TechFlow Logo
Login/ Sign up
ETH Gas
Gwei
Fear
gas
Stablecoins, Agent Commerce, and AI Through Mastercard's Lens

Stablecoins, Agent Commerce, and AI Through Mastercard's Lens

2026.08.21
Share
From Mastercard’s perspective, three things become clear: why the settlement layer gave ground, why the trust and security layer refused to budge, and the significance of acquiring BVNK for $1.8 billion just three days later.

Author: Will Awang

Stablecoins and agent-driven commerce have been the most discussed topics over the past two years, but it’s not the stablecoin issuers nor the agent developers who ultimately have to make trade-offs on their balance sheets. It’s the card networks. Others can treat both as pure commercial opportunities, but before card networks can monetize them, they must first answer a harder question: If money no longer flows through my rails, what do I have left?

Mastercard CEO Michael Miebach addressed this on the same day by answering three questions: How does money move, who is buying, and how are decisions made.

On stablecoins, he said there's "no problem to solve" for everyday consumer spending; on agent commerce, he said cards will prevail; on AI, he said the winners will be companies that hold proprietary data. Though these three answers seem unrelated, they all execute the same strategic maneuver: settlement can be multi-rail, but trust and security must be secured.

On July 30, the day of the second-quarter earnings report, Miebach held a conference call with analysts at 9 a.m. He then sat down in a Motley Fool podcast studio and spoke with Tom Gardner for forty-six minutes. One audience was institutional, the other retail; the wording differed slightly, but the underlying judgment was identical. Three days later, on August 3, Mastercard completed its acquisition of BVNK for $1.8 billion, five months ahead of schedule.

In our previous article on stablecoins and agentic commerce, which did not make it into Visa’s financial reports, we analyzed the card network’s revenue model and Visa’s strategic positioning. This article seeks to clarify three things from Mastercard’s perspective: why the settlement layer stepped back, why the trust and security layer drew the line, and the significance of the $1.8 billion acquisition of BVNK three days later.

I. What Is This Company Actually Selling?

Tom Gardner kicked off by asking him to lay out the stage, clarifying the relationship between banks, merchants, and cardholders. Miebach first corrected a figure: it’s not 4 billion cardholders, but 3.7 billion. Then he delivered a sentence many had never really considered: cardholders are not our customers. Our customers are the banks, or potentially the merchants—could be Walmart, could be JPMorgan Chase.

That statement dictates everything that follows. It means this company has never sold payments themselves, but rather the layer that makes payments trustworthy.

What does trust actually look like?

When you go shopping, you can leave a website waiting for your delivery, or simply walk out of a store with your bags. You’re able to do so because behind the scenes sits a payment guarantee issued by Mastercard, telling the merchant: let this person go, we guarantee you’ll receive the funds. The four-party network transfer happens afterward, where your bank pulls money from your card account and routes it to the merchant’s bank. This system operates across 220 countries and regions and covers 3.7 billion cards, each with distinct regulatory frameworks and infrastructure.

The flip side of that trust guarantee is risk management. If you pay on a fake website, you’re still protected because it wasn’t your fault. To stop fraud at the source, backend systems scan trillions of data points in nanoseconds: could you genuinely be in this location right now? Does this expenditure exceed anything you’ve ever done before?

Miebach himself calls it the operating system of the digital economy. While that sounds like PR speak, its real value lies in how he unpacks it: an operating system has a security layer, a fund-flow layer, and above that, data and insights. But the description of the fund-flow layer demands careful reading—it spans stablecoins, account-to-account transfers, and cards.

The trust and security layer is singular; the fund-flow layer is plural.

The revenue structure is also shifting upward. Second-quarter net income grew 12% on a currency-neutral basis, while value-added services and solutions grew 18%, outpacing the top line by six percentage points, with roughly 60% tied to the network. Security, identity, fraud management, and personalization aren’t payments themselves; they’re decision-making capabilities sold around payments.

On the security side, he provided a magnitude: by 2030, losses from fraud and cyber risk will reach $15.6 trillion. If cyber risk were a country, it would be the third-largest economy globally. His stated strategic shift boils down to one phrase: moving from defense to offense. Built on threat intelligence powered by Recorded Future, his team identified over 7 million test-card transactions across 192 countries in the first three quarters, estimating $172 million in avoided losses.

II. Stablecoins: How Money Moves

Miebach’s most complete framing of stablecoins isn’t in the podcast but in the opening remarks of that same-day conference call, condensed into a single point: stablecoins have massive potential, but to truly scale, they require certain foundational principles—reliability, security, interoperability—and that’s exactly what Mastercard delivers.

