Stablecoins, smart agency business, and artificial intelligence in Mastercard's view

Original author: Will A-Wang

Stablecoins and agency services have been the hottest topics of the past two years, but the ones who truly need to make trade-offs on their balance sheets are not stablecoin issuers or agents, but card issuers. Others might see both as business opportunities, but before commercialization, card issuers must first answer a more thorny question: What will I have left if funds no longer flow through my channels?

Mastercard CEO Michael Mibach answered three questions on the same day: how the funds flow, who is buying, and how we make judgments?

Regarding stablecoins, he stated that there are "no problems to solve" in daily consumption; regarding agency services, he believes bank cards will dominate; and regarding artificial intelligence, he believes companies that control proprietary data will be the winners. These three answers seem unrelated, but they actually lead to the same conclusion—settlement can adopt a multi-track system, but trust and security must be guaranteed.

On July 30th, the day the second-quarter earnings report was released, Michael Miebach held a conference call with analysts at 9:00 AM. Afterwards, he conducted a 46-minute interview with Tom Gardner at the Motley Fool podcast studio. One interview was with institutional investors, the other with retail investors; the tone differed slightly, but the conclusion was the same. Three days later, on August 3rd, Mastercard completed its acquisition of BVNK for $1.8 billion, five months ahead of schedule.

In a previous article titled "Stablecoins and Agentic Not Included in Visa's Financial Statements," we analyzed the revenue models of card organizations and Visa's strategic positioning. This article aims to clarify three things from Mastercard's perspective: why the settlement layer made concessions, why the trust and security layers remained uncompromising, and the significance of the $1.8 billion acquisition of BVNK three days later.

1. What does this company actually sell?

Tom Gardner first asked him to clarify the relationship between banks, merchants, and cardholders. Mibach first corrected a figure: not 4 billion cardholders, but 3.7 billion. Then, he put forward a point that many people have never thought about deeply: cardholders are not our customers. The customers are the banks, and also merchants such as Walmart or JPMorgan Chase.

This statement set the stage for everything that followed. It meant that the company never sold the payment system itself, but rather the underlying technology that made payments trustworthy.

In what specific ways is trust manifested?

When shopping, you can choose to leave the website and wait for delivery, or you can simply walk out of the store with your purchases. This is because Mastercard provides payment guarantees, assuring merchants that you can leave with confidence and that they guarantee you will receive your payment. Next, the system performs a four-way transfer: your bank deducts the payment from your card account, and then the funds are transferred to the merchant's bank. This system needs to be built across 220 countries and regions and 3.7 billion cards, with each market having different regulatory rules and infrastructure.

The other side of trust protection is risk control. Even if you make a payment on a fake website, you are still protected because it's not your fault; to prevent fraud at its source, the backend scans trillions of data points in nanoseconds: Is it really possible that you are here right now? Is this expenditure higher than any previous expenditure?

Mibach himself describes it as the operating system of the digital economy. This sounds like public relations rhetoric, but its value lies in how he then breaks it down: the operating system comprises a security layer, a funds transfer layer, and above that, a data and insights layer. The description of the funds transfer layer should be interpreted word for word; it encompasses various payment methods, including stablecoins, inter-account transfers, and bank cards.

The trust and security layer is singular, while the fund transfer layer is plural.

The revenue structure is also trending upwards. In the second quarter, net revenue grew by 12% at constant exchange rates, while value-added services and solutions grew by 18%, 6 percentage points higher than expected, with approximately 60% of that growth related to the internet. Security, identity verification, fraud management, and personalization—these are not payments themselves, but rather the value judgments surrounding payments.

Regarding security, he presented a staggering figure: by 2030, losses from fraud and cybersecurity risks will reach $15.6 trillion; if cybersecurity risks were a country, it would be the world's third-largest economy. The strategic shift he described can be summarized in one sentence: from defense to offense. Based on threat intelligence capabilities built in partnership with Recorded Future, in the first three quarters, we identified over 7 million test credit card transactions in 192 countries, preventing an estimated $172 million in losses.

