Mastercard
Building the payment rails for machines while twenty-one years of litigation renegotiates the price of the old ones
Almost every institution in this series is deploying artificial intelligence inside its own operations. Mastercard is doing something different: it is building the infrastructure that lets AI agents pay for things, and trying to own the trust layer for machine-initiated commerce before anyone else defines it. That is the first of two facts that govern this report. The second is that the economics of the business it already has are being rewritten in a Brooklyn courtroom, in a settlement that would cut interchange, cap rates and abolish a rule Mastercard has defended for two decades. The figures come from Mastercard's filings with the United States Securities and Exchange Commission. Because Mastercard reports on the calendar year, the governing annual report covers 2025 while the operating picture here is the quarter ended 30 September 2026, and both are cited.
$9.8B
Third-quarter net revenue
Up 14 percent, as reported and currency-neutral
59.8%
Operating margin
On operating income of $5.9 billion, up 16 percent
143B
Transactions scored annually
Through the Decision Intelligence platform
$38B
Interchange settlement
Preliminary approval June 2026, final approval still pending
39,800
Employees at 31 December 2025
-46%
Absolute emissions against 2016
While 2025 revenue grew 16 percent
Section 2
The challenge: inventing the next rail while the old one is repriced
Mastercard's position is unusual among the institutions Finatune has assessed. It does not lend, it does not underwrite, and it does not hold customer deposits. It operates a network that authorises, clears and settles transactions between issuers and acquirers, earns a fee on volume, and sells services on top of the data that passes through. The credit card is the visible product, the payment gateway and the switch behind it are the actual business, and the economics are extraordinary: a 59.8 percent operating margin in the third quarter of 2026 on net revenue of 9.8 billion dollars.
That network was built for humans holding plastic, then for humans holding phones. Gross dollar volume rose 7 percent and purchase volume 9 percent in the third quarter, and the digital wallet has already replaced the card as the point of interaction for a large share of those payments. What has not existed until now is a way for software to pay on a person's behalf without the person being present, with the issuer, the merchant and the network all able to tell what just happened and who authorised it. That is the problem Mastercard has spent eighteen months building for.
Two facts from the same quarter define this report. Mastercard signed a strategic partnership with OpenAI covering Agent Pay, payments, joint services growth and AI transformation. And a federal judge granted preliminary approval to a 38 billion dollar settlement that cuts United States credit interchange by 10 basis points for five years, caps standard consumer credit rates at 125 basis points, ends the Honor All Cards rule and expands surcharging rights.
The connection between those two facts is the subject of this report. Mastercard is extending its network into a category that does not yet generate material revenue, at the same moment a court is reducing what the existing category can charge. Tokenization is the hinge: the same mechanism that replaced card numbers with device tokens is what Mastercard is now using to give an AI agent a scoped, revocable credential. Whether the new rail earns enough to matter before the old one is repriced is the question the 2026 filings do not answer.
Section 3 Β· Score: 95 / 100
Technology integration
This is the highest technology score Finatune has awarded, and the reason is categorical rather than incremental. Other institutions in this series are using AI to do their existing work more cheaply. Mastercard is building a payment product for a customer that does not have hands. Agent Pay, announced on 29 April 2025, lets verified AI agents transact on a consumer's behalf using Agentic Tokens, an extension of the Mastercard Digital Enablement Service, each one scoped to a specific agent and a specific commerce policy. An Agentic Token assumes the human pre-approved the agent and is not present at checkout, which is a different trust model from every card transaction that came before it. Agentic AI here is not a feature of the product. It is the counterparty.
The architecture around it is more developed than the headline suggests. The Agent Pay Trust Framework brings together identity, intent, controls, execution and intelligence, which is AI orchestration described as a layered system rather than a slogan. Agent Pay for Machines, launched in June 2026, permissions, orchestrates and settles machine-driven transactions at machine speed, scaling high-frequency, low-latency, low-value payments, down to fractions of a cent, so an AI agent can buy APIs, compute and data as it works. Shopify's Shop Pay already uses Agent Pay, and under the third-quarter partnership OpenAI agreed to use it too. All of it reaches the network through the same API integration layer that issuers and acquirers already connect to, which is the reason Mastercard can ship this at all.