He doesn’t deny stablecoins; he denies that they currently possess those three conditions.

He immediately drew boundaries around use cases, with the conference call version being sharper than the podcast take: stablecoins have clear utility in certain B2B and P2P transfers, but there’s no problem to solve in P2M (person-to-merchant) transactions.

In the podcast, he fleshed that out visually. Why would you use a stablecoin to buy coffee at the corner café? Flip it to a small business paying a supplier in another country via correspondent banking: high fees, poor transparency, you send $100, unaware that two parties skim $5 each along the way, leaving only $90 to arrive.

His principle is straightforward: this was never about technology, but about whose problems we can solve. He emphasized that Mastercard has invested in account-to-account systems since 2016 and is today one of the largest A2A solution providers. Cards represent a major part of the answer, but not the answer for every payment type. Today, this network can handle dollars, any fiat currency, and even stablecoins—not just by running stablecoins through the system, but by carrying over the protections consumers expect from card payments.

He didn’t say stablecoins don’t work; he said they haven’t yet grown an acquiring network.

Because Mastercard spent sixty years building the world’s largest acquiring network. No one wants a payment solution that reaches only a tiny fraction of endpoints. You want scale, predictability, and protection. Those are things stablecoins simply don’t provide right now.

During the conference call, he structured the same idea more formally: We anticipate a poly-currency world with many tokens and many chains, and all of them will need a trusted interoperability layer because people will transact across different currencies.

Partnerships began with SoFi in March, followed by obtaining New York DFS’s BitLicense in May, and expanding settlement to six regulated stablecoins and eight chains in June. Even earlier came the multi-currency network (MTN), Crypto Credential, and One Credential, which consolidates fiat and stablecoin balances into a single credential.

Despite these rapid moves, where’s the track record?

On the call, he offered only one growth metric tied to digital assets: transaction volume for crypto co-branded cards has more than tripled over the past two years, with new partnerships from Bitget and Kraken added this quarter. The number is real, but it’s crucial to understand what it measures—it tracks how many stablecoins are being spent, and the spending path remains the card itself.

As for the metric that would prove “stablecoins are running on our network,” he withheld it. Analysts pointedly noted that Mastercard has never disclosed stablecoin settlement volumes comparable to Visa’s, suggesting most efforts may still be in pilot phases. Meanwhile, the opposing figures are public: as of April this year, Visa’s stablecoin settlement run rate hit $7 billion annually, up 50% sequentially, spanning nine chains and over 160 stablecoin card programs. More stark is the on-chain card settlement share distribution: Visa commands 97%, Mastercard 3%, despite both supporting nearly equal project counts.

He built a more complete framework than his competitor, but captured less traffic. He’s pitching a methodology for stablecoins, and the evidence he can produce is the rising spend volume from stablecoin holders using cards.

III. Agent Commerce: Who Is Buying?

Analyst Ramsey El-Assal asked the sharpest question of the day on the call: Are there scenarios in agent commerce that absolutely require stablecoins, or can traditional Mastercard credentials cover everything?

Miebach didn’t give a straight yes or no; instead, he split transactions into two categories.

3.1 AI Shopping on Behalf of Humans: Cards Are Sufficient

Imagine you’re planning a camping trip and ask an AI what you should pack. It gives you fifteen items, skipping the tent because it knows you already own one. But after getting the list, you still have to hunt through one website after another until your afternoon is gone. If you could check out instantly at that moment, delegating the purchase to an agent running on a Mastercard behind the scenes, everything clicks into place.

But the problems follow. There’s now an intermediary entity that didn’t exist before. How do you verify this agent is known and not spoofed? How do you ensure it’s buying what you actually wanted? If it orders two grills instead of one, what do you present when the charge posts?

Mastercard Agent Pay’s answer is to migrate existing card system utilities forward: tokenization, zero-liability protection, and dispute resolution. The most critical component, highlighted during the call, is Verifiable Intent. It lets you challenge a transaction by declaring you never intended to buy it, triggering a fresh chargeback process. He added a detail omitted from the podcast: this feature was developed in partnership with Google.

Chargebacks hold value precisely because resolving them is expensive. The reasoning is quite simple: merchants will always need access and predictable user experiences, and so will consumers. That’s why we truly believe cards will win in that world.

He applied the same logic to the enterprise side. Clearly stating on the call, he noted a cohort of B2B agent transactions where an agent handles corporate procurement, all of which can easily run on the card ecosystem under identical logic. Amounts, speed, and use cases align perfectly, requiring the same protections and global reach. In his view, neither AI buying on behalf of individuals nor AI procuring for companies poses a threat to the card rails.

