The value can be real. Everyone can still wait.

By 2015, the United States was one of the last developed card markets to adopt chip cards at scale. The technology was not new. Other countries had used it for years, and its purpose was well understood: instead of relying on reusable data stored on a magnetic stripe, the chip generated authentication data for an individual transaction, making copied card data substantially harder to use at the physical point of sale.¹

The value of the chip was credible. Counterfeit card fraud in the United States was valued at $3.05 billion in 2015 and represented 44 per cent of card fraud by value. Yet only about 2 per cent of in-person general-purpose card payments used chips that year.¹

The problem was not whether a more secure transaction was possible. It was how to make cards, terminals and processing capability move together when each participant faced a different reason to wait.

Everyone could agree. Everyone could still wait.

Issuers had to replace hundreds of millions of cards. Merchants had to replace or upgrade terminals. Acquirers and processors had to support new data, certification and processing requirements. Customers had to learn a different checkout behaviour. The costs arrived before the system-wide benefit.

Issuers historically carried much of the cost of counterfeit fraud and therefore stood to gain directly from reducing it. Merchants faced the expense and disruption of upgrading terminals, but an individual merchant could not ensure that customers would arrive with chip cards. Issuing a chip card created limited benefit where terminals could not accept it; installing a chip-enabled terminal created limited benefit while most cards still relied on magnetic stripes.

Each investment became more valuable as others made theirs. The system would benefit if cards, terminals and processing capability moved together, but no individual issuer or merchant could produce that outcome alone. Waiting was not necessarily resistance to better security. It was a rational response to a transition whose value depended on other participants moving too.

The networks changed the consequence of delay

After years of slow movement, the transition accelerated only when the incentives changed. The major payment networks established processing requirements and a common October 2015 liability shift. Historically, issuers had generally borne the cost of counterfeit fraud. After the shift, liability could fall on whichever party to a counterfeit transaction—issuer or merchant—had not implemented chip technology.²

That changed the calculation for each participant. Before the shift, a merchant considering an upgrade weighed an immediate and certain cost against a benefit that depended on widespread adoption elsewhere. After it, continuing to rely on magnetic-stripe technology could expose the merchant to losses it had previously been able to leave elsewhere in the system. The cost of moving had not disappeared. The cost of waiting had changed.

The common timetable mattered too. Issuers could invest with greater confidence that merchants and processors were moving. Merchants could invest with greater confidence that chip cards would arrive. Processors had a defined sequence around which to organise readiness. The networks' position across the payment system allowed them to coordinate timing and alter who would bear the loss if adoption did not occur.

The chip did not suddenly become better, and the collective value did not suddenly become clearer. The route became more viable because the economics of delay changed.

Adoption made the value visible

Once the surrounding conditions changed, use of the chip rose quickly. The share of chip-to-chip transactions among US card-present transactions increased from less than 5 per cent in 2015 to nearly 90 per cent in 2022 for credit and non-prepaid debit cards. Prepaid debit moved more slowly, reaching 60 per cent. By 2025, EMVCo reported that 95.55 per cent of reported US card-present transactions were EMV chip transactions.³

Credit

Less than 5% in 2015→Nearly 90% in 2022

Non-prepaid debit

Less than 5% in 2015→Nearly 90% in 2022

Prepaid debit

Almost zero in 2015→60% in 2022

Redrawn from the published chart; annual positions are approximate. Source: Federal Reserve Bank of Kansas City, using Board of Governors data.

The technology had not merely been issued. It had become the normal way of authenticating an in-person card payment.

The intended value also began to appear. Counterfeit card fraud declined from $3.05 billion in 2015 to $2.62 billion in 2016. By March 2017, Visa reported that counterfeit-fraud value at merchants that had upgraded their terminals was 58 per cent lower than a year earlier; Mastercard reported a similar result.⁴

The results followed the logic of the intervention. As chip adoption accelerated, counterfeit cards became harder to use and merchants that implemented the technology experienced substantial reductions in the fraud it was built to prevent.