Underneath sits the asset nobody can replicate quickly. Decision Intelligence scores roughly 143 billion transactions annually, and in March 2026 Mastercard added a large tabular foundation model trained on billions of transactions, which is built in-house rather than licensed. The agentic extension of that platform scores transactions on signals that did not exist two years ago: agent identity, session provenance and consent freshness. Decades of machine learning on network data is what makes those signals computable, and risk management on a network of this size has always been a modelling problem rather than a rules problem.
The newest piece is the one Finatune finds most telling. Mastercard has launched an agent initiation probability score that asks not who the agent is but whether a transaction was initiated by an agent at all, currently in testing in the United States, alongside signals indicating whether activity is unusual and whether it requires review. Ann Johnson, an executive vice president, framed the purpose as making agentic commerce trustworthy for the people and businesses whose behalf the systems act on. For a network, AI governance is not an internal policy question. It is the product specification, because an issuer that cannot tell a human from a bot cannot price the risk of either.
Technology snapshot
- βΈAgent Pay, issuing Agentic Tokens through MDES scoped to a specific agent and commerce policy
- βΈAgent Pay for Machines, settling agent and machine payments down to fractions of a cent
- βΈDecision Intelligence scoring roughly 143 billion transactions a year
- βΈA large tabular foundation model trained in-house on billions of transactions
- βΈAn agent initiation probability score, asking whether a buyer is software at all
- βΈA strategic partnership with OpenAI covering Agent Pay and AI transformation
Scorecard
Section 4
Digital transformation timeline
2012 -- The first interchange settlement
Mastercard and the other defendants enter a definitive settlement on the United States MDL cases, agreeing a short-term reduction in default credit interchange rates and modifications to business practices including the no surcharge rule. The court grants final approval in 2013.
2016 -- The climate baseline
The year against which Mastercard's science-based emissions targets are measured, and the starting point for the reductions it reports a decade later.
2018 -- The damages class settlement
The parties to the Damages Class litigation enter a class settlement agreement, resolving the damages claims while the equitable relief claims continue.
2024 -- A settlement rejected
A judge rejects a 30 billion dollar settlement of the merchant interchange litigation, sending the parties back to negotiate.
2025 -- Agent Pay is announced
On 29 April Mastercard introduces Agent Pay, letting verified AI agents transact using Agentic Tokens issued through the Mastercard Digital Enablement Service.
2025 -- A record year on both measures
Full-year net revenue rises 16 percent with value-added services and solutions up 23 percent, while Mastercard surpasses its SBTi-approved 2025 interim emissions targets.
2026 -- The 38 billion dollar settlement
On 9 and 10 June a federal judge grants preliminary approval to a revised 38 billion dollar settlement covering more than 12 million merchants, cutting interchange, capping consumer credit rates and ending the Honor All Cards rule.
2026 -- Agent Pay for Machines
In June Mastercard launches a service allowing agent and machine transactions to be permissioned, orchestrated and settled at machine speed, in payments as small as fractions of a cent.
2026 -- The OpenAI partnership
On the third-quarter call on 29 October, Mastercard discloses a broad strategic partnership with OpenAI covering Agent Pay, payments, joint services growth and AI transformation, alongside the launch of Agentic Trust Intelligence.
Section 5 Β· Score: 93 / 100
Financial performance
Third-quarter 2026 net revenue reached 9.807 billion dollars against 8.602 billion a year earlier, an increase of 14 percent both as reported and on a currency-neutral basis. Net income was 4.628 billion dollars against 3.927 billion, up 18 percent, with diluted earnings per share of 5.28 dollars against 4.34, up 22 percent. Adjusted net income was 4.7 billion dollars and adjusted diluted earnings per share 5.31 dollars, against consensus of 5.16 on revenue consensus of 9.65 billion. Operating income reached 5.9 billion dollars, up 16 percent, at an operating margin of 59.8 percent.
