The End of Card Numbers: How Network Tokenization Is Replacing the 16-Digit PAN
Mastercard aims to tokenize all e-commerce by 2030. Why the 16-digit card number is being retired and what replaces it for online payments.
The 16-digit number printed on every payment card is one of the most widely recognized credentials in finance. It is also one of the most exploited. Card-not-present fraud accounts for the majority of payment fraud losses worldwide, and the root cause is structural: a static number that anyone who sees it can reuse. The payments industry is now converging on a fix that looks remarkably similar to how cryptocurrency networks have worked from the beginning: replace static credentials with context-restricted, cryptographically secured tokens.
Mastercard has committed to 100% e-commerce tokenization by 2030. Visa reports that over 60% of its e-commerce transactions are already tokenized. Juniper Research projects that global network-tokenized transactions will nearly double from 283 billion in 2025 to 574 billion by 2029. The 16-digit PAN (Primary Account Number) is being retired as the primary online payment credential, and understanding what replaces it matters for anyone building or operating in the payments space.
What Is Network Tokenization?
Network tokenization replaces a cardholder's PAN with a surrogate value issued by the card network itself (Visa, Mastercard, or another scheme). Unlike merchant-level tokenization, where a payment vault swaps the PAN for an internal reference, network tokens are recognized end-to-end across the entire authorization chain: from the acquirer through the network to the issuing bank.
The framework was standardized by EMVCo's Payment Tokenisation Specification, first published in 2014. The spec defines three core roles: the token service provider (the network or its delegate), the token requestor (a merchant, wallet, or payment processor requesting tokens), and the issuer that approves the token for its cardholder.
How a Network Token Is Created
When a consumer adds a card to a digital wallet, enrolls in Click to Pay, or saves credentials with a merchant that supports network tokenization, the following happens:
- The token requestor sends the PAN to the card network's token service provider (TSP).
- The TSP validates the request with the issuing bank, which may apply its own risk checks or require cardholder authentication via 3D Secure.
- On approval, the TSP generates a token: a surrogate number in the same format as a PAN but mapped one-to-one to the original account, scoped to a specific domain.
- The token is returned to the requestor along with domain restriction controls that limit where and how it can be used.
From that point forward, the token replaces the PAN in all transaction messages. The original card number never touches the merchant's systems.
Domain Restriction: The Security Core
What makes network tokens meaningfully different from the PAN they replace is domain restriction. Each token is bound to a specific context: a particular device, merchant, or channel. A token provisioned for Apple Pay on a specific iPhone cannot be used at a different merchant's website. A token issued for a subscription merchant cannot be used for in-store contactless payments.
This is enforced by the token requestor ID, a unique identifier the network assigns to scope the token's validity. If a token is presented outside its authorized domain, the network rejects the transaction before it ever reaches the issuer.
Per-Transaction Cryptograms
Beyond domain restriction, each tokenized transaction includes a cryptogram: a one-time cryptographic value generated by the token holder (the device or the merchant's token vault) and validated by the network. The cryptogram proves that the entity presenting the token is the entity that was granted the token. Even if an attacker intercepts the token number, it is useless without the ability to generate valid cryptograms for future transactions.
The key insight: A stolen PAN can be used anywhere in the world, indefinitely. A stolen network token is restricted to a single domain, and each transaction requires a fresh cryptogram the attacker cannot produce. This is the same principle behind digital signatures in cryptocurrency: possession of a credential alone is not enough to authorize a transaction.
The Scale of the Shift
Network tokenization is no longer a pilot program. It is the default trajectory for digital payments.
Industry Projections
According to Juniper Research's Global Network Tokenisation Market report, global network-tokenized transactions are projected to grow from 283 billion in 2025 to 574 billion by 2029: a near-doubling over four years. The associated revenue is forecast to grow from $4.1 billion to $8.9 billion, a 117% increase over the same period.
These numbers reflect adoption across all tokenized channels: contactless tap-to-pay, in-app mobile wallets, e-commerce checkout, and recurring billing. The fastest growth is coming from browser-based e-commerce, where Click to Pay and merchant-initiated tokenization are replacing manual card entry.
Network Milestones
| Milestone | Visa | Mastercard |
|---|---|---|
| Tokenization service launch | 2014 (Visa Token Service) | 2014 (MDES) |
| Tokens issued | 10 billion+ (as of June 2024) | Not publicly disclosed |
| E-commerce tokenization rate | 60%+ (as of September 2026) | 25% globally (as of mid-2024) |
| Full tokenization target | No public date set | 100% e-commerce by 2030 |
| Issuer support | 8,000+ issuers across 200+ markets | Broad global coverage |
| Complementary initiatives | Token Management Service, Token ID | Click to Pay, payment passkeys |
Why Authorization Rates Improve
One of the most compelling arguments for network tokenization is not security: it is revenue. Tokenized transactions consistently show higher authorization rates than raw PAN transactions, translating directly into more successful sales for merchants.