2. Stablecoins: How do funds flow?

Mibach's most complete statement on stablecoins wasn't in the podcast, but in the opening remarks of a conference call, where he said only one thing: stablecoins have enormous potential, but to really make them work, several essential principles are needed—reliability, security, and interoperability, and that's exactly what Mastercard offers.

He does not deny the existence of stablecoins; he denies that stablecoins currently meet these three conditions.

He then defined the boundaries of the scenario, with the version in the conference call being more straightforward than the version in the podcast: stablecoins have clear utility in certain B2B and P2P transaction flows, but there are no issues to address in P2M transactions.

In the podcast, he further illustrated this point with diagrams. Why would you use stablecoins to buy coffee at the coffee shop downstairs? It's entirely different from a small business paying a supplier in another country through a correspondent banking system, a method that is costly and lacks transparency; you send $100, but don't know that both parties in the transaction take $5, leaving you with only $90.

His principle was very clear: this was never a technical issue, but rather a question of whose problem we could solve. He emphasized that Mastercard has been investing in inter-account payment systems since 2016 and is now one of the largest providers of inter-account payment solutions; bank cards are an important part of the solution, but not the only solution for all payment methods. Today, the network can handle US dollars, any fiat currency, and stablecoins. It not only allows stablecoins to operate within the system but also needs to transfer the protections you expect from bank card payments.

He didn't say stablecoins are unworkable; he said stablecoins haven't yet established a widely accepted network.

Mastercard has spent 60 years building the world's largest acceptance network. Nobody wants a payment solution that only covers a limited number of terminals; people need scale, predictability, and security. These are things that stablecoins currently cannot provide.

During the conference call, he articulated the same point in a more structured way: We envision a diverse world with many currencies and many chains, all of which require a trusted interoperability layer as people will transact across different currencies.

In March, they established a partnership with SoFi; in May, they obtained a BitLicense from the New York State Department of Finance; in June, they expanded their settlement scope to six regulated stablecoins and eight blockchains; previously, they had also launched multi-currency networks MTN, Crypto Credential, and One Credential, the latter of which combines fiat currency and stablecoins.

Despite the intensive efforts, what were the results?

During the conference call, he only provided one growth metric related to digital assets: transaction volume on cryptocurrency co-branded cards has more than tripled in the past two years, with two new cards added this quarter, Bitget and Kraken. This figure is accurate, but the important thing is what it measures—it measures stablecoin spending, and the spending channel remains through cards.

He did not provide specific figures to prove "the number of stablecoins operating on our network." Some analyses point out that Mastercard has never disclosed stablecoin settlement volumes comparable to Visa's, and most projects are likely still in the pilot phase. In contrast, the other party's data is publicly available: as of April this year, Visa's annual stablecoin settlement throughput was $7 billion, a 50% increase year-over-year, covering 9 chains and over 160 stablecoin card projects. Even more noteworthy is the distribution of on-chain card settlement share: Visa accounts for 97%, while Mastercard accounts for only 3%, despite both supporting nearly the same number of projects.

The framework is more refined than its competitors, but has less traffic. He is discussing a methodology for stablecoins, and the evidence he can provide is the growing number of stablecoin holders using bank cards for spending.

3. Agency-type business: Who is buying?

Analyst Ramsey El-Assal posed the best question of the day on the conference call: Are there scenarios in the agent business where stablecoins are absolutely necessary? Or can traditional Mastercard credentials meet all the requirements?

Instead of answering "yes" or "no" directly, Mibach divided the transactions into two categories.

3.1 AI-powered shopping: A bank card is enough.

Let's say you want to go camping, so you ask AI what to bring; it lists 15 items, and since it knows you already have a tent, it won't recommend anything else. But even after receiving the list, you still need to search each website individually, wasting a lot of time. Everything would be incredibly convenient if you could checkout directly, delegate your shopping to an agent, and have Mastercard handle it automatically.