The nine-month figures show the same shape without the noise. Net revenue of 27.482 billion dollars against 23.985 billion, and net income of 12.898 billion against 10.908 billion. Across four consecutive quarters net revenue has grown between 14 and 18 percent, and value-added services and solutions grew 19 percent in the third quarter, outpacing the core network. That last number is the one that matters most for the next five years, because it is the part of the business a court cannot reprice. Cash flow conversion at these margins is close to total.
$5.28
Diluted EPS, up 22 percent
+19%
Value-added services revenue, outpacing the network
$27.5B
Nine-month net revenue
Against $24.0 billion a year earlier
$95.2B
Treasury stock at cost
Against retained earnings of $95.6 billion
One balance sheet line deserves attention because it describes the capital strategy better than any ratio. Class A treasury stock stood at 95.2 billion dollars at cost against 83.2 billion a year earlier, while retained earnings were 95.6 billion against 85.0 billion. Mastercard has repurchased very nearly as much stock as it has ever retained in earnings. For a company with no lending book and minimal capital requirements, that is the rational use of equity, and it is why market capitalisation has compounded faster than revenue for most of a decade.
The reservation is not about execution, which is close to flawless, but about what happens to the growth rate when the settlement takes effect. A 10 basis point reduction in United States credit interchange for five years and a 125 basis point cap on standard consumer credit rates do not fall on Mastercard's revenue directly, since interchange flows to issuers rather than to the network. They fall on the economics of the issuers who choose which network to put on a card, and on the merchant pricing that drives acceptance. The second-order effect is real and the filings do not quantify it.
Section 6 Β· Score: 84 / 100
Workforce and human capital
After three consecutive reports describing shrinking workforces, this one is the opposite. Mastercard employed approximately 39,800 people globally at 31 December 2025, of whom approximately 70 percent were outside the United States across more than 90 countries, with an employee base that is predominantly full-time, supplemented by approximately 6,000 contingent workers. Voluntary attrition was approximately 6 percent, and the total cost of the workforce for 2025 was 7.3 billion dollars.
Six percent voluntary attrition is the number that earns this score. It is exceptionally low for a technology-intensive employer competing for the same engineers as the large platforms, and it is disclosed rather than inferred. The 7.3 billion dollar workforce cost is also disclosed in absolute terms, which almost no institution in this series does, and against 2025 net revenue it represents a company that spends roughly a fifth of its revenue on people and keeps them.
Mastercard is the first institution in this series whose AI programme implies hiring rather than attrition. Building a trust framework, a foundation model and a machine payments rail is engineering work, and the company describes acquisitions as a source of talent with differentiated skills rather than a route to headcount reduction.
The reservation is one of disclosure rather than substance. Mastercard reports its workforce cost, its attrition and its geographic spread, but not the split between engineering and commercial roles, nor how much of the 39,800 sits in the services business that is now growing faster than the network. For a company whose argument is that it is becoming a technology and services firm rather than a card scheme, that split would be the most informative number it could publish. The score reflects a genuinely well-run and well-disclosed workforce, held below the high eighties by the absence of that breakdown.
Section 7 Β· Score: 76 / 100
Regulatory compliance
This is the lowest score in the report and the only one that is unambiguously a weakness. Mastercard's interchange fees and other practices are subject to regulatory and legal review or challenge in a number of jurisdictions. The core allegation, pursued since 2005, is that Mastercard, Visa and certain financial institutions conspired to set the price of interchange fees, enacted point-of-sale acceptance rules including the no surcharge rule in violation of antitrust laws, and engaged in unlawful tying and bundling, resulting in merchants paying excessive costs for acceptance. Regulatory compliance for a payment network means defending the rules that constitute the network itself.
The litigation is now at its most consequential point in twenty-one years. A judge rejected a 30 billion dollar settlement in 2024. The revised 38 billion dollar settlement received preliminary approval on 9 and 10 June 2026, with the court finding it a reasonable recovery given the risks, costs and delay of continued litigation. It covers more than 12 million merchants, cuts United States credit interchange by 10 basis points for five years, caps standard consumer credit rates at 125 basis points, ends the Honor All Cards rule and expands surcharging rights. It is not final. Merchant objectors argue it still fails to eliminate key restraints including the Honor All Issuers and No-Bypass rules, and no claims process yet exists.