The problem moved

Fraud did not disappear. It moved. Once copied cards became harder to use at the point of sale, more fraud appeared in transactions where the physical card was not present.

Channel20152016Change

In person

2015$3.68bn

2016$2.91bn

Changedown 21%

Remote

2015$3.40bn

2016$4.57bn

Changeup 34%

US card-fraud value by transaction channel, 2015-2016. Source: Federal Reserve Payments Study.

The pattern has endured. By 2023, remote transactions represented just over a third of covered US debit-card volume but more than half of fraud losses. Across all cards, US fraud losses reached $14.3 billion. The United States accounted for about a quarter of global card purchases, but more than two-fifths of global card-fraud losses.⁵

The lesson is not that the chip failed. It did what it was designed to do, once the market adopted it. Fraud adapted towards channels with different evidence, authentication and liability arrangements.⁶ The next wave of innovation is forming around those channels—and the question of how to achieve widespread adoption begins again.

A familiar value thesis is forming again

Remote commerce now faces another wave of technical responses. Mobile driving licences (mDLs) can provide cryptographically verifiable identity evidence online and support step-up verification when a transaction is high risk or suspected to be fraudulent.⁷ EMVCo is developing a Digital Payment Credential as a payment-specific application of verifiable digital credentials, intended to support a common and interoperable approach to authentication for online card payments.⁸

These capabilities address different parts of remote-commerce risk. An mDL can provide stronger, cryptographically verifiable evidence about the person and support additional verification for a high-value or suspicious transaction. A Digital Payment Credential is intended to strengthen authentication of the payment credential itself. Together, they could make it harder to use stolen identity or payment data online.

Agentic commerce adds another layer. A transaction may need evidence about the customer, the payment credential, an agent's authority and the relationship between them. Neither an mDL nor a Digital Payment Credential alone establishes whether an AI agent is authorised to act, what that authority covers, whether the transaction remained within it or who must correct the outcome when something goes wrong.

Technical work is progressing across those layers, but the value will depend on whether issuers, merchants, wallets, platforms, payment networks and customers have sufficient reason to adopt, integrate and rely on the new mechanisms—and whether liability moves with practical control.

SignalThe alphabet soup of agentic commerce→Examines the emerging credentials, authority models, commercial positions and obligations around agent-initiated transactions. Reference MapWho is assembling the agentic-commerce market?→Maps the organisations shaping, governing, implementing and building commercial positions around the emerging protocols and credentials.

That is the lesson worth carrying forward from the chip transition. The chip did what it promised. The market moved only when cards, terminals, processing rules and liability moved around it. Mobile driving licences, Digital Payment Credentials and agentic-payment controls are different capabilities, but none escapes that dependency: technical credibility is not the same as a viable route to adoption.

What the case shows

The US chip transition was, in several respects, a relatively tractable coordination challenge. The card market already existed. The participants and transaction model were established. Fraud losses made the value observable. The payment networks had sufficient reach to establish common requirements, coordinate timing and redistribute liability. Even then, technical maturity and credible collective value were not enough.

The case turns on four questions:

• Is the value of the outcome real?
• Which participants must act before it becomes accessible?
• What does each participant gain, fund, risk or become liable for?
• Who can change the conditions that make waiting rational?

A strong aggregate value case answers only the first. The route becomes viable when enough participants have a reason to move, sufficient confidence that others will move and a sequence in which their investments can reinforce one another. In the chip transition, the liability shift supplied a particularly clear lever. It made delay consequential and reduced uncertainty about the behaviour of others.

That does not mean every coordination problem requires a liability shift. It means that explaining the collective benefit is not a substitute for changing participant economics. It also means that success against one risk does not make the surrounding system static: leaders have to watch where incentives, behaviour and exposure move next.

The value can be real while the route to it remains unviable.

SignalAdoption is the value thesis→Develops the wider argument about how participant economics, coordination and authority determine whether a value pool becomes accessible.

The value of the chip was credible before the US market moved. What changed was not the possibility of a more secure transaction but the route through which the market could adopt it. The payment networks altered the consequence of delay; participation began to reinforce participation; and the intended value appeared.