But then the problems arise. Now there's an entity that didn't exist before; how do you know this agent is legitimate and not a fraud? How do you know it's buying what you actually want? If it ordered two grills instead of one, what proof do you have after the bill is deducted?

Mastercard's agent payment solution is to port the system from the existing card organization framework: tokenization, zero-liability protection, and dispute resolution mechanisms. The most crucial element, which he mentioned in the conference call, is the verifiable intent feature. This feature allows users to challenge a transaction and declare that they never intended to purchase the item, thus restarting the chargeback process. He also added a point not mentioned in the podcast: this feature was developed in partnership with Google.

Chargebacks are valuable because processing them is costly. The reason is simple: merchants need to constantly understand and predict user experience, and so do consumers. Therefore, we firmly believe that bank cards will continue to dominate the world today.

He applied the same logic to the enterprise side. In the conference call, he explicitly pointed out that there are numerous agency transactions in the B2B sector—that is, agents making purchases on behalf of companies. These transactions can operate entirely within the credit card ecosystem, following the same logic. The amounts, speed, and purposes are all matched, and they also require the same security and global coverage. In other words, in his view, neither personal AI shopping nor enterprise AI shopping poses a challenge to the credit card pipeline.

Where is the business? Tokenization itself is a service offered by Mastercard. In the second quarter, token penetration just exceeded 40% of all routed transactions, leaving room for growth in the remaining 60%.

But in essence, people won't buy five more tents just because there are agents; it's more about filling the gaps in existing customer traffic.

3.2 Machine Purchase: On-chain authorization, off-chain settlement

The dividing line is drawn here.

Why would a company send invoices when it buys APIs, computing power, data, and content? The ideal model is pay-as-you-go: for example, needing to increase computing power by 10%, then reducing it as needed, paying only for the portion actually used. This immediately improves working capital efficiency. This is exactly what chief procurement officers and CFOs want. But this requires a payment ecosystem that supports always-on, high-frequency, and extremely low-cost payments, a system that currently doesn't exist.

In his opening remarks for the conference call, he summarized the architecture in one sentence:

On-chain permission, off-chain settlement.

Mastercard is the only network that supports machine-to-machine payments.

These eight characters define the division of labor. Authorization takes place on-chain because machines need to immediately recognize transactions; settlement takes place off-chain because the flow of funds is another matter. He expressed the latter more generally during the Q&A session: Settlement will take place on different types of tracks, potentially involving stablecoins or other methods, and we are open to this.

The podcast version is more straightforward:

The underlying track and infrastructure may differ from the bank card track. It could be stablecoins or other types of payment methods. It depends on the company's choice; we remain neutral on this, but the upper-level mechanisms that maintain trust and interoperability are crucial.

The initial ecosystem includes over 30 companies, such as Adyen, Ant International, BVNK, Checkout.com, Coinbase, OKX, and Cloudflare. The inclusion of CDN companies in the payment protocol list indicates that the ecosystem's focus is not on traditional procurement processes, but rather on computing power and API billing.

Therefore, his stance on concessions is very clear. As long as the buyer is a person, whether individual or corporate, credit cards will prevail; only when the buyer becomes a machine will he acknowledge the existence of multiple payment channels, but he still wants to maintain trust and interoperability at the top.

Settlement can be done in multiple ways, but trust and interoperability must be ensured.

4. Artificial Intelligence: How to Make Judgments

Tom Gardner raises a thorny question: Elon Musk has stated that artificial intelligence will surpass human intelligence within five years, and almost all jobs can be done at a lower cost, yet the tech companies with the strongest cash flow and balance sheets are downsizing. What does this mean for consumer spending?

Mibach's answer first focused on the technology itself. This technology certainly deserves in-depth exploration; if used properly, it can lead to prosperity and growth, but it also has drawbacks, as generative artificial intelligence is emboldening fraudsters, scammers, and hackers. The same technology can be used for both attack and defense, thus creating an arms race.