21 years
Duration of the interchange litigation
The settlement is aimed at ending it
12M+
Merchants covered
By the settlement granted preliminary approval in June 2026
$5.17B
Damages settlement fund at 30 June
After an initial distribution of $418.9 million
The separate damages settlement shows how long this takes to resolve even once agreed. The fund held 5.17 billion dollars at 30 June following an initial distribution of 418.9 million dollars to merchants, with roughly 599,000 claims in that first distribution and about 40,000 merchants awaiting a special master's review. In April 2026 Mastercard and Visa settled the final claims of around 65 individual retailers, including Amtrak, Crate and Barrel, Dick's Sporting Goods and Nike, in a case set for trial that month, and a further group led by GrubHub remained in discussions. A score of 76 reflects no prudential enforcement, no consent orders and no finding of wrongdoing, against two decades of antitrust exposure whose structural remedies are about to take effect.
Section 8 Β· Score: 86 / 100
ESG and sustainability
This is the strongest environmental record Finatune has assessed, and unlike the exchange that operates the carbon market, Mastercard earns it on its own footprint rather than its market position. Its net zero target for 2040 is validated by the Science Based Targets initiative and covers Scope 1, Scope 2 and Scope 3 emissions across the entire value chain, a decade ahead of the 2050 date most of this series uses. In 2025 the company achieved and surpassed its SBTi-approved 2025 interim targets, which were a 38 percent reduction in Scope 1 and 2 emissions and a 20 percent reduction in Scope 3 against a 2016 base year.
The decoupling is the part that is hard to argue with. Mastercard has reduced absolute total greenhouse gas emissions by 46 percent against 2016 while growing into a far larger company, and in 2025 it cut total emissions a further 1 percent while net revenue grew 16 percent, the third consecutive year of falling emissions alongside an expanding business. All company-owned and operated facilities worldwide run on renewable electricity, achieved through onsite generation, renewable energy certificates and clean energy investments, with further solar capacity secured adjacent to its St. Louis data centre operations.
2040
SBTi-validated net zero target
Covering scopes 1, 2 and 3 across the value chain
3 years
Consecutive years of falling emissions
While revenue grew 16 percent in 2025 alone
71%
Of top-emitting suppliers on science-based targets
Against a 65 to 75 percent target band
What separates this from the rest of the series is that the targets are wired into pay. The proxy discloses an environmental, social and governance modifier that can adjust the executive financial score upward or downward within a range of 10 percentage points, with the metrics, target bands and actual results published: a 44 to 55 percent band for the Scope 1 and 2 reduction against 46 percent actual, and a 65 to 75 percent band for suppliers representing the top 85 percent of supply chain emissions having set or committed to science-based targets against 71 percent actual. Finatune has not previously seen a company in this series publish the band, the outcome and the compensation consequence together. The score is 86 rather than higher because the hardest part, Scope 3, is still fourteen years from its deadline and most of the reduction to date has come from operations and electricity.
Section 9 Β· Score: 84 / 100
SDG alignment
Mastercard's development alignment is the most measurable in this series because the company counted it. It connected one billion people to the digital economy, a commitment it set, tracked and reported as completed, and the number of individuals newly connected is one of the metrics in the executive compensation framework rather than a figure in a brochure. For a payments network, the development contribution and the commercial strategy are the same activity: a person without a payment credential is both excluded and not a customer.
Select a goal to view supporting evidence
The limit is the same one that applies to any inclusion claim made by the party doing the including. Connecting a billion people to the digital economy means issuing them credentials that generate transaction fees, and the company does not disclose what those users cost or what they earn. The alignment is real and counted, which is more than most, and it is not disinterested. The score of 84 reflects measurement and scale rather than altruism.