So where’s the business? Tokenization itself is a service Mastercard sells, and with token penetration just crossing 40% of total switching transactions in Q2, the remaining 60% represents clear runway.

Ultimately, people won’t buy five extra tents just because an agent exists. This is largely a substitution of existing traffic.

3.2 Machines Buying: On-Chain Authorization, Off-Chain Settlement

The line is drawn here.

A company buys digitized assets like APIs, compute power, data, and content—why mail a physical invoice? You’d prefer usage-based pricing: dial up compute by 10%, then down again, paying exactly for what you consume, immediately boosting working capital efficiency. That’s what a CPO wants, and what a CFO wants. But it requires a payment ecosystem capable of handling always-on, high-frequency, fractional-dollar transactions. Such a payment layer simply doesn’t exist today.

In the opening remarks of the conference call, he summarized this architecture in a single sentence:

on-chain permissioning, off-chain settlement.
Mastercard is the only network supporting machine-to-machine payments.

Eight characters perfectly delineate the division of labor. Authorization happens on-chain because machines need instant transaction validation. Settlement occurs off-chain because choosing the monetary route is a separate matter. In the Q&A, he softened the second half: settlement will occur across various types of rails, potentially involving stablecoins or alternative methods. We remain highly open to this.

The podcast version was even more direct:

The underlying rails and infrastructure will likely differ from card rails. It could be stablecoins or something else. That depends entirely on corporate choice. We remain highly neutral, but the layer above—protocols that maintain trust and interoperability—is key.

The launch ecosystem includes over thirty partners, naming Adyen, Ant International, BVNK, Checkout.com, Coinbase, OKX, and Cloudflare. A CDN provider appearing on a payment protocol launch list signals that this track isn’t targeting traditional procurement workflows, but rather compute and API billing.

Consequently, the precise boundary of his concession becomes clear. As long as humans—whether individual or corporate—are doing the buying, cards will prevail. Only when buyers become machines does he acknowledge non-card rails, but he insists on maintaining the upper-layer trust and interoperability.

Settlement can be multi-rail; trust and interoperability must be won.

IV. AI: How to Make Judgments

Tom Gardner posed a heavy question: Musk claims AI will surpass the sum of human intelligence within five years, allowing nearly all work to be done cheaper, while tech giants with the strongest cash flows and balance sheets in history are reducing headcount. What does this mean for consumer spending?

Miebach’s answer initially lands on technology itself. It’s an uncharted territory; when deployed correctly, it can drive paths toward prosperity and growth, but it equally carries downside risks, as generative AI empowers scammers, fraudsters, and hackers. The same technology used for attack can be wielded for defense, making it an arms race.

He elaborated on this duality more concretely during the call. Cybersecurity currently dominates CEO and boardroom conversations. Debates surrounding frontier models hinge precisely on whether they constitute threats or tools that help uncover your own vulnerabilities. His answer: both. Mastercard itself uses frontier models to accelerate internal vulnerability detection, sharing this practice as a best practice with clients.

At the human level, he argued that this industry, and indeed most industries, ultimately hinges on possessing the best talent, which means upskilling workers continuously. He defined the direction as “human-centric AI applications”—using tools to enhance work, eliminating repetitive tasks. He cited his own example: he built an AI assistant for processing emails. He still reviews outputs, but the mundane parts are automated. He also noted that numerous clients currently seek discussions on agent commerce and stablecoins. The team leverages AI to compile public materials for preparation, freeing up time to focus on what truly matters.

Regarding why this company isn’t afraid, his exact words were that Mastercard has always been about technology, not headcount. Because it’s fundamentally a network company, its workforce is exceptionally light relative to market capitalization, yet it operates across 220 countries and regions. Thus, “our industry isn’t one that requires fundamental reinvention.”

He also conveniently distanced himself from the AI trade: “We’re not in the AI trade business, nor AI infrastructure. We do applied AI.” This context explains the $4.9 billion share repurchase program executed in Q2. When AI sectors dominated market attention, Mastercard became a liquidity sink as investors dumped shares to chase AI stocks, pushing the price from $570 down to $470. He described the buyback as opportunistic: “We don’t do buybacks as a business model; we only execute them when it makes sense.”

His definition of winners in the AI era proved the most concise remark in the session: “Companies that stand out will be those capable of utilizing diverse models while possessing proprietary data to feed them.” Mastercard holds one of the most unique datasets available: transaction data. He stated this guarantees the company’s long-term vitality and seat at the table.

V. The Significance of Acquiring BVNK

The answer to that three-day-later closing acquisition actually resides within the four-party model.