He elaborated on this duality during the conference call. Currently, the CEO and board are most concerned with cybersecurity, and the discussion surrounding cutting-edge models revolves around whether they pose a threat or can be used as tools for identifying vulnerabilities. His answer is both. Mastercard itself uses cutting-edge models to accelerate the identification of internal vulnerabilities and shares these best practices with customers.

He pointed out from a humanistic perspective that this industry, and indeed most industries, ultimately relies on having the best talent, making skills enhancement crucial. He defined the development direction as "human-centered AI applications," that is, using tools to better accomplish tasks, rather than repetitive ones. He gave the example of developing an AI assistant to handle emails; the assistant still scans through them, but handles the trivial parts. He also mentioned that many clients are currently looking to discuss agency business and stablecoins with Mastercard, and the team uses AI to organize publicly available information and prepare for these conversations, saving time and allowing them to focus on what truly matters.

As for why the company isn't afraid, his exact words were that Mastercard has always been technology-centric, not data-centric, because it is essentially an internet company with a relatively lean workforce relative to its market capitalization, yet its business spans 220 countries and regions. Therefore, "our industry doesn't need fundamental changes."

He also cleverly distanced himself from AI transactions: "We neither engage in AI transactions nor build AI infrastructure; we focus on applied AI." This is the context of the $4.9 billion stock buyback in the second quarter: when the AI sector dominated the market, Mastercard became a source of funds for stock sales, with its share price falling from $570 to $470. He stated that this buyback was an opportunistic act; "We are not engaged in stock buyback business; we only do it when the time is right."

His definition of winners in the age of artificial intelligence is the simplest and clearest: those companies that can utilize various models and possess proprietary data to drive these models are the ones that truly stand out. Mastercard possesses one of the most unique datasets—transaction data. He states that it is this data that gives the company its long-term viability and the right to compete.

5. The significance of acquiring BVNK

The acquisition, which was completed three days later, actually concealed the answer within the four-party model.

In traditional fiat currency systems, the final transfer of funds is completed by banks, while Mastercard handles the instructions and clearing logic. In the "you can walk out of the store with your goods" scenario described in Chapter 1, the transfer of funds is jointly completed by your bank and the merchant's bank, who provide guarantees and routing services. Member banks form a fiat currency conduit through which Mastercard does not hold licenses or handle funds—it never controls the settlement.

The problem is that when he claims settlements can be made on blockchain stablecoins, there is no corresponding member banking system on other tracks to do this for him.

The acquisition of BVNK was precisely to fill this gap. He explained in the conference call: With BVNK, Mastercard will act as a trusted interoperability layer, enabling customers to send, receive, store, and convert assets.

5.1 Current Status of Stablecoin Operations in the Market

The transaction was completed on August 3, with a base consideration of $1.5 billion plus a $300 million performance guarantee, five months ahead of the original year-end plan. The transferred assets include: approximately $30 billion in annualized stablecoin trading volume, access to over 130 markets (holding more than 25 licenses), the MiCA authorization obtained in February, and direct access to the SEPA Eurozone via Lithuania. Clients include Worldpay, Deel, Rapyd, Flywire, and Visa Direct.

After being incorporated into Mastercard, it will undertake three specific tasks:

• Provides 24/7 stablecoin settlement services for payment processors and acquiring institutions.

• Integrate stablecoin payment functionality into Mastercard's payment gateway.

• And facilitate the exchange between fiat currency and stablecoins.

What member banks do with fiat currency—receiving payments, settling accounts, and exchanging currencies—BVNK will do the same with stablecoins.

Mibach himself gave a key reason for the acquisition: BVNK's payment orchestration capabilities, licensing portfolio, and connectivity are highly differentiated, and these advantages are now being fully realized in the market. This $1.8 billion wasn't buying technology, but rather time.