Section 10 Β· Score: 92 / 100
Competitive positioning
Mastercard is the smaller half of a duopoly, and being the smaller half of this particular duopoly is one of the best competitive positions in global finance. Two networks carry the overwhelming majority of card-based commerce outside China, both earn around 60 percent operating margins, and neither can be displaced by capital because what they own is acceptance on both sides at once. Mastercard's advantage over Visa is not scale, which it does not have, but growth rate: net revenue up 14 percent in the third quarter with value-added services up 19 percent.
The services business is where the position is being widened rather than defended. Mastercard sells threat intelligence, economic analysis, cyber security, loyalty programming, fraud prevention and open finance capability on top of the network, and its open banking and embedded finance businesses reach institutions that are not card issuers at all. That matters because every one of those revenue lines is priced on value delivered rather than on a basis point of interchange, which is precisely the exposure the settlement constrains.
At the edges the company is buying optionality in currencies and rails it does not control. The new Global Money Platform arrives through the BVNK acquisition, stablecoin capability has been expanding since early 2026, and the cross-border and remittance business is where a network with 90 country coverage has structural advantages over any single-market challenger. The honest reservation is that Visa is doing most of these things too, on a larger base, and that the two companies' strategic disclosures have rarely been more similar than they are now.
Section 11 Β· Score: 89 / 100
Brand and reputation
The interlocking circles are among the most recognised marks in the world, and the Priceless campaign is one of the longest-running pieces of brand positioning in financial services. The proxy describes the promise of reliability and protection that the interlocking circles represent to people everywhere, which is a fair summary of what the brand actually does commercially: it tells a consumer in an unfamiliar country that a transaction will work and that they will be made whole if it does not.
The reputational exposure is structural and almost invisible to consumers. Mastercard has spent twenty-one years as a defendant in litigation brought by the merchants who accept it, over the fees they pay to do so. Very few cardholders know this, and that asymmetry is itself the asset: the brand is loved by the side that does not pay it and litigated by the side that does. Honor All Cards ending and surcharging rights expanding will make the fee visible at the till for the first time in many markets, which is the first real test of whether consumer goodwill survives contact with the price.
Priceless
The campaign behind the brand
One of the longest-running positions in financial services
90+
Countries with employees
With 70 percent of staff outside the United States
Shop Pay
First major merchant platform on Agent Pay
With OpenAI added under the third-quarter partnership
21 years
As an antitrust defendant
Brought by the merchants who accept the brand
Section 12 Β· Synthesis
The Finatune lens
Mastercard scores 95 on technology, the highest Finatune has awarded, and the margin over the rest of the series is categorical rather than a matter of degree. Every other institution assessed is applying AI to work it already does. Mastercard has built a payment product whose customer is software: Agentic Tokens scoped to a named agent and a named policy, a trust framework covering identity, intent, controls, execution and intelligence, a machine payments rail settling fractions of a cent, a tabular foundation model trained in-house on billions of transactions, and an agent initiation probability score that asks a question no network had needed to ask. The OpenAI partnership is the external validation that this is the standard others will build against.
Financial performance at 93 reflects a business that is close to the theoretical ideal. Net revenue up 14 percent, net income up 18, diluted earnings per share up 22, a 59.8 percent operating margin, four consecutive quarters of double-digit revenue growth, and services growing faster than the network. The capital strategy is equally clear: treasury stock at cost of 95.2 billion dollars against retained earnings of 95.6 billion, which is a company that has returned nearly everything it has ever earned because it has nothing capital-intensive to do with it.
The two halves of this report point in opposite directions and both are true. Mastercard is building the rails for machine commerce and has signed the most significant AI partner in the market to use them. And a court is about to cut interchange for five years, cap consumer credit rates and abolish Honor All Cards, in a settlement ending twenty-one years of litigation brought by the merchants who accept the brand. One is a new business that earns almost nothing yet. The other is the business that earns everything.
ESG at 86 is also a series high, and it is the one score Finatune would defend most readily, because the evidence is third-party validated rather than self-reported. An SBTi-validated 2040 net zero target covering Scope 3, interim targets surpassed rather than withdrawn or missed, a 46 percent absolute reduction against 2016, three consecutive years of falling emissions while revenue grew 16 percent, and the target band, the actual outcome and the compensation consequence all published together in the proxy. Compare that to the bank in this series that discontinued its financed emissions targets, and the fourteen point gap in the scores understates the difference in posture.