In traditional fiat systems, final fund transfers are executed by banks, while Mastercard handles instructions and clearing logic. In the chapter-one scenario of “walking out of a store with your bags,” your bank and the merchant’s bank move the money; Mastercard provides the guarantee and routing. Member banks form a fiat rail, and on that rail, Mastercard doesn’t need its own licenses or touch capital outright—it has never controlled settlement.

The problem arises when he states settlement might occur on blockchain stablecoins: the alternative rail lacks a corresponding member bank framework. No one else performs that function on his behalf.

BVNK was acquired to fill that exact gap. His positioning on the call was clear: with BVNK, Mastercard will serve as that trusted interoperability layer, enabling clients to send, receive, store, and convert assets.

5.1 A Stablecoin Rail Already Operating in the Market Today

Closing on August 3 for a $1.5 billion base consideration plus a $300 million performance earnout, five months ahead of the originally scheduled year-end timeline. The asset portfolio brought aboard includes approximately $30 billion in annualized stablecoin transaction volume, over 25 licenses across 130 markets, MiCA authorization secured in February, and direct integration into the SEPA euro rail via Lithuania. The client roster features Worldpay, Deel, Rapyd, Flywire, and Visa Direct.

Integrated into Mastercard, it undertakes three specific functions:

  1. Providing 24/7 stablecoin settlement for processors and acquirers,
  2. Adding stablecoin checkout capabilities to Mastercard’s payment gateway,
  3. And facilitating fiat-to-stablecoin exchanges.

Member banks perform those very functions on the fiat side—collection, settlement, FX—which BVNK replicates on the stablecoin side.

Among Miebach’s own rationale for the acquisition, one line stands out: “BVNK’s payment orchestration capabilities, license portfolio, and connectivity are highly differentiated, and it’s already live in the market.” That is exactly what $1.8 billion purchased—not technology, but time.

5.2 Deal Details Across Stakeholders

The bidding process for this deal was explained more clearly by BVNK’s earliest institutional shareholders than by either party involved. Concentric founding partner Kjartan Rist entered at a $4 million valuation in 2019 and hasn’t sold a single share in eight years. The three facts he presented each carry more weight than any press release.

The pressure originated from Stripe. Late last year, Stripe acquired Bridge for $1.1 billion, letting the cat among the pigeons. Rist noted Mastercard holds Stripe in the highest regard, then translated that sentiment himself: it’s code for saying they’re watching intently. He broke the threat down into three components: execution capability, product simplicity, and lack of historical baggage.

Coinbase offered a higher price but lost on culture. Reports indicate Coinbase briefly pushed to $2.5 billion, but founders prioritize chemistry when selecting acquirers. Coinbase is an exchange; Mastercard is a financial services firm. BVNK tested the waters with Coinbase unsuccessfully, while Mastercard remained patiently waiting on the porch.

Visa sits on the board but chose not to compete. Holding both equity and a board observer seat, Visa ultimately refrained from bidding. Rist interpreted this as Visa pursuing a different strategy: refusing to own operators, preferring collaboration across multiple ones.

Same asset, same negotiating table. One rents pipes; the other buys them.

$1.8 billion isn’t cheap, yet within the same quarter, Mastercard repurchased $4.9 billion of its own stock—the acquisition cost amounted to just over a third of that buyback. But there’s a fundamental difference: shares bought back can be reissued, while a company acquired cannot be returned.

The May BitLicense serves as another signal. Card networks traditionally never needed their own licenses; regulators held them within member banks. Yet Mastercard independently applied to New York DFS for clearance to settle tokenized deposits and payment stablecoins. Coupled with the 25 licenses brought by BVNK and direct SEPA integration, this company is actively constructing new rails, progressing even faster than many fintechs.

On this stablecoin rail, lacking member banks, it steps in to act as its own member bank.

Conclusion

Miebach once uttered a seemingly plain statement that actually contained the entire answer: What consumers want from payments hasn’t changed in decades—it’s simplicity, safety, and knowing exactly what happens when things go wrong.

That is precisely what card networks have been selling for sixty years. Not clearing, not switching, but “who bears responsibility when problems arise.”

Stablecoins can’t deliver this answer because there are no chargebacks on-chain. Agents can’t either, which is why they partnered with Google to build verifiable intent. AI fails here too; models can recommend, but they can’t assign liability.

His three answers thus converge into one: pick whichever rail you want, but someone must take accountability when failures occur, and I will charge that accountable party. He bets this entire proposition on the trust and security protocol layer, then purchases a brand-new rail—one he previously claimed unnecessary—just to give that protocol something to connect to.

But this wager rests on a premise that has never been stress-tested: will accountability itself retain its former value?