In the acquisition press release, Chief Product Officer Jorn Lambert provided the official description: In a multi-currency world where fiat currency, stablecoins, and tokenized deposits coexist, the next generation of payment models will depend on the efficiency of connections between different payment pipelines. In other words, we are not creating payment pipelines, but rather connections. But to create connections, we must first have our own payment pipelines.

5.2 Transaction Details of All Parties

Compared to the two parties involved in the transaction, BVNK's earliest institutional shareholders provided a clearer account of the bidding process. Kjartan Rist, founding partner of Concentric, invested in the company in 2019 at a valuation of $4 million and has not sold a single share in the eight years since. He offered three facts, each more valuable than any press release.

The pressure comes from Stripe. Stripe's $1.1 billion acquisition of Bridge by the end of 2024 will cause a major stir in the industry. Rist stated that Mastercard has deep respect for Stripe, further explaining that this means they are re-evaluating Stripe. He attributes the threat to three points: execution capabilities, product simplicity, and a lack of historical baggage.

Coinbase offered a higher price but ultimately failed due to a lack of cultural fit. Reports indicate that Coinbase offered $2.5 billion, but the founders prioritized compatibility with the acquirer, and Coinbase is an exchange while Mastercard is a financial services company. BVNK also attempted to acquire Mastercard but ultimately failed; Mastercard has been waiting for an opportunity.

Visa, as a board member, chose not to take further action. Visa is both a shareholder and has an observer seat on the board, yet ultimately took no action. Rist interprets this as Visa adopting a different strategy, not directly owning operators but partnering with multiple operators.

Same goal, same goal. One rents a pipeline, the other buys one.

$1.8 billion isn't expensive; during the same period, Mastercard repurchased $4.9 billion worth of stock, with the acquisition cost only slightly more than a third of the repurchase amount. But there's a fundamental difference: repurchased shares can be reissued, while acquired companies cannot be reclaimed.

The BitLicense obtained in May is another sign. Traditionally, credit card institutions do not need to hold licenses; licenses are usually held by member banks. Mastercard has applied to the New York State Department of Financial Services (DFS) for the qualification to clear tokenized deposits and pay stablecoins. In addition, BVNK has obtained 25 licenses and a direct connection to SEPA; this company is blazing a new trail, growing even faster than some fintech companies.

In the stablecoin space, it has no member banks, so it acts as a member bank itself.

in conclusion

Mibach once said something that sounds simple but is actually the whole answer: For decades, consumers’ needs for payments have not changed—simplicity, security, and knowing what will happen if something goes wrong.

This is what credit card companies have been really selling for the past sixty years. It's not about liquidation or routing, but about "who's responsible when problems arise."

Stablecoins cannot provide this answer; there is no on-chain chargeback mechanism. Agents also cannot provide this answer, so they need to submit a verifiable statement of intent to Google. Artificial intelligence also cannot provide this answer; models can make suggestions, but cannot hold anyone accountable.

Therefore, his three answers were essentially the same: you can choose any track, but if something goes wrong, someone must be held responsible, and I will charge the person in charge. He put this system at the layer of trust and security agreements, and then, to ensure that the agreements had a connection point, he bought a new track that he had previously said he didn't need.

But this gamble has an untested premise: will responsibility itself still be as valuable as it was in the past?

Unless otherwise stated, Miebach’s remarks in this article are from the Motley Fool podcast, while remarks marked “Conference Call” are from Mastercard’s Q2 2026 earnings call.

This article does not constitute any investment advice. The stablecoin industry is currently undergoing changes in regulation, competitive landscape, and interest rate environment; any advancement or delay of these changes may render the judgments in this article—including those the author is most confident in—invalid. The calculations in this article are intended to aid understanding of the relevant issues, not as a promise regarding stock price movements. Markets are always far more complex than frameworks suggest; please consider this article as a starting point for thinking, not a basis for action.