SignalThe position is taken→Examines how present investment decisions can expand, preserve, narrow or foreclose the positions an organisation can occupy later.

The emerging ecosystem has a harder version of the coordination problem. Its value depends on multiple actors moving, but no equivalent coordinating centre is yet visible with sufficient reach to establish a common timetable or change the economics of delay. Each participant can therefore look at the uncertainty around it and conclude that waiting is rational.

But collective adoption is not the only thing moving. Standards, interfaces and commercial positions are already being shaped by the organisations that act. The cold-start problem can delay collective adoption without preserving every participant’s place in the market.

The next wave of payment and identity capabilities will unlock real value—and not only through fraud reduction. That will not, by itself, make the route viable.

Everyone can still wait. But the position may be taken while they do.

Notes

Footnotes

1. “EMV” originally referred to Europay, Mastercard and Visa, which developed the original chip specifications and formed EMVCo. EMVCo now manages the global EMV specifications. Federal Reserve sources support the US adoption and fraud figures used here. Sources: EMVCo — Why EMV; Federal Reserve — 2016 Card Payments Study; Federal Reserve — Changes in US Payments Fraud. ↩ ↩²

2. Visa records that issuers historically bore counterfeit-fraud costs and that, after the October 2015 shift, liability could fall on the party that had not implemented chip technology. The Federal Reserve records the shared date adopted by the major networks. Sources: Visa — US chip-card roadmap; Visa — US chip-card migration and liability; Federal Reserve — 2016 Card Payments Study. ↩

3. The Federal Reserve Bank of Kansas City reports that chip-to-chip transactions rose from less than 5 per cent of US card-present credit and non-prepaid debit transactions in 2015 to nearly 90 per cent in 2022; prepaid debit reached 60 per cent. EMVCo reports that 95.55 per cent of reported US card-present transactions were EMV chip transactions in 2025. Sources: Federal Reserve Bank of Kansas City — EMV migration and card-present fraud; EMVCo — Worldwide EMV deployment statistics. ↩

4. Federal Reserve data support the reported 2015–2016 changes and attribute the movement between channels to both stronger in-person chip protection and increasingly sophisticated cyber-fraud techniques. The CFPB reported Visa's finding that counterfeit fraud at upgraded merchants fell 58 per cent year on year by March 2017, with Mastercard reporting a similar result. Sources: Federal Reserve — Changes in US Payments Fraud; CFPB — Consumer Credit Card Market Report 2017. ↩

5. The Federal Reserve reports that card-not-present transactions represented 34.4 per cent of covered debit-card transaction volume, almost half of value and 54.3 per cent of fraud losses in 2023. The CFPB reports total US card-fraud losses of $14.3 billion in 2023, up from $13.6 billion in 2022, and—citing the Nilson Report—that the United States accounted for 25.3 per cent of global card volume but 42.3 per cent of worldwide card-fraud losses. Sources: Federal Reserve — 2023 debit-card transactions and fraud; CFPB — Consumer Credit Card Market Report 2025. ↩

6. US merchants are generally liable for card-not-present fraud, although authentication through 3-D Secure can shift liability to the issuer. Federal Reserve Bank of Kansas City analysis also found that 3-D Secure has been more narrowly adopted in the United States than in Australia, the United Kingdom and European Union markets. Source: Federal Reserve Bank of Kansas City — Card-not-present fraud rates after migration to chip cards. ↩

7. NIST demonstrates online uses of mDLs for remote identity proofing, authentication and step-up verification for high-value or high-risk transactions. An mDL provides verifiable identity evidence; it is not itself a payment credential or evidence of an agent's authority. Sources: NIST — How mobile driver's licences are changing online; NCCoE — mDL demonstration scenarios. ↩

8. EMVCo describes its Digital Payment Credential as a payment-card-specific application of verifiable digital credentials intended to support interoperable online card authentication. Current work covers credential provisioning, request and verification; payment initiation remains under investigation. Sources: EMVCo — Defining an EMV Digital Payment Credential; EMVCo — Agentic payments and Digital Payment Credentials. ↩