent transactions become clear, the weight of this declaration will be clear too.