Regulatory at 76 is the counterweight and it is not a footnote. Mastercard has no consent orders and no prudential enforcement, which is better than every bank assessed here. It has also been an antitrust defendant since 2005, had a 30 billion dollar settlement rejected by a judge, and is now a party to a 38 billion dollar settlement that is not final, that merchant groups are objecting to, and whose structural remedies change rules the network has defended for two decades. The damages fund has distributed 418.9 million dollars of 5.17 billion with 40,000 claims still in dispute. This is what it looks like when the regulatory risk is not misconduct but the business model itself.
What decides the next three years is whether services and agentic commerce grow faster than interchange economics compress. The services line, up 19 percent, is already the faster half and is priced on value rather than basis points, which is the right exposure to have. Agent Pay is strategically early and commercially immaterial, and will stay immaterial until agents transact at volume, which nobody can date. If both work, Mastercard stops being a card scheme and becomes a network and services company whose pricing is not set in court. If neither moves fast enough, it is a magnificent toll road whose toll was just lowered by a judge for five years.
Finatune scorecard
Sources and disclosures
Where these numbers come from
Every financial and operational figure in this report comes from the filings listed below, except where identified otherwise in the text. Because Mastercard reports on the calendar year, both annual and quarterly sources are cited, and the period is identified wherever a figure appears. Emissions data, the status of the science-based targets, product announcements made outside the filings, and figures reported by the court-appointed settlement administrator are identified at the point of use.
- Annual Report on Form 10-K
For the fiscal year ended 31 December 2025, certified 11 February 2026.
- Quarterly Report on Form 10-Q
For the quarterly period ended 30 September 2026, filed 29 October 2026.
- Third Quarter 2026 Results on Form 8-K
Earnings release furnished 29 October 2026 for the quarter ended 30 September 2026.
- Proxy Statement on Form DEF 14A
Dated 27 April 2026, with compensation and ESG modifier disclosures covering fiscal year 2025.
Resources
Downloads and Tools
Executive Summary PDF
A two page synthesis of the key findings from this report, formatted for circulation to an investment committee or executive team.
Finatune Assessment Framework Template
Apply the same nine dimension assessment framework to any financial institution, with the scoring rubric and evidence requirements for each dimension.
Related content
Continue across Finatune
Credit cards
The credit card category in Finatune's Financial Services module, covering the products that carry Mastercard's brand to consumers.
Remittance and cross-border transfer
Our coverage of cross-border payment products, the category Mastercard's new Global Money Platform is built for.
Accounting and payments MCP servers
The MCP servers covering payments, the interface layer through which AI systems reach payment rails, and the closest category to Agent Pay for Machines.
Compliance and RegTech agents
The compliance and monitoring agents Finatune covers, against which to read Decision Intelligence and the agentic trust signals.
Financial data platforms
The market and financial data platforms covered in our Data and BI module, where Mastercard's value-added services business competes.
Finance glossary
Definitions for the payment terms used in this report, from tokenization to payment gateway.
This report is an independent assessment prepared by Finatune for informational purposes only. It is not investment advice, and not a recommendation to buy, sell or hold any security. Finatune has no commercial relationship with Mastercard Incorporated. Financial figures are drawn from Mastercard's filings with the United States Securities and Exchange Commission. Because Mastercard reports on the calendar year, annual figures are stated as of the fiscal year ended 31 December 2025 and quarterly figures as of the period ended 30 September 2026, as indicated at each point of use. The interchange settlement described in Section 7 received preliminary approval in June 2026 and has not received final approval; its terms may change and objections remain pending, and nothing here should be read as a prediction of its outcome. Emissions figures and the status of science-based targets are drawn from Mastercard's own impact reporting and from the Science Based Targets initiative, and are identified as such in the text. Scores are Finatune's own judgement, applied on a consistent framework to every institution in this series, and reasonable analysts will dispute some of them.