The improvement comes from multiple factors. Network tokens are pre-validated by the issuer during provisioning, so the bank has already confirmed the token is legitimate before any transaction occurs. Tokens also solve the stale-credential problem: when a cardholder's physical card is reissued (new expiry date, new CVV), the network automatically updates the token mapping through a process called lifecycle management. Merchants using raw PANs would see those transactions decline; tokenized merchants see uninterrupted approvals.
Measured Improvements
Authorization rate gains vary by network, region, and merchant category. Published figures from industry sources indicate:
- Visa reports a card-not-present authorization uplift of approximately 4.6% for tokenized transactions, with an overall six-basis-point improvement in global payment approval rates
- Mastercard reports a card-not-present uplift of approximately 2.1%
- Visa has attributed $40 billion in incremental e-commerce revenue globally to tokenization, driven primarily by reduced false declines
For a merchant processing $100 million annually in card-not-present volume, a 3% authorization rate improvement represents $3 million in recovered revenue that would otherwise be lost to unnecessary declines.
Fraud Surface Reduction
The fraud case is equally significant. Network tokenization attacks the problem at the credential layer rather than the detection layer. Traditional fraud prevention relies on scoring engines that analyze transaction patterns and flag anomalies. Tokenization makes the stolen credential itself less useful, reducing the attack surface regardless of how sophisticated the fraud operation is.
Visa reported $650 million in fraud prevented over the year preceding its June 2024 milestone announcement. Across industry sources, fraud reduction figures for tokenized transactions range from 28% to 60% compared to non-tokenized equivalents. The variance reflects differences in methodology and baseline, but the direction is consistent: tokenized transactions are substantially harder to exploit.
Why this works: When a merchant's database is breached, attackers get network tokens instead of PANs. Those tokens are domain-restricted (they only work at that specific merchant) and require valid cryptograms (which the attacker cannot generate). A breach that once exposed millions of reusable card numbers now yields data that is functionally useless at any other merchant.
Mastercard's 2030 Roadmap
Mastercard's approach combines three technologies into a unified checkout transformation. The company has publicly committed to achieving 100% e-commerce tokenization by 2030, initially announced for Europe in June 2024 and expanded as a global ambition in early 2025.
The Three Pillars
Mastercard's strategy rests on three mutually reinforcing technologies:
- Network tokenization via MDES (Mastercard Digital Enablement Service) replaces the PAN in all digital transaction flows
- Click to Pay embeds one-click checkout into merchant sites, eliminating manual card entry and storing credentials as tokens rather than raw PANs
- Payment passkeys use device-based biometric authentication (fingerprint, face recognition) to replace passwords and one-time codes, making checkout faster while strengthening authentication
The end state Mastercard envisions: a consumer adds their card once through their issuer's banking app or during checkout. From that point forward, they authenticate with a biometric and the card number is never entered, displayed, or transmitted in cleartext again.
Visa's Token Infrastructure
Visa's approach mirrors Mastercard's direction but emphasizes infrastructure breadth. By its own reporting, Visa had issued over 10 billion tokens as of June 2024, with more than 1.5 million e-commerce merchants transacting with Visa tokens daily. Its FY2026 10-K filing disclosed that over 60% of e-commerce transactions processed by Visa were tokenized as of September 2026.
Visa is extending its token infrastructure into two emerging areas. Identity-based authentication, branded as Token ID, links tokens to verified digital identities rather than just card accounts. Agentic commerce support enables AI agents to hold and present tokens on behalf of consumers, using the same domain restriction and cryptogram framework that secures human-initiated transactions.
Traditional PAN vs. Network Token
The architectural difference between a raw PAN transaction and a tokenized transaction is fundamental, not incremental:
| Property | Raw PAN | Network Token |
|---|---|---|
| Credential type | Static 16-digit number | Surrogate number scoped to domain |
| Reusability if stolen | Unlimited: works at any merchant | Restricted: only valid at the authorized merchant or device |
| Transaction authentication | CVV (static) or 3DS challenge | Per-transaction cryptogram (dynamic) |
| Card reissuance impact | Transactions decline until merchant updates credentials | Automatic lifecycle update: no merchant action required |
| PCI scope for merchant | Full PCI DSS compliance for PAN storage | Reduced PCI scope: tokens are not classified as cardholder data |
| Breach exposure | Compromised PANs are reusable globally | Compromised tokens are functionally useless outside their domain |
The Crypto Parallel: Tokens All the Way Down
The irony of network tokenization is that the card networks are converging on a security model that cryptocurrency pioneered. Bitcoin and other blockchain networks have never used static, reusable credentials. From the beginning, crypto transactions have been authorized through public-key cryptography: a sender proves ownership of funds by producing a digital signature using a private key, without ever revealing the key itself.
The structural parallels are striking:
| Concept | Network Tokenization | Crypto (Bitcoin / Stablecoins) |
|---|---|---|
| Static credential replaced | PAN → domain-restricted token | No static credential exists: public key is not a secret |
| Transaction authorization | Per-transaction cryptogram | Per-transaction digital signature |
| Credential reuse attack | Mitigated by domain restriction | Impossible: each UTXO can only be spent once |
| Key compromise impact | Token is scoped: limited blast radius | Private key controls all funds: self-custody tradeoff |
| Intermediary requirement | Network acts as TSP and arbiter | Protocol validates: no intermediary needed |
| Settlement finality | T+1 or T+2 via batch clearing | Seconds to minutes via on-chain settlement |
The card networks are retrofitting their decades-old infrastructure with cryptographic authorization, adding layers of indirection (tokens, TSPs, domain controls) to approximate what crypto achieves natively through asymmetric cryptography. The result works, but it requires coordination among issuers, networks, processors, and merchants to function. Crypto's approach eliminates most of these intermediaries by design.
What This Means for Stored Credentials
The transition to network tokenization fundamentally changes how merchants handle stored payment credentials. Today, millions of merchants store raw PANs in PCI-compliant vaults for recurring billing, one-click checkout, and subscription management. Under full tokenization, these stored PANs are replaced by network tokens that benefit from automatic lifecycle updates and reduced breach exposure.
For merchants, this is largely positive: reduced PCI scope, lower compliance costs, fewer declines from expired cards, and reduced fraud liability. The transition cost is real but manageable: payment processors and gateways are increasingly handling tokenization transparently, so merchants may not need to change their integration at all.
For the payments ecosystem more broadly, the shift raises questions about who controls the token infrastructure. Network tokens are issued and managed by the card schemes themselves. This deepens merchants' dependency on Visa and Mastercard, which already set interchange and scheme fee structures that merchants have limited ability to negotiate.
Infrastructure Dependencies and Risks
Network tokenization is not without challenges. Full adoption requires every participant in the payment chain to support tokens:
- Issuers must register with the token service provider and approve token provisioning requests in real time
- Acquirers and processors must route token-based transactions correctly, including passing cryptograms and domain data
- Merchants must integrate with token requestor APIs or use processors that handle tokenization on their behalf
- Regional adoption is uneven: one industry analysis found that only 20% to 30% of European issuers currently support network tokens, compared to broader coverage in the United States
There is also the question of fallback. During the transition period, merchants must support both tokenized and non-tokenized flows. The dual infrastructure adds complexity, particularly for payment orchestration platforms that route transactions across multiple processors and networks.
Where Stablecoin Payments Fit
Traditional payments are converging toward the token-based, cryptographic authorization model that crypto pioneered. But the convergence is asymmetric: card networks are layering cryptographic security onto a system that still depends on batch settlement, intermediary coordination, and scheme-controlled infrastructure. Stablecoin payments start from cryptographic-first principles and add the stability of dollar-denominated value.
On Spark, stablecoin payments use public-key cryptography and FROST threshold signatures for the same security guarantees that card networks are now retrofitting: every transaction is cryptographically authorized, credentials cannot be reused, and settlement is instant rather than batched over days. The difference is that these properties are inherent to the protocol, not bolted on after decades of operating with static credentials.
For developers building payment infrastructure, both models are worth understanding. Network tokenization and crypto-native payments are solving the same fundamental problem: making digital payments secure without relying on secret, reusable credentials. The approaches differ in architecture and tradeoffs, but the direction is the same. Wallets like General Bread, built on Spark, demonstrate what payments look like when cryptographic authorization is the foundation rather than an afterthought. Developers interested in building on this model can explore the Spark SDK documentation.
The Road Ahead
The 16-digit card number is not disappearing overnight. Physical cards will continue to carry embossed or printed PANs for the foreseeable future, and many merchants and regions are years away from full tokenization support. But the direction is clear: static credentials are a liability, and every major payment network is investing heavily in replacing them with cryptographically secured, domain-restricted tokens.
By 2030, if Mastercard's target holds, manually entering a card number for an online purchase may feel as outdated as writing a check. The question for the broader payments industry is whether the right architecture is one that layers cryptographic security onto legacy rails or one that starts from cryptographic principles and builds the payment experience on top. Both approaches are advancing simultaneously, and the next five years will determine which model captures the most volume.
This article is for educational purposes only. It does not constitute financial or investment advice. Bitcoin and Layer 2 protocols involve technical and financial risk. Always do your own research and understand the tradeoffs before using any protocol.

